PDA

View Full Version : NET News (January 28): Who would the No. 1 seeds be based on blind resumes?



scottdude8
01-28-2019, 12:12 PM
There hasn't been too much to shake up the NET rankings this week beyond two major results: Kentucky's win over Kansas and Michigan State's upset loss to Purdue (I told you all it was coming, haha). Thus, instead of doing the NET breakdown like I've been doing over the past few weeks I figured it might be a good time to do a "blind resume" exercise to see if how we perceive the strength of the top teams corresponds with the data provided on the NCAA Team Sheets, which are primarily driven by the NET rankings instead of the RPI now.

Below, I've made a table with the resumes of the seven teams that I think the consensus is are reasonably in the running for No. 1 seeds (note that I'm moving to seven teams despite talking about a "top-6" recently... you may have an inkling as to why, but I won't spoil that to maintain the anonymity of the exercise!). All of the data is taken directly via the NET Rankings and the Team Sheets through games played on January 27, meaning that if the season ended today, these would be the primary data points provided to the committee. When I list wins/losses, I'll again keep things anonymous by listing them as something like "H19", which would be a home win (or loss) against the team ranked 19 in the NET. Obviously you may be able to suss out who is who, but try not to spoil it for everyone, haha.

Based off of this, I'm curious to see what people think about not only who the 4 No. 1 seeds should be, but also who the No. 1 overall seed should be. I'll put the later question as a "poll" (since there are 35 possible outcomes of 7 choose 4 for those mathematically inclined out there, haha), so keep your responses about the 4 No. 1 seeds to replies. Sometime later today I'll try to compile everyone's thoughts and reveal who is who and see if anything interesting has come about from looking at team resume's not through the lens of the big names, but instead through the NET data. Enjoy!



Team
A
B
C
D
E
F
G


Record
17-2
16-3
18-1
18-3
19-2
19-1
17-1


NET Rank
3
8
1
6
2
4
5


KPI Rank
2
10
1
3
4
5
6


"Q1" Record
6-1
5-2
4-1
8-3
3-2
5-1
4-1


"Q2" Record
3-1
2-1
6-0
4-0
2-0
5-0
4-0


Q1 Wins
H1, N8, N18, A33, N26, A63
N9, A12, A26, H16, H27
H10, N13, A22, A59
H11, A25, A30, A36, A37, H22, H30, N43
N3, A47, A58
H9, H11, A21, A41, A56
N2, N12, N16, A36, A54


Losses
N2, H48
N3, A44, N64
A3
A11, A12, N16
N5, A9
A13
N16


Non-Conference SOS
8
38
222
113
67
153
30


Overall SOS
6
33
67
24
58
64
39

uh_no
01-28-2019, 12:32 PM
A c d g

BandAlum83
01-28-2019, 12:34 PM
A good example of why the blind resume should not always be the best way to go. A clear case may be made for D, with A as #2. The "D" losses coming on the road to 2 top 12 teams, and neutral to another top 20, doesn't look so bad against all of the quality wins. The out of conference SOS is pathetic, but overall 24 mitigates that some.

Not sure why you added the #10 Net team. You thought we could guess, but I for one can't guess why, other than it's a team we would NEVER think belongs that high,

Acymetric
01-28-2019, 12:36 PM
There hasn't been too much to shake up the NET rankings this week beyond two major results: Kentucky's win over Kansas and Michigan State's upset loss to Purdue (I told you all it was coming, haha). Thus, instead of doing the NET breakdown like I've been doing over the past few weeks I figured it might be a good time to do a "blind resume" exercise to see if how we perceive the strength of the top teams corresponds with the data provided on the NCAA Team Sheets, which are primarily driven by the NET rankings instead of the RPI now.

Below, I've made a table with the resumes of the seven teams that I think the consensus is are reasonably in the running for No. 1 seeds (note that I'm moving to seven teams despite talking about a "top-6" recently... you may have an inkling as to why, but I won't spoil that to maintain the anonymity of the exercise!). All of the data is taken directly via the NET Rankings and the Team Sheets through games played on January 27, meaning that if the season ended today, these would be the primary data points provided to the committee. When I list wins/losses, I'll again keep things anonymous by listing them as something like "H19", which would be a home win (or loss) against the team ranked 19 in the NET. Obviously you may be able to suss out who is who, but try not to spoil it for everyone, haha.

Based off of this, I'm curious to see what people think about not only who the 4 No. 1 seeds should be, but also who the No. 1 overall seed should be. I'll put the later question as a "poll" (since there are 35 possible outcomes of 7 choose 4 for those mathematically inclined out there, haha), so keep your responses about the 4 No. 1 seeds to replies. Sometime later today I'll try to compile everyone's thoughts and reveal who is who and see if anything interesting has come about from looking at team resume's not through the lens of the big names, but instead through the NET data. Enjoy!

Clear #1 seeds: A, D

#1 seeds (marginal): G, F

In the hunt: C, E

Not in contention: B

Overall #1: A, but D is very close.

Breakdown:

A & D have a ton of good Q1 wins, and the two best SOS.

G & F have fewer losses, but also fewer big wins and so are a tier lower but still #1 seeds.

C & E have only lost to really good teams, but don't have enough big wins to overtake G & F while having more losses.

B has two of the three worst losses and are thus disqualified.

scottdude8
01-28-2019, 12:44 PM
A good example of why the blind resume should not always be the best way to go. A clear case may be made for D, with A as #2. The "D" losses coming on the road to 2 top 12 teams, and neutral to another top 20, doesn't look so bad against all of the quality wins. The out of conference SOS is pathetic, but overall 24 mitigates that some.

Not sure why you added the #10 Net team. You thought we could guess, but I for one can't guess why, other than it's a team we would NEVER think belongs that high,

I think you got the NET and KPI mixed up, because the team I "added" is No. 8 in the NET and No. 10 in the KPI.

Also, I'm tempted to post who I'd choose, but that might make it a little more obvious who is who in the anonymous table. So I'll save that (with an explanation, of course, haha) for later :)

JohnJ
01-28-2019, 01:37 PM
Is anyone else only seeing four teams listed in the table?

scottdude8
01-28-2019, 01:40 PM
Is anyone else only seeing four teams listed in the table?

8989

This is what I see... I hope I didn't make a mistake in creating the table! Are you perhaps viewing via the mobile browser?

Ian
01-28-2019, 01:43 PM
For me it's A C D F

Acymetric
01-28-2019, 01:44 PM
For what it's worth, if I ran the numbers again in three hours I would probably come up with entirely different results.

uh_no
01-28-2019, 01:44 PM
For me it's A C D F

the NCSOS turned me off from F. Not that either the arguments or meta-arguments matter here, but I'd rather reward someone with 1 fewer top win but who actually scheduled people when they didn't have to.

december basketball makes me vomit.

JohnJ
01-28-2019, 01:44 PM
8989

This is what I see... I hope I didn't make a mistake in creating the table! Are you perhaps viewing via the mobile browser?

Thanks for a adding this. Yes, viewing with my iPad.

Acymetric
01-28-2019, 01:48 PM
the NCSOS turned me off from F. Not that either the arguments or meta-arguments matter here, but I'd rather reward someone with 1 fewer top win but who actually scheduled people when they didn't have to.

december basketball makes me vomit.

See, I'm the opposite. Well, I mean I much prefer teams play better teams so that we have better games, I agree with you about December basketball. I don't really like making that part of seeding criteria though. To me, overall SOS is best used as a tiebreaker after considering quality of wins vs. quality of losses.

scottdude8
01-28-2019, 01:56 PM
the NCSOS turned me off from F. Not that either the arguments or meta-arguments matter here, but I'd rather reward someone with 1 fewer top win but who actually scheduled people when they didn't have to.

december basketball makes me vomit.

The discussion of SOS is interesting, and brings up one of the things that I'm sure will be hotly debated when it comes to seeding: with regards to non-conference scheduling what matters more, the overall SOS or the "top-tier" SOS (and corresponding ability to get top-tier victories)? One of the biggest flaws in the RPI was that your RPI could be artificially inflated if you chose to play a bunch of middling RPI teams that you were still definitely going to beat in your non-conference schedule while never playing any marquee, lose-able games... in contrast, a team could play a handful of solid opponents from power conferences, but have their RPI weighed down by a couple of games against teams with sub-300 RPIs. To me the NET has done a better job of correcting for this (it isn't by any means perfect, but it's an improvement) in its overall rankings, but the question itself still poses itself on the team sheet with how different committee members will weigh the overall NC SOS vs specific NC wins or losses.

Also, as a more specific sidebar... if the SOS turned you off from F, why wouldn't it also turn you off from C, which has a significantly worse NC SOS (in fact, it's the worst amongst the 7 teams listed here) as well as a slightly worse overall SOS? Just curious.

Nugget
01-28-2019, 02:19 PM
For me it's A C D F

Me, too.

BandAlum83
01-28-2019, 02:28 PM
I think you got the NET and KPI mixed up, because the team I "added" is No. 8 in the NET and No. 10 in the KPI.

Also, I'm tempted to post who I'd choose, but that might make it a little more obvious who is who in the anonymous table. So I'll save that (with an explanation, of course, haha) for later :)

Gotcha, and now I know why you added them. They are playing really well right now, but perhaps against not-so-tough competition.

Nugget
01-28-2019, 02:29 PM
The discussion of SOS is interesting, and brings up one of the things that I'm sure will be hotly debated when it comes to seeding: with regards to non-conference scheduling what matters more, the overall SOS or the "top-tier" SOS (and corresponding ability to get top-tier victories)? One of the biggest flaws in the RPI was that your RPI could be artificially inflated if you chose to play a bunch of middling RPI teams that you were still definitely going to beat in your non-conference schedule while never playing any marquee, lose-able games... in contrast, a team could play a handful of solid opponents from power conferences, but have their RPI weighed down by a couple of games against teams with sub-300 RPIs. To me the NET has done a better job of correcting for this (it isn't by any means perfect, but it's an improvement) in its overall rankings, but the question itself still poses itself on the team sheet with how different committee members will weigh the overall NC SOS vs specific NC wins or losses.

I don't think historically that non-conference SOS (or SOS at all for that matter) has historically been a significant factor in seeding of the teams at the top of the bracket, other than the oddities of a Gonzaga or Nevada or Wichita St. outlier from a conference that doesn't provide opportunities against top teams.

When comparing, for instance, a 28-3 Virginia against a 27-4 Tennessee and a 26-5 Kansas, I don't recall their respective non-con SOSs weighing into the discussion much. I think the number of Q1/Q2 wins and the number of "bad" losses (plus conference championships) has really driven the decisions at the top of the bracket

Non-con SOS has typically been much more of a discussion point for the debate between the last few bubble teams -- and the Committee has often cited that as a factor for why a team gets left out. But, I'm hard put to recall any example of the Committee citing non-con SOS as a factor in choosing between major conference teams for #1 seeds.

ikiru36
01-28-2019, 02:36 PM
Fun exercise. For me, ACGD in that order.

One thing that the committee is allowed to take into account (that is not on the table) is injuries, which I'm pretty sure played a role in Team A's most troublesome loss! Honestly, without knowing about that injury impact (or the head-to-head matchup, including injury), I'd probably make C the top seed, by a smidge.

To me, G and D raise the interesting comparative issue of relying too much on the tier-ing without looking at the quality of wins/loses within the tier. Of course, team D's eight Tier 1 wins seems really impressive at first until I notice that their best win was #11 and their second best win was #22. Whereas Team G has only four Tier 1 wins, but that includes wins against #2, #12, and #16. And in terms of the scheduling they can control, Team D played the 30th most difficult non-conference schedule.

I guess that if I could add any data category (other than injuries to significant players during games lost), some kind of mean victory margin against Tier 1 (or something like that) could be important for differentiation as well? Thanks for doing this!

FerryFor50
01-28-2019, 03:09 PM
I think team G is missing a win...

Ian
01-28-2019, 03:43 PM
the NCSOS turned me off from F. Not that either the arguments or meta-arguments matter here, but I'd rather reward someone with 1 fewer top win but who actually scheduled people when they didn't have to.

december basketball makes me vomit.

Yeah, the the SOS to me is a phony issue. F played 11 Q1+2 games, G played 9. So the SOS different is made entirely of team G playing more Q3 vs team F playing more Q4 games. And so what, for a team vying for a #1 seed, there is essentially no difference between Q3 and Q4, as neither group poses a serious threat for a win. I don't consider playing more Q3 teams instead of Q4 to be "playing a tougher schdule" when evaluating the SOS for a #1 seed.

uh_no
01-28-2019, 03:45 PM
Yeah, the the SOS to me is a phony issue. F played 11 Q1+2 games, G played 9. So the SOS different is made entirely of team G playing more Q3 vs team F playing more Q4 games. And so what, for a team vying for a #1 seed, there is essentially no difference between Q3 and Q4, as neither group poses a serious threat for a win. I don't consider playing more Q3 teams instead of Q4 to be "playing a tougher schdule" when evaluating the SOS for a #1 seed.

fair enough...i didn't look tooo deeply into how the SOS was calculated (and am generally an opponent of blind looking at SOS, since not all SOS are created equal). It would be nice to have some metric for games played within a range of ones own .NET or something.

scottdude8
01-28-2019, 05:30 PM
I think team G is missing a win...

Darn, I knew there would end up being a mistake there somewhere... and I canít edit the original post anymore :(

scottdude8
01-29-2019, 10:52 AM
So here's who each of the teams were (I know many of you probably sussed it out on your own):
A=Duke
B=Kentucky
C=Virginia
D=Michigan State
E=Gonzaga
F=Michigan
G=Tennessee

As far as the No. 1 overall seed, a vast majority of people picked us... with MSU in second, Virginia in third, and one vote for Michigan (which wasn't me surprisingly, haha).

When it came to who the No. 1 seeds were, almost everyone who participated said Duke, Virginia and Tennessee would be on the top line (which corresponds with the general consensus from Bracketologists, it seems, so nice job!). The fourth No. 1 seed was a bit more of a debate, with Michigan and Michigan State being the most picked teams (some people actually had both of those as No. 1 seeds instead of Tennessee, which is an interesting scenario).

For those who might be interested, when I was looking through things (trying my best to keep things "blind", but I was also looking at the Team Sheets) I thought Duke, Tennessee, Michigan, and Virginia should be the top seeds (in that order), with MSU the next closest. How did I come to this decision? Well:

Kentucky was the first eliminated. While they have the potential to jump up to the 1 line with the way they're playing and two matchups against Tennessee left to go, obviously they aren't there yet. I included them just to show that they've quickly turned their season around to even put themselves back in the running for a top seed. I would suspect that at most one SEC team will get a No. 1 seed, and it could very well come down to the UK-Tennessee matchups.
Gonzaga is also eliminated because they just don't have enough top-tier victories. A one-loss Gonzaga team might be in the running, but two losses with only 5 total Q1/Q2 wins is pretty tough to overcome. Gonzaga only has a shot at a No. 1 seed at this point, in my analysis, if a lot of top teams falter considerably.
Among the remaining 5 teams, Duke clearly differentiates itself based on overall SOS, Q1 victories, and the relative quality of those Q1 victories. Duke wins in Q1 resume both in quality and quantity, which overcomes the extra loss.
Among the four remaining teams, MSU has the most Q1 wins. However, I eliminated MSU based on two factors: first, the relative quality of those Q1 wins (their most impressive win came at home, and all of their away/neutral court wins were amounted to "lower tier" Q1 wins), and the fact that you can't discount two additional losses. I recognize that this is not the prevailing opinion on this board or amongst "bracketologists", but I think that losses have to count for something. IMHO for a team with two more losses than another team from a power conference to overtake them on the seeding line, their resume would have to be clearly superior. While you can argue that MSU has the best resume based on Q1 wins, I don't think that resume is clearly superior to any of the 1 loss teams. The closest team is probably Michigan, but considering they're in the same conference, with the same record, and Michigan had a superior NC resume, I don't see how you choose MSU over Michigan.

All that said, a lot of this will sort itself out, particularly Michigan vs. MSU and Kentucky vs. Kansas. I think there's a scenario in which the SEC cannibalizes itself (i.e. Kentucky sweeps Tennessee and shows up for those big games, but also drops a couple more bad upsets as they're prone to do) and leaves the 1 line open just for the ACC and B1G. But for me, the most interesting part of this exercise was how much people liked MSU's resume despite having the most losses on the board... I found it very surprising that not only were they consistently chosen as a No. 1 seed, but that many thought they deserved the OVERALL No. 1 seed despite 3 losses. Again, I may be in the minority but I think losses have to count for something. It'll be interesting to see which way the committee goes come Selection Sunday if a choice like this presents itself!

jv001
01-29-2019, 11:24 AM
Thanks for taking the time to put this thread/chart together. I didn't know Virginia's SOS was that bad with 222 out of conference rating. The conference schedule helped get them into the 60s. It will be good to see how this shapes up in the 2nd half schedules. As some have said, TN and the Mildcats can separate themselves with a sweep. The same can be said of Duke-VA. If the Hoos win in Charlottesville, it could come down to the ACCT. No matter what happens, I love this Duke team. GoDuke!

Acymetric
01-29-2019, 11:47 AM
When it came to who the No. 1 seeds were, almost everyone who participated said Duke, Virginia and Tennessee would be on the top line (which corresponds with the general consensus from Bracketologists, it seems, so nice job!). The fourth No. 1 seed was a bit more of a debate, with Michigan and Michigan State being the most picked teams (some people actually had both of those as No. 1 seeds instead of Tennessee, which is an interesting scenario).


Interestingly, I had Michigan and Michigan State in over Virginia rather than Tennessee. It is hard for me to remember exactly why, and I'm not sure I agree with myself.

I originally had Duke(A), Michigan State (D), Tennessee (G), Michigan (F) in that order.

In hindsight, it should be Duke (A), Michigan State (D), Tennessee (G), Virginia (C). Michigan's SOS takes them just barely out, and I'm not sure why I edged them over UVA (although I noted in my originally post that C was very close to F, so I was undecided when I made the decision as well). I think I might have drifted columns and attributed Kentucky's losses to UVA when doing my analysis on the C team.

scottdude8
01-29-2019, 12:43 PM
Interestingly, I had Michigan and Michigan State in over Virginia rather than Tennessee. It is hard for me to remember exactly why, and I'm not sure I agree with myself.

I originally had Duke(A), Michigan State (D), Tennessee (G), Michigan (F) in that order.

In hindsight, it should be Duke (A), Michigan State (D), Tennessee (G), Virginia (C). Michigan's SOS takes them just barely out, and I'm not sure why I edged them over UVA (although I noted in my originally post that C was very close to F, so I was undecided when I made the decision as well). I think I might have drifted columns and attributed Kentucky's losses to UVA when doing my analysis on the C team.

I think what you just outlined is a very interesting example of how blind vs. non-blind can influence things. I would argue (and I think the numbers back me up) that Michigan has a superior resume over Virginia in every aspect except the actual NET and KPI rankings, and perhaps maybe the "eye test". As much as Michigan's SOS might hurt them, Virginia's is significantly worse, especially in the non-conference. Michigan also has two "marquee" NC victories (at Villanova and home against UNC) while UVA's best NC win was against Wisconsin on a neutral site. But UVA plays in the ACC and has been talked up in the media as a contender a lot more than Michigan. So I think the fact that you didn't have Virginia as a 1-seed when things were blind (even though it was close) but afterwards you do is a very interesting development... and again, something interesting to consider when we think about whether or not this will impact the committee.

scottdude8
01-29-2019, 12:44 PM
Thanks for taking the time to put this thread/chart together. I didn't know Virginia's SOS was that bad with 222 out of conference rating. The conference schedule helped get them into the 60s. It will be good to see how this shapes up in the 2nd half schedules. As some have said, TN and the Mildcats can separate themselves with a sweep. The same can be said of Duke-VA. If the Hoos win in Charlottesville, it could come down to the ACCT. No matter what happens, I love this Duke team. GoDuke!

Appreciate it! Putting that chart together took more time than I'd like to admit, haha, but I think it was a very interesting exercise. Maybe I'll have to do it again in a month or so when things start to shake themselves out!

English
01-29-2019, 04:09 PM
I think what you just outlined is a very interesting example of how blind vs. non-blind can influence things. I would argue (and I think the numbers back me up) that Michigan has a superior resume over Virginia in every aspect except the actual NET and KPI rankings, and perhaps maybe the "eye test". As much as Michigan's SOS might hurt them, Virginia's is significantly worse, especially in the non-conference. Michigan also has two "marquee" NC victories (at Villanova and home against UNC) while UVA's best NC win was against Wisconsin on a neutral site. But UVA plays in the ACC and has been talked up in the media as a contender a lot more than Michigan. So I think the fact that you didn't have Virginia as a 1-seed when things were blind (even though it was close) but afterwards you do is a very interesting development... and again, something interesting to consider when we think about whether or not this will impact the committee.

Spoiler alert: it will impact the committee. As long as the committee has human people on it, and the team sheets are NOT blind, the names of the schools will have a tangible impact. Perhaps not quantifiable, but it'll be there. And, it'll not be uniform across the membership, just as things like conference regular season and tournament champs will have an uneven impact across membership. And good wins vs bad losses, and home vs neutral vs away. And how much value is ascribed to KPI and NET vs the eye test.

This was a really thoughtful exercise, thanks for setting it up.

scottdude8
01-30-2019, 11:35 AM
Spoiler alert: it will impact the committee. As long as the committee has human people on it, and the team sheets are NOT blind, the names of the schools will have a tangible impact. Perhaps not quantifiable, but it'll be there. And, it'll not be uniform across the membership, just as things like conference regular season and tournament champs will have an uneven impact across membership. And good wins vs bad losses, and home vs neutral vs away. And how much value is ascribed to KPI and NET vs the eye test.

This was a really thoughtful exercise, thanks for setting it up.

Good point. It's so hard to suss out what's actually important when it comes to seeding because you get so many mixed messages from the NCAA and the committee, both before and after Selection Sunday. Obviously looking purely at the numbers (i.e. keeping things blind) isn't ideal, because via the eye-test we know things like Houston is overrated in the NET (for example)... but I'm of the mind that once you use that external information to sort of split teams into general tiers, the committee has an obligation to keep things as blind as possible so that they're truly rewarding resume and not name. How you accomplish that I'm not sure (as with most issues with the NCAA, it's straightforward to diagnose the problem, but not so straightforward to cure it, haha), but I think the easiest step forward would be more transparency... transparency in terms of how the NET is calculated, transparency in terms of what the actual directives to the committee members are, and maybe even transparency in the thought process on particular examples in seeding or bubble teams. Who knows if that'll ever happen. For now we'll have to satisfy ourselves with being armchair bracketologists I guess!

OldPhiKap
02-05-2019, 07:42 AM
NET changing end-of-game strategy?

https://twitter.com/cardchronicle/status/1092639683731968001?s=21

HereBeforeCoachK
02-05-2019, 07:52 AM
NET changing end-of-game strategy?

https://twitter.com/cardchronicle/status/1092639683731968001?s=21

This is typical NCAA stupidity....that leads to unintended consequences...design a system where a 1 point loss and a 9 point loss are equal...yet a 9 point loss and a ten point loss are totally different things.

What could possibly go wrong?

ice-9
02-05-2019, 10:29 AM
I wouldíve put C as either first or second in the pecking order ó hard to be objective since it was obvious Duke was A.

I find it interesting that people would look at the overall rank, and then qualify it by looking at other metrics like SOS which was already captured in the calculation of said rank! Itís double counting in short.

My view is to go with the overall rankings as the North Star, and if you disagree with itís methodology then go ahead make another, but donít mix the algorithm with other variables. It might seem logical and more robust but instead all youíre doing is misrepresenting intuition as data.

I donít think I can come up with something better than NET or KP so I wouldíve put C as the top team.

English
02-05-2019, 10:37 AM
The latest Jordan Sperber pod "Solving Basketball" (guest Jeff Goodman) again addresses the NET--some (relevant) cliffs notes:

- They cover the NCSU issue again of scheduling a laughably garbage NCSOS and pummeling all the sub-200 KP teams--Goodman has been, perhaps, the most vocal critic of this strategy and State for executing it; Sperber is more thoughtful about it, and explains that, when you analyze historically comparable situations where a team massacred a similarly egregious schedule, those teams ended up being pretty good (he used the post-NCAAT Champs* Louisville team as a comp). Although he doesn't totally editorialize about the merits of the strategy or the likelihood that other teams will see how it's fared for State and replicate it going forward, he doesn't fully agree with Goodman that this is a flaw of the NET.

- They cover the temptation to run up the score as a means of padding unadjusted efficiency, and thus, NET ranking. Sperber seems very confident that the NCAA will revisit unadjusted efficiency as a component because of its impact. He does mention something that struck me--he said that if they remove or tweak the efficiency component of the NET, the likelihood of getting outliers WRT KenPom goes up. This makes sense, obviously, but I hadn't considered it too much in those terms. Something tells me that the further astray from KenPom that a future NET might end up, the more head scratching there would be around here and among the analysts.

- Sperber explains that one of the things he finds most lacking of the NET, in addition to the opaqueness of their formula (which, like myself and many here, he speculates will take care of itself with more data and reverse engineering the rankings), is the NCAA defining what the intent of the NET is--we've alluded to this before here, but is the goal to create a ranking system that selects and seeds the teams most inline to what their future performance will be (i.e., predictive, a la KenPom)? Or, conversely, is it to reward/punish teams for their results and performance throughout the season. The two men agree that the latter makes more sense, so including such a (seemingly high) premium on efficiency is curious.

- To conclude, Goodman asks Sperber to give him a recommendation of an advanced stat that he would find useful but probably doesn't already know about--Sperber (and he says he and John Gasaway have spoken about this, and both find it enlightening) cites the "Wins Above Bubble" stat. I won't go into the details, mostly because I haven't delved into the formula too deeply at this point, but I plan to check it out in the next week or two.

There are a bunch of other interesting nuggets in there. Anyway, as always, I recommend a listen for folks interested in an analytical discussion of hoops.

ETA:
- Sperber also notes that coaches have, almost as soon as the NET was announced, asked him point blank if it was in their interests to run up the score. He said initially it was unclear, but that isn't the case anymore. There is an unambiguous interest in running up the score (although, as he mentioned in an earlier pod, there is some consideration in running up the score too much to the detriment of your opponent in the quad system). He emphasized that, in fact, the coaches are very much paying attention to the NET and are adjusting strategy (as you would guess) to optimize their teams' ranking.

- Sperber also mentions that, while he doesn't think the NET is the bee's knees (my phrase, not his), he says as long as the NCAA is using the quad system, the NET is a good--and certainly worlds better than RPI--metric. He explicitly says he doesn't like the quad system and wishes the NCAA would toss it.

Ian
02-05-2019, 11:01 AM
When coaches are making salaries in the millions/yr, and when keeping or getting such a job depends on the making the tournament, how can you expect them to not claw and grasp for every advantage possible. Of course they are going to game the selection criteria when to comes to running up the score and playing to the +10 cap rule. Anyone who expected different are fooling themselves.

Hingeknocker
02-05-2019, 11:17 AM
- Sperber explains that one of the things he finds most lacking of the NET, in addition to the opaqueness of their formula (which, like myself and many here, he speculates will take care of itself with more data and reverse engineering the rankings), is the NCAA defining what the intent of the NET is--we've alluded to this before here, but is the goal to create a ranking system that selects and seeds the teams most inline to what their future performance will be (i.e., predictive, a la KenPom)? Or, conversely, is it to reward/punish teams for their results and performance throughout the season. The two men agree that the latter makes more sense, so including such a (seemingly high) premium on efficiency is curious.


I may have to listen to this whole section, because I actually think there's a reasonable compromise to be had between efficiency (NET, KenPom, etc) and results-based evaluations. Selecting the teams for the tournament should be based on a host of factors, including your black-and-white wins and losses, that efficiency metrics aren't designed to include. As a hypothetical, it would be possible for a team to go 10-20 on the season, but end up with a relatively high KenPom rating; perhaps they played a very hard schedule, and all of their 20 losses came by 1 point. But, obviously, this team should not make the tournament.

However, seeding the teams absolutely, without a doubt, should be done via the efficiency metrics. The predictive metrics are all that matter once the tournament has its 68 teams. If you're a team who earns a 1 seed under this metric, you deserve to play a team truly of the caliber of an 8/9 seed in the 2nd round, 4/5 in the S16, etc. etc. I would be perfectly happy if the tournament seeds were a straight S-curve of the KenPom rankings at the end of the year (with the usual bracketing rules in place, of course).

All in all, very intelligent observation and conversation about the systems. Thank you for the recap!

Troublemaker
02-05-2019, 12:53 PM
I may have to listen to this whole section, because I actually think there's a reasonable compromise to be had between efficiency (NET, KenPom, etc) and results-based evaluations. Selecting the teams for the tournament should be based on a host of factors, including your black-and-white wins and losses, that efficiency metrics aren't designed to include. As a hypothetical, it would be possible for a team to go 10-20 on the season, but end up with a relatively high KenPom rating; perhaps they played a very hard schedule, and all of their 20 losses came by 1 point. But, obviously, this team should not make the tournament.

However, seeding the teams absolutely, without a doubt, should be done via the efficiency metrics. The predictive metrics are all that matter once the tournament has its 68 teams. If you're a team who earns a 1 seed under this metric, you deserve to play a team truly of the caliber of an 8/9 seed in the 2nd round, 4/5 in the S16, etc. etc. I would be perfectly happy if the tournament seeds were a straight S-curve of the KenPom rankings at the end of the year (with the usual bracketing rules in place, of course).

All in all, very intelligent observation and conversation about the systems. Thank you for the recap!

I (mostly) disagree here. I think resume should still matter for seeding. For example, if Duke wins by a small margin at UVA (not an unlikely outcome), UVA will probably stay ahead of us in kenpom despite our sweep of them. But Duke would obviously deserve the overall #1 seed and the Washington D.C. bracket ahead of the Hoos at that point.

I think it makes sense to use predictive analytics to seed, say, the 13 thru 16 seeds. I don't trust the Committee to have watched those teams enough or to even compare their resumes sensibly. It stinks when a 1 seed has to face a 16 seed that should've been a 14 seed based on kenpom, for example, and likewise, if a 14 seed should've been a 16 seed, then the 3 seed it is paired with gets an undeserved break.

Hingeknocker
02-05-2019, 01:53 PM
I (mostly) disagree here. I think resume should still matter for seeding. For example, if Duke wins by a small margin at UVA (not an unlikely outcome), UVA will probably stay ahead of us in kenpom despite our sweep of them. But Duke would obviously deserve the overall #1 seed and the Washington D.C. bracket ahead of the Hoos at that point.

I think it makes sense to use predictive analytics to seed, say, the 13 thru 16 seeds. I don't trust the Committee to have watched those teams enough or to even compare their resumes sensibly. It stinks when a 1 seed has to face a 16 seed that should've been a 14 seed based on kenpom, for example, and likewise, if a 14 seed should've been a 16 seed, then the 3 seed it is paired with gets an undeserved break.

I agree, it would feel weird if Duke beats UVA twice but stayed behind them in KenPom, when evaluating the traditional way overall #1 seeds are awarded. My counter would just be, is it actually so obvious that Duke deserves the #1 seed instead of UVA in that scenario? The reason UVA had such a huge lead in KenPom (up until the last 2 games) is that they were absolutely destroying people. It might actually be the case that UVA is a better team, even with two losses to Duke.

But the real reason I'd advocate using efficiency ratings to seed the entire bracket is that I don't trust the Committee to do anything right, not just those 13-16 seeds. Up and down the bracket, the tournament is mis-seeded. Just last year, Xavier was an abomination as a 1 seed. Butler was brutally underseeded as a 10 (#20 in KenPom). It's unfair to everyone when teams are mis-seeeded: the teams directly involved, their future opponents elsewhere in the bracket, etc.

And if it makes us feel better, we could use the Massey Composite ratings, which currently have Duke at #1. :)

HereBeforeCoachK
02-05-2019, 01:57 PM
I agree, it would feel weird if Duke beats UVA twice but stayed behind them in KenPom, when evaluating the traditional way overall #1 seeds are awarded. My counter would just be, is it actually so obvious that Duke deserves the #1 seed instead of UVA in that scenario? The reason UVA had such a huge lead in KenPom (up until the last 2 games) is that they were absolutely destroying people. It might actually be the case that UVA is a better team, even with two losses to Duke.)

Virginia has been destroying people, by average of 20. Duke has been destroying by an average of almost 22 - probably against a better schedule.

Kedsy
02-05-2019, 02:50 PM
Virginia has been destroying people, by average of 20. Duke has been destroying by an average of almost 22 - probably against a better schedule.

The numbers you cite don't tell the whole story, because Virginia plays at a much slower pace than Duke. If Virginia played as many possessions as Duke has, that 20 point average margin would be almost 25.

But be that as it may, if Duke had performed better than Virginia against a better schedule, by definition Duke would be ahead of Virginia in the computer rankings. At this point, I assume the real reason Duke is behind Virginia is that our efficiencies were worse while Tre Jones was out with his injury. But if that weren't true, and UVa was still ahead of Duke, it would be because UVa performed better against their schedule than Duke did against their's. The only reason to say that a team lower in the computer rankings should be considered better than a team or teams ahead of them is because the computer system is flawed. But my big issue is, if the system is that flawed, it doesn't make sense to use that same system for the ancillary evaluation.

For example, if you're using a team's performance against the "top 50" as secondary information because you don't entirely trust the overall ranking, but you determine which teams are in the top 50 based on the same rating system you don't entirely trust, that doesn't make a lot of sense, at least to me.

And, as ice-9 said, if, e.g., schedule strength is a major part underlying the computer ranking, using schedule strength as a separate, supporting evaluator is double-counting, essentially saying that schedule strength should be counted more than it already is in the ranking calculation. And again, if you think that, it means you think the computer system is flawed, presumably too flawed to be relied upon (which begs the question, why are you even using it?).

The obvious answer is to find a computer ranking system that isn't flawed (or at least is minimally flawed) and then live with what it tells us. But apparently that's either too difficult or not acceptable, so I guess we'll have to live with what we have.

CDu
02-05-2019, 03:01 PM
And, as ice-9 said, if, e.g., schedule strength is a major part underlying the computer ranking, using schedule strength as a separate, supporting evaluator is double-counting, essentially saying that schedule strength should be counted more than it already is in the ranking calculation. And again, if you think that, it means you think the computer system is flawed, presumably too flawed to be relied upon (which begs the question, why are you even using it?).

The obvious answer is to find a computer ranking system that isn't flawed (or at least is minimally flawed) and then live with what it tells us. But apparently that's either too difficult or not acceptable, so I guess we'll have to live with what we have.

I think the counterargument COULD be that the committee wants to reward teams for playing tougher schedules. So if you had one team rated #4 by the computer ranking and one team #5, but that #5 team played a MUCH tougher schedule, you might be inclined to give the #5 team a better seed as a reward for playing a tougher schedule.

Not saying that is what is going on, but I could see that logic.

That being said, there's basically no system that is going to be perfect, because we have too little information. There are only about 5-6 nonconference games per team with relevance for which to gauge a team's value, and almost all of those happen in November or December. So it's really hard to evaluate, no matter how intricate your calculations are. Now, you can get pretty reasonable guessimates, but a high degree of accuracy is going to be virtually impossible to achieve.

Another of the problems with the system is that efficiency margins may be inherently somewhat tempo- and opponent-dependent. But that's another discussion altogether.

HereBeforeCoachK
02-05-2019, 03:09 PM
The numbers you cite don't tell the whole story, because Virginia plays at a much slower pace than Duke. If Virginia played as many possessions as Duke has, that 20 point average margin would be almost 25.
.

Interesting, but another way to norm it is this:
Duke is scoring about 1.34 points for every opponent point....UVa is scoring about 1.37, so it's very very close. Close enough so that the argument that "Virginia is destroying people" as a reason to put them ahead of Duke...doesn't hold water. There may be other reasons to do so, but that is not one of them. And THAT was the specific post I was responding to.

I also think norming for pace is not necessarily indicated for this comparison. I understand the idea behind it, but at the end of the day, a FT counts 1, a hoop counts 2 and a trey counts 3...and in the last couple minutes of a game, that's all that matters and that's when the margins are determined.

Kedsy
02-05-2019, 03:11 PM
I think the counterargument COULD be that the committee wants to reward teams for playing tougher schedules. So if you had one team rated #4 by the computer ranking and one team #5, but that #5 team played a MUCH tougher schedule, you might be inclined to give the #5 team a better seed as a reward for playing a tougher schedule.

Not saying that is what is going on, but I could see that logic.

That being said, there's basically no system that is going to be perfect, because we have too little information. There are only about 5-6 nonconference games per team with relevance for which to gauge a team's value, and almost all of those happen in November or December. So it's really hard to evaluate, no matter how intricate your calculations are. Now, you can get pretty reasonable guessimates, but a high degree of accuracy is going to be virtually impossible to achieve.

Another of the problems with the system is that efficiency margins may be inherently somewhat tempo- and opponent-dependent. But that's another discussion altogether.

I agree with what you're saying here, except for one thing. If you want to reward teams for tougher schedules, then make schedule strength a more influential factor in your rating system.

This was obviously a bigger problem with the RPI. That system made schedule strength worth 75% of the rating (putting aside that the way they calculated schedule strength was stupid). And then the committee used schedule strength as an important factor on top of it. If they felt that strongly, they should have made schedule strength worth 80% or 90% of the rating and not counted it twice. The same problem exists here, it's just more subtle, in that (a) schedule strength is worked into the system in a more complicated way; and (b) the way they calculate schedule strength is less stupid. But still, if they don't think schedule strength is prominent enough in the rankings, the solution would be to change the ranking system, not count one particular factor twice.

Kedsy
02-05-2019, 03:19 PM
Close enough so that the argument that "Virginia is destroying people" as a reason to put them ahead of Duke...doesn't hold water. There may be other reasons to do so, but that is not one of them. And THAT was the specific post I was responding to.

OK, yes, I apologize. My observation wasn't entirely useful or relevant in relation to your point, and I don't like it when people do that to me.



I also think norming for pace is not necessarily indicated for this comparison. I understand the idea behind it, but at the end of the day, a FT counts 1, a hoop counts 2 and a trey counts 3...and in the last couple minutes of a game, that's all that matters and that's when the margins are determined.

With this, I disagree. If in a hypothetical game each team got the ball only five times and one team won by 10, that's a LOT more impressive than winning by 20 in a game when each team gets the ball 80 times. Sure it all counts the same in win/loss records, but in any system evaluating which team is better (based on past performance), pace matters.

CDu
02-05-2019, 03:20 PM
I agree with what you're saying here, except for one thing. If you want to reward teams for tougher schedules, then make schedule strength a more influential factor in your rating system.

This was obviously a bigger problem with the RPI. That system made schedule strength worth 75% of the rating (putting aside that the way they calculated schedule strength was stupid). And then the committee used schedule strength as an important factor on top of it. If they felt that strongly, they should have made schedule strength worth 80% or 90% of the rating and not counted it twice. The same problem exists here, it's just more subtle, in that (a) schedule strength is worked into the system in a more complicated way; and (b) the way they calculate schedule strength is less stupid. But still, if they don't think schedule strength is prominent enough in the rankings, the solution would be to change the ranking system, not count one particular factor twice.

Right, I'm not saying it's the BEST way to do it. Just A way. And an easy way for them to have an overt excuse for any discrepancies in seeding.

Another reason I think they do this is to allow themselves for some wiggle room come selection time. If the NCAA suddenly decided "hey, we're going to use this super-awesome formula as the one and only measure for at-large selection" then there would be some fallout from that too:
1. You lose the "excitement" of Selection Sunday; and
2. You lose argument ground in how you finagle seeding to account for the non-basketball stuff they account for

By keeping several different metrics in play, they give themselves enough subjectivity to justify different selections and different seeding arrangements.

That said, it will be interesting to see how the committee seeds teams with respect to NET.

Kedsy
02-05-2019, 03:26 PM
Right, I'm not saying it's the BEST way to do it. Just A way. And an easy way for them to have an overt excuse for any discrepancies in seeding.

Another reason I think they do this is to allow themselves for some wiggle room come selection time. If the NCAA suddenly decided "hey, we're going to use this super-awesome formula as the one and only measure for at-large selection" then there would be some fallout from that too:
1. You lose the "excitement" of Selection Sunday; and
2. You lose argument ground in how you finagle seeding to account for the non-basketball stuff they account for

By keeping several different metrics in play, they give themselves enough subjectivity to justify different selections and different seeding arrangements.

That said, it will be interesting to see how the committee seeds teams with respect to NET.

You make a lot of sense here.

But that also means when doing an exercise such as the one that started this thread, people need to distinguish between which teams deserve the better seeds and predicting what the committee will do. Because those might be two entirely different things.

HereBeforeCoachK
02-05-2019, 03:26 PM
With this, I disagree. If in a hypothetical game each team got the ball only five times and one team won by 10, that's a LOT more impressive than winning by 20 in a game when each team gets the ball 80 times. Sure it all counts the same in win/loss records, but in any system evaluating which team is better (based on past performance), pace matters.

I fully understand that point...and by ratio you are certainly correct. However, consider this: there are high paced Duke games where the margins often shrink by 8-10 points very late because Duke ran off and buried the team early, and you have mop up time for extended minutes. I would not think that this is as true as often for UVa, though I don't know that for sure.

In some games, I suspect you would find that at the possession mark of most Virginia games, that Duke was further ahead than UVa's 20 point average. That's why in a final score situation, there can be a lot of deception related to how the game actually went. I know Duke has won a good number of games this season when the final margin was significantly lower than a margin held midway through the second half - or at the 8 minute mark, or 5 minute mark, what have you.

This is also why betting is such a gamble...you can have a very correct feel for how the game will play out, and be right, and yet have things that happen that mitigate the final margin.

CDu
02-05-2019, 03:35 PM
I fully understand that point...and by ratio you are certainly correct. However, consider this: there are high paced Duke games where the margins often shrink by 8-10 points very late because Duke ran off and buried the team early, and you have mop up time for extended minutes. I would not think that this is as true as often for UVa, though I don't know that for sure.

In some games, I suspect you would find that at the possession mark of most Virginia games, that Duke was further ahead than UVa's 20 point average. That's why in a final score situation, there can be a lot of deception related to how the game actually went. I know Duke has won a good number of games this season when the final margin was significantly lower than a margin held midway through the second half.

This is also why betting is such a gamble...you can have a very correct feel for how the game will play out, and be right, and yet have things that happen that mitigate the final margin.

I will say that it is generally hard to truly tease out what the "true" efficiency is in blowouts. Because teams sub in their benches at different times, and teams' starters can relax at different times. So I don't know one way or the other where that would go in discussing UVa and Duke. It could go either way.

However, I will also say that a good team that plays slow will tend to get an efficiency boost in closer games due to late-game fouling. If you are ahead near the end, but it is close enough to foul for it, that difference is going to be magnified for a slow-pace team than for a fast-pace team.

But that is a very specific instance, and I'm not sure how many games that has been relevant for UVa this year.

I think the bigger driver is that we've had a series of games hurt by key injuries/illness to our starters, either early in the game or before the game.

Kedsy
02-05-2019, 03:39 PM
I fully understand that point...and by ratio you are certainly correct. However, consider this: there are high paced Duke games where the margins often shrink by 8-10 points very late because Duke ran off and buried the team early, and you have mop up time for extended minutes. I would not think that this is as true as often for UVa, though I don't know that for sure.

In some games, I suspect you would find that at the possession mark of most Virginia games, that Duke was further ahead than UVa's 20 point average. That's why in a final score situation, there can be a lot of deception related to how the game actually went. I know Duke has won a good number of games this season when the final margin was significantly lower than a margin held midway through the second half.

This is also why betting is such a gamble...you can have a very correct feel for how the game will play out, and be right, and yet have things that happen that mitigate the final margin.

I suppose most of this is true, though if Virginia is winning their games by an average of 20 points I imagine they are similarly affected by the same phenomenon. Though it's also true that in many games the final margin is larger than the closeness of the game would suggest, either because of excessive fouling at the end of the game or because the end of game lineups for the winning team are superior to the end of game lineups for the losing team.

I remember reading once where a guy talked about creating a rating system based on minute-by-minute win probabilities -- there are people who calculate the chance of each team winning at each minute of the game -- which would presumably take pace into account and would also eliminate the problem you're talking about, as well as the need for diminishing returns for larger point spreads. I've never actually seen such a rating system, though, so either (a) it was too difficult to create such a system that made sense; (b) the information necessary for such a system wasn't readily available; or (c) it's out there but I haven't seen it and it never got popular enough to matter.

HereBeforeCoachK
02-05-2019, 03:43 PM
I will say that it is generally hard to truly tease out what the "true" efficiency is in blowouts. Because teams sub in their benches at different times, and teams' starters can relax at different times. So I don't know one way or the other where that would go in discussing UVa and Duke. It could go either way.

However, I will also say that a good team that plays slow will tend to get an efficiency boost in closer games due to late-game fouling. If you are ahead near the end, but it is close enough to foul for it, that difference is going to be magnified for a slow-pace team than for a fast-pace team.

But that is a very specific instance, and I'm not sure how many games that has been relevant for UVa this year.

I think the bigger driver is that we've had a series of games hurt by key injuries/illness to our starters, either early in the game or before the game.

I agree with all of that, with one little quibble....a tiny one....and that is that it can be truly hard to tease out the true efficiency in many games, not just blow outs. Or let me back up....hard to tease out the final margin in many games. I'll use a betting example, but it works for this too: a 10 point underdog plays a great game and takes a heavily favored team to overtime. In OT, the favored team jumps out to a lead and ends up winning by 11 with late FTs.

Here you have a game where A: those bettors who correctly analyzed that the game would be close and took the underdog....actually lost the bet....and B: the last couple minutes really skewed the overall efficiency and margin stats for the whole game.

I hope I explained that in a way that makes sense.

CDu
02-05-2019, 03:46 PM
I agree with all of that, with one little quibble...a tiny one...and that is that it can be truly hard to tease out the true efficiency in many games, not just blow outs. Or let me back up...hard to tease out the final margin in many games. I'll use a betting example, but it works for this too: a 10 point underdog plays a great game and takes a heavily favored team to overtime. In OT, the favored team jumps out to a lead and ends up winning by 11 with late FTs.

Here you have a game where A: those bettors who correctly analyzed that the game would be close and took the underdog...actually lost the bet...and B: the last couple minutes really skewed the overall efficiency and margin stats for the whole game.

I hope I explained that in a way that makes sense.

That's true. Even in semi-close games, things can be quirky. And in general, the fewer the possessions, the quirkier things tend to be, because any outlier occurrence takes on a larger proportion of the final result.

HereBeforeCoachK
02-05-2019, 03:46 PM
I remember reading once where a guy talked about creating a rating system based on minute-by-minute win probabilities -- there are people who calculate the chance of each team winning at each minute of the game -- which would presumably take pace into account and would also eliminate the problem you're talking about, as well as the need for diminishing returns for larger point spreads. I've never actually seen such a rating system, though, so either (a) it was too difficult to create such a system that made sense; (b) the information necessary for such a system wasn't readily available; or (c) it's out there but I haven't seen it and it never got popular enough to matter.

I love it, though that may be taking it to the absurd - but it would be a mathematically valid measure if for no other reasons than the fact that you'd have 40 reference points per game instead of one. Analytics was born in baseball because they are more predictive when your universe of reference points is 162 games versus 20 or 40 or whatever, depending on the sport.

And if you watch those ESPN trackers, someone somewhere (or some machine) is calculating odds of winning after every play. So this could be done.

jhmoss1812
02-05-2019, 04:13 PM
I fully understand that point...and by ratio you are certainly correct. However, consider this: there are high paced Duke games where the margins often shrink by 8-10 points very late because Duke ran off and buried the team early, and you have mop up time for extended minutes. I would not think that this is as true as often for UVa, though I don't know that for sure.

In some games, I suspect you would find that at the possession mark of most Virginia games, that Duke was further ahead than UVa's 20 point average. That's why in a final score situation, there can be a lot of deception related to how the game actually went. I know Duke has won a good number of games this season when the final margin was significantly lower than a margin held midway through the second half - or at the 8 minute mark, or 5 minute mark, what have you.

This is also why betting is such a gamble...you can have a very correct feel for how the game will play out, and be right, and yet have things that happen that mitigate the final margin.

This has happened a lot to UVA this year as well. I don't know if it's happened more or less than it's happened to Duke but Bennett usually pulls starters when up big and the lead shrinks. As one example, we were up 29 against FSU with around 3 minutes left in the game. FSU then goes on a 16-0 run in the last 3 minutes (it was pathetic by our backups) to make the final score look somewhat respectable. Let's just say that there have been a lot of backdoor covers by the opponent in UVA games this year. I'm sure the people in the degenerate gambling thread can attest to that.

uh_no
02-05-2019, 04:25 PM
Interesting, but another way to norm it is this:
Duke is scoring about 1.34 points for every opponent point...UVa is scoring about 1.37, so it's very very close. Close enough so that the argument that "Virginia is destroying people" as a reason to put them ahead of Duke...doesn't hold water. There may be other reasons to do so, but that is not one of them. And THAT was the specific post I was responding to.

I also think norming for pace is not necessarily indicated for this comparison. I understand the idea behind it, but at the end of the day, a FT counts 1, a hoop counts 2 and a trey counts 3...and in the last couple minutes of a game, that's all that matters and that's when the margins are determined.

the difference between Duke and uva is easily within the margin of whos third string played better.

Hingeknocker
02-05-2019, 04:51 PM
To clarify my semantic choice that semi-derailed this thread, the reason I said "they were destroying people" in reference to Virgina was, well, they were. (Emphasis on the past tense). Their KenPom rating before the OT win against State was 37-point-something, which is the highest I can ever remember seeing during a season, and would have been the highest ever rating for any end-of-season team.

Anyway, when comparing Duke to UVA this year, obviously the margins are very close. And have only gotten closer over the past 2 weeks. What I still believe is that the efficiency ratings (choose your favorite, or choose a composite of some kind) are inherently superior to anything else we have as a rankings system. They are not perfect, of course. But for the purposes of seeding the tournament, I think a straight S-curve on, say, the Massey Composite ratings would be superior to every single tournament that the committee has ever bracketed.

ETA: The discussion about how much "garbage time" impacts a team's final margin of victory, and thus, their efficiency rating is very interesting. One thing I've enjoyed following this year is Bart Torvik's Game Score and +/-. These two metrics do a very good job of showing how "in control" a particular victory was, and would (in general) penalize you less for giving up points in garbage time.

jv001
02-05-2019, 05:03 PM
This has happened a lot to UVA this year as well. I don't know if it's happened more or less than it's happened to Duke but Bennett usually pulls starters when up big and the lead shrinks. As one example, we were up 29 against FSU with around 3 minutes left in the game. FSU then goes on a 16-0 run in the last 3 minutes (it was pathetic by our backups) to make the final score look somewhat respectable. Let's just say that there have been a lot of backdoor covers by the opponent in UVA games this year. I'm sure the people in the degenerate gambling thread can attest to that.

I watched the game you guys played against FSU and you were destroying the Noles. Bennett pulled the starters and the game ended up looking a lot closer than it really was. I don't know if those bench players are that bad or just had a bad game. GoDuke!

ElliottHoo
02-05-2019, 05:08 PM
This has happened a lot to UVA this year as well. I don't know if it's happened more or less than it's happened to Duke but Bennett usually pulls starters when up big and the lead shrinks. As one example, we were up 29 against FSU with around 3 minutes left in the game. FSU then goes on a 16-0 run in the last 3 minutes (it was pathetic by our backups) to make the final score look somewhat respectable. Let's just say that there have been a lot of backdoor covers by the opponent in UVA games this year. I'm sure the people in the degenerate gambling thread can attest to that.

Adding a detail: UVA has held at least a 20 point lead in every ACC game except vs Duke and the last 2. Dunno how that compares with Duke.

English
02-05-2019, 05:19 PM
The numbers you cite don't tell the whole story, because Virginia plays at a much slower pace than Duke. If Virginia played as many possessions as Duke has, that 20 point average margin would be almost 25.

But be that as it may, if Duke had performed better than Virginia against a better schedule, by definition Duke would be ahead of Virginia in the computer rankings. At this point, I assume the real reason Duke is behind Virginia is that our efficiencies were worse while Tre Jones was out with his injury. But if that weren't true, and UVa was still ahead of Duke, it would be because UVa performed better against their schedule than Duke did against their's. The only reason to say that a team lower in the computer rankings should be considered better than a team or teams ahead of them is because the computer system is flawed. But my big issue is, if the system is that flawed, it doesn't make sense to use that same system for the ancillary evaluation.

For example, if you're using a team's performance against the "top 50" as secondary information because you don't entirely trust the overall ranking, but you determine which teams are in the top 50 based on the same rating system you don't entirely trust, that doesn't make a lot of sense, at least to me.

And, as ice-9 said, if, e.g., schedule strength is a major part underlying the computer ranking, using schedule strength as a separate, supporting evaluator is double-counting, essentially saying that schedule strength should be counted more than it already is in the ranking calculation. And again, if you think that, it means you think the computer system is flawed, presumably too flawed to be relied upon (which begs the question, why are you even using it?).

The obvious answer is to find a computer ranking system that isn't flawed (or at least is minimally flawed) and then live with what it tells us. But apparently that's either too difficult or not acceptable, so I guess we'll have to live with what we have.

I think this bears some emphasis--the efficiency rankings do not discount games with significant injuries, but the Committee (rightly or wrongly...and certainly subjectively) does in the case that a team was minus one or more players who have since returned and shown that they're back to health/productivity. That's relevant in the Duke/UVa example, and plenty of others. Likewise, there are other examples (see Martin, Kenyon & the Bearcats back in the day) where the efficiency rankings might have a team much higher than its predictive potential because its efficiency margin was largely earned before a key contributor was lost. It would be tough to quantitatively build that into a formula (and that's been the subject of some discussion recently among KenPom and others, specifically WRT Tre Jones this season).

ElliottHoo
02-05-2019, 05:29 PM
The discussion about how much "garbage time" impacts a team's final margin of victory, and thus, their efficiency rating is very interesting. One thing I've enjoyed following this year is Bart Torvik's Game Score and +/-. These two metrics do a very good job of showing how "in control" a particular victory was, and would (in general) penalize you less for giving up points in garbage time.

Iíve got a question re: efficiency. Obviously margin of victory is key, winning by 5 is better than by 2. But how valuable is the win itself? Winning by 5 instead of 2 is 3, with some degree of sliding scale, but surely that ~3 is less valuable than winning by 1 instead of losing by 2, right? How much better? Any idea?

Hingeknocker
02-05-2019, 05:46 PM
I’ve got a question re: efficiency. Obviously margin of victory is key, winning by 5 is better than by 2. But how valuable is the win itself? Winning by 5 instead of 2 is 3, with some degree of sliding scale, but surely that ~3 is less valuable than winning by 1 instead of losing by 2, right? How much better? Any idea?

I could butcher this, so someone please correct me if wrong, but I'm pretty sure the efficiency systems do not "reward" or "punish" a loss, per se. Any boost to your efficiency score would come from your adjusted per-possession performances on offense and defense. Obviously, a net positive per-possession performance during a game is just a fancy way of saying you won the game. If you're talking about the difference between a win and a loss, an efficiency-based system like KenPom or Torvik just cares about how you did on an adjusted per-possession basis, and how much better or worse that was than your season average was.

HereBeforeCoachK
02-05-2019, 05:48 PM
ETA: The discussion about how much "garbage time" impacts a team's final margin of victory, and thus, their efficiency rating is very interesting. One thing I've enjoyed following this year is Bart Torvik's Game Score and +/-. These two metrics do a very good job of showing how "in control" a particular victory was, and would (in general) penalize you less for giving up points in garbage time.

That sounds like a fascinating, and valid metric. I'm not too familiar with that....gonna Google that guy and learn more. I think "in control" - if his numbers capture it - is incredibly valuable and predictive information.

Kedsy
02-05-2019, 06:04 PM
Iíve got a question re: efficiency. Obviously margin of victory is key, winning by 5 is better than by 2. But how valuable is the win itself? Winning by 5 instead of 2 is 3, with some degree of sliding scale, but surely that ~3 is less valuable than winning by 1 instead of losing by 2, right? How much better? Any idea?


I could butcher this, so someone please correct me if wrong, but I'm pretty sure the efficiency systems do not "reward" or "punish" a loss, per se. Any boost to your efficiency score would come from your adjusted per-possession performances on offense and defense. Obviously, a net positive per-possession performance during a game is just a fancy way of saying you won the game. If you're talking about the difference between a win and a loss, an efficiency-based system like KenPom or Torvik just cares about how you did on an adjusted per-possession basis, and how much better or worse that was than your season average was.

Yeah, Hingeknocker got it right. Efficiency-based systems like Pomeroy and Torvik don't give any extra boost for winning the game.

uh_no
02-05-2019, 06:06 PM
That sounds like a fascinating, and valid metric. I'm not too familiar with that...gonna Google that guy and learn more. I think "in control" - if his numbers capture it - is incredibly valuable and predictive information.

you definitely could attempt to regress teams efficiency vs incontrolledness, and begin to discount negative impacts of letting off the gas.

HereBeforeCoachK
02-05-2019, 06:06 PM
Yeah, Hingeknocker got it right. Efficiency-based systems like Pomeroy and Torvik don't give any extra boost for winning the game.

That sounds like a little bit of a fly in the ointment there....how would you tweak those systems to include some kind of adjustment based on winning or losing....

uh_no
02-05-2019, 06:11 PM
That sounds like a little bit of a fly in the ointment there...how would you tweak those systems to include some kind of adjustment based on winning or losing...

actually winning the game provides very little predictive value above efficiency.

if you want to give bonus points for winning, net does exactly that.

HereBeforeCoachK
02-05-2019, 06:22 PM
actually winning the game provides very little predictive value above efficiency.

if you want to give bonus points for winning, net does exactly that.

Right, but it does, and should, provide a little extra.......garbage time and all.....

House P
02-05-2019, 06:47 PM
I remember reading once where a guy talked about creating a rating system based on minute-by-minute win probabilities -- there are people who calculate the chance of each team winning at each minute of the game -- which would presumably take pace into account and would also eliminate the problem you're talking about, as well as the need for diminishing returns for larger point spreads. I've never actually seen such a rating system, though, so either (a) it was too difficult to create such a system that made sense; (b) the information necessary for such a system wasn't readily available; or (c) it's out there but I haven't seen it and it never got popular enough to matter.

In his T-Rank system (http://barttorvik.com), Bart Torvik does a couple of things to account for scenarios where the final couple minutes are not played like a "normal" game (e.g. when a large lead narrows during garbage time or when a single digit lead grows to double digits when the trailing team fouls out of desperation). Here is my understanding of how we addresses a couple specific scenarios.

1. Garbage time: Possession which occur after a lead has been determine to be 100% safe are significantly de-emphasized in T-Rank. The point at which a lead is considered 100% safe is based on a formula developed by Bill James which says that a 20 point lead is safe with ~5 minutes left and a 30 point lead is 100% safe with ~12 min left (anyone interested in more details can click this link (http://www.slate.com/articles/sports/sports_nut/2008/03/the_lead_is_safe.html)). Using the Duke/Clemson game as an example, Duke's lead was considered 100% safe when Duke led by 28 with 9.5 minutes left. Therefore, the fact that Duke was outscored 23-14 for the rest of the game doesn't impact Duke's overall T-Rank rating very much.

2. Game Control: Instead of using only the final score as an input, T-Rank also uses the "average lead" over the course of the game. This can help differentiate a tight game which ended up with a 10 point margin due to fouling from a game where one team led by double digits most of the way, but the final margin ended up at 10 points. As Hingeknocker points out, T-Rank lists the average margin as "+/-" on each team page. Two good examples for Duke are the Texas Tech and Pitt games. The final margin was 11 vs Texas Tech and 15 vs Pitt, but the average margin was -0.8 for Texas Tech and 13 for Pitt. This reflects the fact that the Texas Tech game was significantly more competitive than the Pitt game.

3. Blow outs between mismatched teams: If the margin of victory is more than 10 points and the overall ranking between two teams is above a certain threshold, the result of the game starts getting discounted. . However, as an example, the influence of Duke/Stetson game is discounted by 75% compared to a "normal" game. I should point out that Duke's blowout of Kentucky is not discounted because the two teams are not considered to be "mismatched". Anyone interested in the specific calculation can check out this link (http://adamcwisports.blogspot.com/p/every-possession-counts.html).

I know that KenPom adjusts for blowouts, but I am not sure he makes the other adjustments.

Finally, while KenPom is still the go-to guy for college basketball analytics, I am growing to really like the (free) content T-Rank provides. I would highly encourage anyone interested in this kind of stuff to check out his site (and listen to the episode of Jordan Sperber's podcast where Bart Torvik was a guest ).

Nugget
02-05-2019, 07:23 PM
Thanks! Really appreciate the clarity/detail of the explanation. And, agree that Sperber's podcast is very well done and illuminating.

uh_no
02-05-2019, 07:28 PM
Right, but it does, and should, provide a little extra....garbage time and all....

NET, sure. Rewarding teams for actually winning is important.

KP is a predictor. w/l provides no marginal predictive value, and thus is not factored in. He had posts a while back where he did a study on clutch and close games, and clutchness and winning close games are not evidence of clutchness or winning close games moving forward.

Ian
02-06-2019, 12:26 AM
I'm just glad we aren't using the RPI any more. Kansas is still #1 RPI even after their loss tonight.

BandAlum83
02-06-2019, 12:51 AM
I'm just glad we aren't using the RPI any more. Kansas is still #1 RPI even after their loss tonight.

Seriously? Are you sure you donít need to refresh you browser? I would have thought they would have dropped at least 3 losses ago!

Thatís really bad.

ElliottHoo
02-06-2019, 07:11 AM
Seriously? Are you sure you donít need to refresh you browser? I would have thought they would have dropped at least 3 losses ago!

Thatís really bad.

RPI really, really loved SoS and Kansas has that in spades.

camion
02-06-2019, 07:44 AM
RPI really, really loved SoS and Kansas has that in spades.

I know I've said this before. My understanding of the RPI is that it was not designed to rank the teams in order of their own strength or likelihood of winning. It was created to discourage teams from loading their schedules with cupcakes. As such it was better than nothing.

HereBeforeCoachK
02-06-2019, 07:48 AM
I know I've said this before. My understanding of the RPI is that it was not designed to rank the teams in order of their own strength or likelihood of winning. It was created to discourage teams from loading their schedules with cupcakes. As such it was better than nothing.

But the NCAA used it for the former, even as it was designed for the latter. That was the problem.

camion
02-06-2019, 08:00 AM
But the NCAA used it for the former, even as it was designed for the latter. That was the problem.

The NCAA used RPI as a tool in ranking teams. That doesn't mean that the ranking was necessarily an ordering of the teams' likelihood of winning. Part of that ranking could be how well a team conforms to the NCAA's scheduling wishes.



Note: I really don't enjoy defending the NCAA on any front, but I'm okay with this particular aberration.

HereBeforeCoachK
02-06-2019, 08:16 AM
The NCAA used RPI as a tool in ranking teams. That doesn't mean that the ranking was necessarily an ordering of the teams' likelihood of winning. Part of that ranking could be how well a team conforms to the NCAA's scheduling wishes.



Note: I really don't enjoy defending the NCAA on any front, but I'm okay with this particular aberration.

So in your opinion, the RPI never led to any unfair/bad seedings?

scottdude8
02-06-2019, 09:44 AM
Glad to see this discussion going strong! Sorry for the lack of a new post this week, but itís vacation time, haha. Maybe next week Iíll try to do a RPI/NET comparison or something like that.

uh_no
02-06-2019, 10:06 AM
So in your opinion, the RPI never led to any unfair/bad seedings?

i don't think he said that at all ...

HereBeforeCoachK
02-06-2019, 10:08 AM
i don't think he said that at all ...

Actually, when you look at my post that he responded to...then looked at his...there is no other reason to make that reply unless that was the point. I don't think he really means that, but context is everything...and there was no reason to debate my point unless you are willing to stand by that. The context was that the RPI would from time to time lead to bad seedings......not all the time....so my burden of proof was low.

Nugget
02-06-2019, 05:08 PM
I may have to listen to this whole section, because I actually think there's a reasonable compromise to be had between efficiency (NET, KenPom, etc) and results-based evaluations. Selecting the teams for the tournament should be based on a host of factors, including your black-and-white wins and losses, that efficiency metrics aren't designed to include. As a hypothetical, it would be possible for a team to go 10-20 on the season, but end up with a relatively high KenPom rating; perhaps they played a very hard schedule, and all of their 20 losses came by 1 point. But, obviously, this team should not make the tournament.

However, seeding the teams absolutely, without a doubt, should be done via the efficiency metrics. The predictive metrics are all that matter once the tournament has its 68 teams. If you're a team who earns a 1 seed under this metric, you deserve to play a team truly of the caliber of an 8/9 seed in the 2nd round, 4/5 in the S16, etc. etc. I would be perfectly happy if the tournament seeds were a straight S-curve of the KenPom rankings at the end of the year (with the usual bracketing rules in place, of course).

All in all, very intelligent observation and conversation about the systems. Thank you for the recap!

To your point, the interview Matt Norlander posted this morning with Committee Chair Bernard Muir at CBS Sports.com includes this from Muir indicating that they see the predictive metrics as having specific value related to seeding:

"Muir: I would say I'm happy with our process. With how this NET plays out, what we glean from it this year will be interesting. It's been helpful the past couple of years as we've introduced predictive metrics in our discussion. More so I think that's been a good thing for us especially in terms of seeding and getting that seeding right because that's so important to creating a balanced bracket and a great national tournament. I...."

BandAlum83
02-06-2019, 05:13 PM
RPI really, really loved SoS and Kansas has that in spades.

What good is SoS if you lose the games?

HereBeforeCoachK
02-06-2019, 05:14 PM
What good is SoS if you lose the games?

That does seem like the logical question, doesn't it?

camion
02-07-2019, 07:54 AM
Actually, when you look at my post that he responded to...then looked at his...there is no other reason to make that reply unless that was the point. I don't think he really means that, but context is everything...and there was no reason to debate my point unless you are willing to stand by that. The context was that the RPI would from time to time lead to bad seedings...not all the time...so my burden of proof was low.

No.

I don't see your logic. For one thing, your definition of good/bad seeding is not THE definition. The seeding choices of the NCAA don't have to be based solely on perceived strength. Seeding could be their main leverage for encouraging teams to schedule with the RPI in mind.

Note: This argument doesn't mean I always agree with NCAA seeding choices.

HereBeforeCoachK
02-07-2019, 08:33 AM
. Seeding could be their main leverage for encouraging teams to schedule with the RPI in mind..

And if that's the case, that's a bastardization of the NCAAT seeding process, and is absolutely improper and frankly, total BS. If they are doing anything with seeding other than properly seeding teams on merit, then they are perverting the process.

Mtn.Devil.91.92.01.10.15
02-07-2019, 08:51 AM
And if that's the case, that's a bastardization of the NCAAT seeding process, and is absolutely improper and frankly, total BS. If they are doing anything with seeding other than properly seeding teams on merit, then they are perverting the process.

Wait, what? Are you saying that scheduling more difficult teams and then performing well against them is something other than merit?

HereBeforeCoachK
02-07-2019, 09:30 AM
Wait, what? Are you saying that scheduling more difficult teams and then performing well against them is something other than merit?

No, to clarify: the point was made by another poster, or so it seemed, that the NCAA used scheduling as a reward/punishment for teams OUTSIDE of, or in addition to, their deserved performance. I believe he said that the RPI rewarded scheduling more than results, and that the NCAA would sometimes seed teams partially on scheduling over merit as a way to try and change their scheduling behavior.

Some coaches prefer to bring their teams along slowly, scheduling more cupcakes...while others like to jump right into the fire early. Both are legitimate ways to run a program and the NCAA should only concern themselves with the merits of the team at seeding time...not so much how they got there.

If a team is a 10 seed on merit, after examining their results including the schedule they faced, then to "punish" them on top of that for a schedule the NCAA doesn't like by giving them a 12 or 13 seed is simply inappropriate - because a teams seeding impacts other teams as well. No 4th seed should have to face a legit ten seed simply because the NCAA punished that team with a 13 seed. And no legit ten seed should have to start out with a 4 seed simply because the NCAA didn't really like the way they got to their merited ten.

My understanding of the other poster's point was that the NCAA used scheduling as a carrot and a stick, above and beyond the normal merit considerations (which of course inherently contemplates sos.) I further understood him to say this was proper. If that was his point, then I think that's totally improper. If I missed his point, then, well, never mind...

uh_no
02-07-2019, 10:21 AM
And if that's the case, that's a bastardization of the NCAAT seeding process, and is absolutely improper and frankly, total BS. If they are doing anything with seeding other than properly seeding teams on merit, then they are perverting the process.

Given there is no morals or customs associated with seeding, i'm not sure how one could say any practice is objectively proper or improper.

The NCAA has a huge vested interest in good games, and they can use both the RPI and NET have factors that help reward the kind of scheduling they want to see. RPI, obviously is SOS. With NET, the 10 point cap serves to limit the bonus you get for beating up on a crappy team...so you'll get a far better boost by playing a better team and beating them by about 10.

Frankly, given that I hate watching crappy basketball in november, I'm all for the NCAA rewarding teams for scheduling tough. Make it double, triple more important than it is today...IDC...

English
02-07-2019, 12:50 PM
Given there is no morals or customs associated with seeding, i'm not sure how one could say any practice is objectively proper or improper.

The NCAA has a huge vested interest in good games, and they can use both the RPI and NET have factors that help reward the kind of scheduling they want to see. RPI, obviously is SOS. With NET, the 10 point cap serves to limit the bonus you get for beating up on a crappy team...so you'll get a far better boost by playing a better team and beating them by about 10.

Frankly, given that I hate watching crappy basketball in november, I'm all for the NCAA rewarding teams for scheduling tough. Make it double, triple more important than it is today...IDC...

Except that NET doesn't really do that because of its unadjusted efficiency component, which absolutely does reward a team for beating the absolute life out of crappy teams. Even a series of crappy teams. That's the entire crux of this NCSU example. They dismantled a handful of woeful, sub-300 KP teams in their NCSOS and were rewarded handsomely in the NET for it.

And hey, listen, I'm with you on the scheduling. Any time I look at the game slate and there are no top-25 matchups on a given day, I'm majorly bummed. I'm also VERY surprised by the seeming stout defense of the RPI here, because, huh?! Stacking SoS on top of SoS on top of SoS to rank and weigh teams was and is a vile, misguided way to handle the selection and seeding. That's (until I saw this thread) a universal opinion held by the galaxy of college basketball people. Even the NCAA people who were using it until this season. So, again, huh?!

Bluedog
02-07-2019, 01:12 PM
Except that NET doesn't really do that because of its unadjusted efficiency component, which absolutely does reward a team for beating the absolute life out of crappy teams. Even a series of crappy teams. That's the entire crux of this NCSU example. They dismantled a handful of woeful, sub-300 KP teams in their NCSOS and were rewarded handsomely in the NET for it.

And hey, listen, I'm with you on the scheduling. Any time I look at the game slate and there are no top-25 matchups on a given day, I'm majorly bummed. I'm also VERY surprised by the seeming stout defense of the RPI here, because, huh?! Stacking SoS on top of SoS on top of SoS to rank and weigh teams was and is a vile, misguided way to handle the selection and seeding. That's (until I saw this thread) a universal opinion held by the galaxy of college basketball people. Even the NCAA people who were using it until this season. So, again, huh?!

While beating up on poor teams may help their "raw NET ranking," I don't think that's the primary thing the committee looks at when making seedings/decisions. They look at the W-L records of the various quadrants. So, a team with an inflated NET ranking, but a record of something 0-2 against Q1 and 2-6 against Q2 (with the rest of their games being 100% victories against Q3 and Q4) will not get a very good seed. That's on the team sheets and what they look at. This is what happens to top teams in lower-tier conferences to some extent as well.

Agree with you on the second paragraph.

uh_no
02-07-2019, 01:17 PM
Except that NET doesn't really do that because of its unadjusted efficiency component, which absolutely does reward a team for beating the absolute life out of crappy teams. Even a series of crappy teams. That's the entire crux of this NCSU example. They dismantled a handful of woeful, sub-300 KP teams in their NCSOS and were rewarded handsomely in the NET for it.

And hey, listen, I'm with you on the scheduling. Any time I look at the game slate and there are no top-25 matchups on a given day, I'm majorly bummed. I'm also VERY surprised by the seeming stout defense of the RPI here, because, huh?! Stacking SoS on top of SoS on top of SoS to rank and weigh teams was and is a vile, misguided way to handle the selection and seeding. That's (until I saw this thread) a universal opinion held by the galaxy of college basketball people. Even the NCAA people who were using it until this season. So, again, huh?!

I'm not trying to say they did it perfectly, or that the ranking is not problematic in other ways....just that there ARE components of it which incentivize playing better teams, and that I'm okay with that.

I think unadjusted efficiency is the biggest flaw and hope they adjust it moving forward. I'm completely fine with the 10 point limit and would love if they even more directly factored in NCSOS to encourage playing good teams. Yeah you could get burned if some of the teams you schedule end up stinking, but such is life.

HereBeforeCoachK
02-07-2019, 09:32 PM
Given there is no morals or customs associated with seeding, i'm not sure how one could say any practice is objectively proper or improper.
.

With due respect, that's incredibly cynical. There are indeed proper and improper contemplations for seeding......and there is an inherent ethic to it (moral is the wrong word here) - because there is only one reason for seeding to even exist, as opposed to just random scheduling.

Hingeknocker
02-10-2019, 11:04 AM
After all this discussion, we finally have some news, in that the NCAA released their as-of-today Top 16 seeds yesterday:

East Region: 1. Duke (1 overall), 2. Michigan (6), 3. Marquette (12), 4. Iowa St. (13)

South Region: 1. Tennessee (2), 2. North Carolina (7), 3. Purdue (9), 4. Nevada (14)

West Region: 1. Gonzaga (4), 2. Michigan St. (8), 3. Kansas (10), 4. Louisville (15)

Midwest Region: 1. Virginia (3), 2. Kentucky (5), 3. Houston (11), 4. Wisconsin (16)

Certainly, much of this will change between now and March, but it's still interesting to get a little snapshot of how the committee is thinking. Especially now that we have the NET rankings updating relatively frequently.

No major surprises in here, but one that jumps out at me is Kansas as the 10th overall team. Neither Kenpom nor Torvik have them that high, and even NET has them at #18 (as of 2/8). Seemingly, what props them up with the committee is their SOS. As ever, this is extremely disappointing to me. If you didn't have NET, and only looked at W-L records, then of course SOS would be relevant. But if you have NET, it doesn't make any sense to carve SOS out and give it such priority. SOS is already factored into the NET ratings!

Ok, rant over. Great win for Duke yesterday which keeps us above the fray, for now.

MChambers
02-10-2019, 11:27 AM
After all this discussion, we finally have some news, in that the NCAA released their as-of-today Top 16 seeds yesterday:

East Region: 1. Duke (1 overall), 2. Michigan (6), 3. Marquette (12), 4. Iowa St. (13)

South Region: 1. Tennessee (2), 2. North Carolina (7), 3. Purdue (9), 4. Nevada (14)

West Region: 1. Gonzaga (4), 2. Michigan St. (8), 3. Kansas (10), 4. Louisville (15)

Midwest Region: 1. Virginia (3), 2. Kentucky (5), 3. Houston (11), 4. Wisconsin (16)

Certainly, much of this will change between now and March, but it's still interesting to get a little snapshot of how the committee is thinking. Especially now that we have the NET rankings updating relatively frequently.

No major surprises in here, but one that jumps out at me is Kansas as the 10th overall team. Neither Kenpom nor Torvik have them that high, and even NET has them at #18 (as of 2/8). Seemingly, what props them up with the committee is their SOS. As ever, this is extremely disappointing to me. If you didn't have NET, and only looked at W-L records, then of course SOS would be relevant. But if you have NET, it doesn't make any sense to carve SOS out and give it such priority. SOS is already factored into the NET ratings!

Ok, rant over. Great win for Duke yesterday which keeps us above the fray, for now.

I agree with you on Kansas. I also find the relatively high seedings of Nevada, Houston, and Marquette discouraging. Reminiscent of the RPI. Yes, I root for Marquette, but still don't think they're that good. Of course, we still have a month for things to sort out a bit.

CDu
02-10-2019, 11:31 AM
After all this discussion, we finally have some news, in that the NCAA released their as-of-today Top 16 seeds yesterday:

East Region: 1. Duke (1 overall), 2. Michigan (6), 3. Marquette (12), 4. Iowa St. (13)

South Region: 1. Tennessee (2), 2. North Carolina (7), 3. Purdue (9), 4. Nevada (14)

West Region: 1. Gonzaga (4), 2. Michigan St. (8), 3. Kansas (10), 4. Louisville (15)

Midwest Region: 1. Virginia (3), 2. Kentucky (5), 3. Houston (11), 4. Wisconsin (16)

Certainly, much of this will change between now and March, but it's still interesting to get a little snapshot of how the committee is thinking. Especially now that we have the NET rankings updating relatively frequently.

No major surprises in here, but one that jumps out at me is Kansas as the 10th overall team. Neither Kenpom nor Torvik have them that high, and even NET has them at #18 (as of 2/8). Seemingly, what props them up with the committee is their SOS. As ever, this is extremely disappointing to me. If you didn't have NET, and only looked at W-L records, then of course SOS would be relevant. But if you have NET, it doesn't make any sense to carve SOS out and give it such priority. SOS is already factored into the NET ratings!

Ok, rant over. Great win for Duke yesterday which keeps us above the fray, for now.

This would be a somewhat annoying draw. Despite having the #1 overall seed, we would have (by their rating and assuming results held to seed) the toughest Sweet-16 opponent and the second-toughest elite-8 opponent. And the second-toughest sum of remaining seeds. In terms of summing seeds, the foursomes in the Midwest and West are notably worse than those in the East and South.

HereBeforeCoachK
02-10-2019, 11:33 AM
This would be a somewhat annoying draw. Despite having the #1 overall seed, we would have (by their rating and assuming results held to seed) the toughest Sweet-16 opponent and the second-toughest elite-8 opponent. And the second-toughest sum of remaining seeds.

True, but for reasons related to the law of averages and regression to the mean, I'm fine with avoiding Michigan State.

UrinalCake
02-10-2019, 11:38 AM
East Region: 1. Duke (1 overall), 2. Michigan (6), 3. Marquette (12), 4. Iowa St. (13)

That would make for an interesting second weekend. Michigan has the whole fab five storyline which the media would eat up. Marquette of course has Wojo. And Iowa State can absolutely bomb away from three, which always makes for a dangerous tournament opponent.

uh_no
02-10-2019, 11:39 AM
I agree with you on Kansas. I also find the relatively high seedings of Nevada, Houston, and Marquette discouraging. Reminiscent of the RPI. Yes, I root for Marquette, but still don't think they're that good. Of course, we still have a month for things to sort out a bit.

KP has nevada at 17. A 4 seed is reasonable
KP has houston at 19...the 3 seed is slightly high
KP has marquette at 29....a stupid overseed.

So marquette is the only real head scratcher here, especially since their SOS isn't particularly good either....but i'm quite happy if the committee wants to put them in our region :D :D

tbyers11
02-10-2019, 11:47 AM
I agree with you on Kansas. I also find the relatively high seedings of Nevada, Houston, and Marquette discouraging. Reminiscent of the RPI. Yes, I root for Marquette, but still don't think they're that good. Of course, we still have a month for things to sort out a bit.

I don't have that much of a problem with Kansas and Marquette as 3 seeds.

Kansas is 18-6. All the other teams in the 3-4 range (except Nevada and Houston) have at least 6 losses. Kansas has more quality wins (Michigan St, Tennessee, Marquette, Villanova, Iowa St, Texas Tech) then any of those other teams. I am just glad RPI is done. With Kansas at #1 in those rankings I feel the committee would have a hard time placing them below a 2 seed because of that

Marquette only has 4 losses and has beat Wisconsin, Louisville, Kansas St, Buffalo (and now Villanova). I actually thought that inclusion of Marquette as a 3 and Villanova not in the top 16 at all was interesting. But then Marquette beat Nova yesterday. Marquette predictive rankings are around 30 so the committee is putting them a bit high based on W-L. But they do have 5 quality wins against a not horrible non-con schedule.

I agree that Nevada and Houston are hard to quantify. They only have 1 loss each. But Houston has only one win over a surefire NCAA team (LSU) while Nevada has none (Utah St and Arizona St are their best wins). KenPom and T-Rank both have them as about 4 seeds.

MChambers
02-10-2019, 11:55 AM
KP has nevada at 17. A 4 seed is reasonable
KP has houston at 19...the 3 seed is slightly high
KP has marquette at 29...a stupid overseed.

So marquette is the only real head scratcher here, especially since their SOS isn't particularly good either...but i'm quite happy if the committee wants to put them in our region :D :D

I stand corrected on Nevada, I guess. I was looking at T-Rank, which has them at 22, but I think I'm right on Houston and Marquette. Maybe time will tell.

DavidBenAkiva
02-10-2019, 11:55 AM
After all this discussion, we finally have some news, in that the NCAA released their as-of-today Top 16 seeds yesterday:

East Region: 1. Duke (1 overall), 2. Michigan (6), 3. Marquette (12), 4. Iowa St. (13)

South Region: 1. Tennessee (2), 2. North Carolina (7), 3. Purdue (9), 4. Nevada (14)

West Region: 1. Gonzaga (4), 2. Michigan St. (8), 3. Kansas (10), 4. Louisville (15)

Midwest Region: 1. Virginia (3), 2. Kentucky (5), 3. Houston (11), 4. Wisconsin (16)

Certainly, much of this will change between now and March, but it's still interesting to get a little snapshot of how the committee is thinking. Especially now that we have the NET rankings updating relatively frequently.

No major surprises in here, but one that jumps out at me is Kansas as the 10th overall team. Neither Kenpom nor Torvik have them that high, and even NET has them at #18 (as of 2/8). Seemingly, what props them up with the committee is their SOS. As ever, this is extremely disappointing to me. If you didn't have NET, and only looked at W-L records, then of course SOS would be relevant. But if you have NET, it doesn't make any sense to carve SOS out and give it such priority. SOS is already factored into the NET ratings!

Ok, rant over. Great win for Duke yesterday which keeps us above the fray, for now.

Thanks for putting this on here.

There are so many scenarios that could play out in the final month of the season. Here are a few things I think:

Write in pen that Gonzaga is going to get a 1 Seed. They are blowing out their conference opponents and may end up as the #1 overall seed if Duke stumbles down the stretch.

The ACC and the SEC will get at least 2 of the other 1 Seeds.

I would say that Duke has wrapped up a 1 Seed, but there are still a lot of tough games ahead, including the two against UNC, at Louisville, at Virginia Tech, and at Syracuse left to play. Meanwhile, Virginia or UNC could put themselves into position for another 1 Seed after the game on Monday. It would really help Duke if Virginia wins that game.

The SEC has two teams in contention with Tennessee and Kentucky. They will play each other as many as three times before the SEC Tournament is over. Tennessee is ripe for an upset in my opinion. They have won a ton of games but mostly against the bottom tier of teams in their conference. In the final month, they will start playing the better SEC teams, including Kentucky, LSU, Ole Miss, Mississippi State, and Auburn. Kentucky could vault them for a 1 Seed by splitting the regular season over Tennessee while the Vols drop a game or two in their other matchups.

The B1G still has several teams in contention for a 1 Seed in Michigan and Michigan State. The Wolverines have home-and-home matchups with MSU and Maryland yet ahead of them and a tough game at Minnesota. Meanwhile, after stopping their skid of 3 straight losses, MSU still has that home-and-home with Michigan and has a tough game at Wisconsin on Tuesday night. My guess is that both of the Michigan schools accumulate another pair of losses in conference play before the season ends while Purdue sneaks in with a soft schedule to grab the B1G regular season title. Given all that, I think the two Michigan schools and Purdue will end the year on the 2 and 3 Seed lines.

My prediction from today is that you'll see Duke, Gonzaga, Virginia, and Kentucky as 1 seeds with Tennessee, Michigan, North Carolina, and Purdue as 2-seeds. This is also a pretty good scenario for Duke.

East: (1) Duke, (2) Purdue
South: (1) Virginia, (2) Tennessee
Midwest: (1) Kentucky, (2) Michigan
West: (1) Gonzaga, (2) North Carolina

Nugget
02-11-2019, 04:14 PM
I agree with you on Kansas. I also find the relatively high seedings of Nevada, Houston, and Marquette discouraging. Reminiscent of the RPI.


I don't view this as RPI driven at all. In my observation, the Committee has tended to do the Top 4 seeds in a way that much closer tracks the AP poll, and rewards wins and losses, than it did the RPI or any metrics approach.

Nevada was highly ranked at the beginning of the season and remains so (#7 this week, and has just 1 loss, albeit a major clunker). Houston is something like 22-1 against a decent schedule and is ranked #9 in the AP Poll, and Marquette, while never a great computer team because of its iffy defense, has won alot of games, including a number of good non-con wins over Louisville, Wisconsin, Kansas St. and Buffalo and is #10 in the AP Poll.

So, those teams getting top 3-4 seeds is no mystery at all.

MChambers
02-11-2019, 04:58 PM
I don't view this as RPI driven at all. In my observation, the Committee has tended to do the Top 4 seeds in a way that much closer tracks the AP poll, and rewards wins and losses, than it did the RPI or any metrics approach.

Nevada was highly ranked at the beginning of the season and remains so (#7 this week, and has just 1 loss, albeit a major clunker). Houston is something like 22-1 against a decent schedule and is ranked #9 in the AP Poll, and Marquette, while never a great computer team because of its iffy defense, has won alot of games, including a number of good non-con wins over Louisville, Wisconsin, Kansas St. and Buffalo and is #10 in the AP Poll.

So, those teams getting top 3-4 seeds is no mystery at all.

That's a good way of looking at it. Having said that, we've discussed on these boards the problems with the polls, so it is discouraging that the Committee doesn't pay more attention to better measures of quality.

wavedukefan70s
02-11-2019, 05:30 PM
I believe Kentucky may shake things up a little this weekend.