Page 1 of 2 12 LastLast
Results 1 to 20 of 35
  1. #1
    scottdude8's Avatar
    scottdude8 is offline Contributor, Zoubek disciple, and resident Wolverine
    Join Date
    Jan 2010
    Location
    Toronto

    NET News following an upset-filled week (January 20)

    Figured if I'm going to keep doing these I might as well make them somewhat organized, haha... and it's definitely worth doing again since there were some major shakeups in the ratings this week following a bunch of upsets!

    First, in terms of the rankings themselves Duke comes in at No. 2. Some other interesting notes on that front:
    • UVA is at No. 1 despite the loss, which isn't all that surprising considering I don't believe the NET directly takes head-to-head into account (although I believe the committee does down the line).
    • The Top 6 matches the consensus "top-tier", with Gonzaga at No. 3, followed by Tennessee, Michigan State and Michigan. (Side note: I'm surprised that Michigan fell so far compared to Virginia, considering both of their first losses were tough road games, and Michigan played a superior non-conference slate I believe... this is one where I wish I could dig into the details of the algorithm to see what caused this).
    • Duke's big victories still look good, with Kentucky at No. 8, Texas Tech at 11, Auburn at 22, FSU at 33, and Indiana at 34. There's also a ton of room for more impressive wins with a lot of ACC teams at the top of the rankings.


    So it appears that there are really six teams separating themselves from the pack in the running for No. 1 seeds. How does this pan out on the team sheets? Well:
    • Duke still has 5 Q1 victories, and even those are high-caliber victories within the Q1 designation. Interestingly, if you look at the sheet they are now splitting things even within each "quadrant" into the top and bottom half, so this is a potentially significant differentiator. Our two losses are Q1 (Gonzaga) and Q2 (Syracuse).
    • Virginia, Tennessee, and Michigan all have 4 Q1 wins. Michigan State has 7, although 3 are "lower caliber" Q1 wins. Gonzaga has 3 Q1 wins and given their conference will probably be stuck there.
    • Kansas could prove to be a real test for the NET, because they have 7 Q1 wins yet are ranked 19. This is especially odd considering they're ranked higher than 19 in all of the other computer rankings, so there has to be something in the NET calculator that is really hurting Kansas. If this holds up it could be a very interesting question for the committee.
    • There are a few other teams with 4 Q1 wins (including UNC), but most other top teams either have 3 or less Q1 wins or are starting to pile up losses.


    The big takeaways: first, despite the Syracuse loss Duke is still in great shape for a No. 1 seed. Considering who we are going to still play, a Duke team that finishes with 6 or 7 losses will still have a major argument for a top seed considering who those victories will likely be against. Second, it's looking more and more unlikely that the Big 12 gets a top seed, and third it's likely going to be hard to give Gonzaga a No. 1 seed with only 3 Q1 wins alongside 2 losses. This means it's looking more and more likely that the ACC (or, in a less likely scenario, the B1G) will get 2 No. 1 seeds. So the rematch against UVA, alongside the two Michigan-Michigan State matchups (both of which happen very late in the year!) are going to have huge ramifications. But all of this combined means we essentially still control our own destiny with regards to seedings.

    This leads me to wonder if we should aim our sights higher, i.e. we should aim to get one of the "top two" seeds to avoid a bad 2 seed matchup. Given the clear break between the Top 6 teams and the rest of the NCAA, both qualitatively and quantitatively, I think Duke ideally would want to avoid one of of the borderline one seeds in their bracket.

    Interesting to see how people think the NET dealt with all the upsets this week/weekend, and debate on that final point!
    Last edited by scottdude8; 01-20-2019 at 04:38 PM.
    Scott Rich on the front page

    Trinity BS 2012; University of Michigan PhD 2018
    Duke Chronicle, Sports Online Editor: 2010-2012
    K-Ville Blue Tenting 2009-2012

    Unofficial Brian Zoubek Biographer
    If you have questions about Michigan Basketball/Football, I'm your man!

  2. #2
    Quote Originally Posted by scottdude8 View Post
    This leads me to wonder if we should aim our sights higher, i.e. we should be aiming to get one of the "top two" seeds to avoid a bad 2 seed matchup. Given the clear break between the Top 6 teams and the rest of the NCAA, both qualitatively and quantitatively, I think Duke ideally would want to avoid one of of the borderline one seeds in their bracket.
    Thanks for the analysis but the above comment isn't accurate. That is, the top one seeds don't necessarily get the bottom two seeds. It is based on geography only so as the top 1 seed we get the top geographic preference then when they get to the two line they also give that two seed geographic preference taking into consideration conference affiliation and other rules. They do some balancing on the three and four lines if necessary so that a particular region cannot have all the top 1 through 4 seeds. Note that I think they may have changed the rule last year where the very top one seed could not get the top two seed but somebody please validate.

  3. #3
    scottdude8's Avatar
    scottdude8 is offline Contributor, Zoubek disciple, and resident Wolverine
    Join Date
    Jan 2010
    Location
    Toronto
    Quote Originally Posted by Bluedog View Post
    Thanks for the analysis but the above comment isn't accurate. That is, the top one seeds don't necessarily get the bottom two seeds. It is based on geography only so as the top 1 seed we get the top geographic preference then when they get to the two line they also give that two seed geographic preference taking into consideration conference affiliation and other rules. They do some balancing on the three and four lines if necessary so that a particular region cannot have all the top 1 through 4 seeds. Note that I think they may have changed the rule last year where the very top one seed could not get the top two seed but somebody please validate.
    I wasn't 100% sure on this either, but if memory serves this is something they implemented last year that the snake gets factored into the decision making, even if it isn't the primary factor. In fact, I think a new rule last year was that the top overall seed couldn't get the top 2 seed, or something like that. So you're right that being one of the top two seeds won't guarantee us avoiding the top two seeds, but I believe it would give us a better chance at doing so.
    Scott Rich on the front page

    Trinity BS 2012; University of Michigan PhD 2018
    Duke Chronicle, Sports Online Editor: 2010-2012
    K-Ville Blue Tenting 2009-2012

    Unofficial Brian Zoubek Biographer
    If you have questions about Michigan Basketball/Football, I'm your man!

  4. #4
    Join Date
    Feb 2007
    Location
    Steamboat Springs, CO
    Quote Originally Posted by Bluedog View Post
    Thanks for the analysis but the above comment isn't accurate. That is, the top one seeds don't necessarily get the bottom two seeds. It is based on geography only so as the top 1 seed we get the top geographic preference then when they get to the two line they also give that two seed geographic preference taking into consideration conference affiliation and other rules. They do some balancing on the three and four lines if necessary so that a particular region cannot have all the top 1 through 4 seeds. Note that I think they may have changed the rule last year where the very top one seed could not get the top two seed but somebody please validate.
    Yes, I hate it when the NCAA changes the rules I worked so hard to learn. There is no longer an "S curve," whereby the strongest #1 gets the weakest #2 should form prevail. Isn't this so?
    Sage Grouse

    ---------------------------------------
    'When I got on the bus for my first road game at Duke, I saw that every player was carrying textbooks or laptops. I coached in the SEC for 25 years, and I had never seen that before, not even once.' - David Cutcliffe to Duke alumni in Washington, DC, June 2013

  5. #5
    scottdude8's Avatar
    scottdude8 is offline Contributor, Zoubek disciple, and resident Wolverine
    Join Date
    Jan 2010
    Location
    Toronto
    Quote Originally Posted by sagegrouse View Post
    Yes, I hate it when the NCAA changes the rules I worked so hard to learn. There is no longer an "S curve," whereby the strongest #1 gets the weakest #2 should form prevail. Isn't this so?
    This is from Wikipedia so take it with a grain of salt, but there are citations at least:
    A number of complex rules govern the seeding process, so it is not as simple as merely following the S-curve, although that is the top priority according to the NCAA's rules.[11] Better teams have priority in remaining close to home, but no hosting institution's team can actually play at the location where the institution is hosting tournament games (generally, games are hosted on neutral courts, so this is not usually a problem). Sometimes a top team may be a short drive away from its games; in 2006 Villanova played its first and second round games in Philadelphia at an arena where they had played three games that year, one fewer than the four required for a site to be considered a "home court" for a team, and in 2002 the Pitt Panthers played their first and second round games in the city of Pittsburgh at Mellon Arena (which was not their home court after the opening of their on-campus arena).
    Scott Rich on the front page

    Trinity BS 2012; University of Michigan PhD 2018
    Duke Chronicle, Sports Online Editor: 2010-2012
    K-Ville Blue Tenting 2009-2012

    Unofficial Brian Zoubek Biographer
    If you have questions about Michigan Basketball/Football, I'm your man!

  6. #6
    scottdude8's Avatar
    scottdude8 is offline Contributor, Zoubek disciple, and resident Wolverine
    Join Date
    Jan 2010
    Location
    Toronto
    Here's more details straight from the NCAA:

    2. The committee will then place the No. 2 seeds in each region in true seed list order. The committee may relax the principle of keeping teams as close to their area of natural interest for seeding teams on the No. 2 line to avoid, for example, the overall No. 5 seed being sent to the same region as the overall No. 1 seed. The committee will not compromise the principle of keeping teams from the same conference in separate regions.
    ...
    5. After the top four seed lines have been assigned, the committee will review the relative strengths of the regions by adding the “true seed” numbers in each region to determine if any severe numerical imbalance exists. Generally, no more than five points should separate the lowest and highest total.
    So the answer to the question about the S-curve is... maybe? Sometimes? If it's the third night of a blood moon and it rained on Selection Sunday?
    Scott Rich on the front page

    Trinity BS 2012; University of Michigan PhD 2018
    Duke Chronicle, Sports Online Editor: 2010-2012
    K-Ville Blue Tenting 2009-2012

    Unofficial Brian Zoubek Biographer
    If you have questions about Michigan Basketball/Football, I'm your man!

  7. #7
    scottdude8's Avatar
    scottdude8 is offline Contributor, Zoubek disciple, and resident Wolverine
    Join Date
    Jan 2010
    Location
    Toronto
    I thought it might be interesting to see how the NET is matching up with the AP Poll (maybe I'll wait to do these write-ups until Mondays in the future, haha). The "top 6" are consistent between the two rankings, as we would expect. From there things get a little wonky. For example:
    • The NET has Nevada at No. 22, while the AP has them at No. 7.
    • The NET has Kansas at No. 19, while the AP has them at No. 9.
    • The NET has Nebraska at No. 13, while they're unranked in the AP.
    • The NET has LSU at No. 14, while the AP has them at No. 25.
    • The NET has Purdue at No. 15, while they're unranked in the AP.

    All things considered those aren't complete dealbreakers, especially compared to the mistakes the RPI would make. But if the NET continues to overvalue B1G and SEC teams while undervaluing Big 12 teams that could lead to some interesting conversations come March (importantly for us, it could keep the Big 12 out of the one-seed conversation entirely, which is what it looks like things are trending towards, but also could help the argument for two No. 1 seeds out of the B1G as opposed to ACC).
    Scott Rich on the front page

    Trinity BS 2012; University of Michigan PhD 2018
    Duke Chronicle, Sports Online Editor: 2010-2012
    K-Ville Blue Tenting 2009-2012

    Unofficial Brian Zoubek Biographer
    If you have questions about Michigan Basketball/Football, I'm your man!

  8. #8
    Join Date
    Feb 2007
    Location
    Steamboat Springs, CO
    Quote Originally Posted by scottdude8 View Post
    I thought it might be interesting to see how the NET is matching up with the AP Poll (maybe I'll wait to do these write-ups until Mondays in the future, haha). The "top 6" are consistent between the two rankings, as we would expect. From there things get a little wonky. For example:
    • The NET has Nevada at No. 22, while the AP has them at No. 7.
    • The NET has Kansas at No. 19, while the AP has them at No. 9.
    • The NET has Nebraska at No. 13, while they're unranked in the AP.
    • The NET has LSU at No. 14, while the AP has them at No. 25.
    • The NET has Purdue at No. 15, while they're unranked in the AP.

    All things considered those aren't complete dealbreakers, especially compared to the mistakes the RPI would make. But if the NET continues to overvalue B1G and SEC teams while undervaluing Big 12 teams that could lead to some interesting conversations come March (importantly for us, it could keep the Big 12 out of the one-seed conversation entirely, which is what it looks like things are trending towards, but also could help the argument for two No. 1 seeds out of the B1G as opposed to ACC).
    Well, there's the SEC-Big 12 Challenge later this month that should provide some alignment data. Not sure how results are likely to affect NET.
    Sage Grouse

    ---------------------------------------
    'When I got on the bus for my first road game at Duke, I saw that every player was carrying textbooks or laptops. I coached in the SEC for 25 years, and I had never seen that before, not even once.' - David Cutcliffe to Duke alumni in Washington, DC, June 2013

  9. #9
    scottdude8's Avatar
    scottdude8 is offline Contributor, Zoubek disciple, and resident Wolverine
    Join Date
    Jan 2010
    Location
    Toronto
    Quote Originally Posted by sagegrouse View Post
    Well, there's the SEC-Big 12 Challenge later this month that should provide some alignment data. Not sure how results are likely to affect NET.
    That's an interesting point... I had forgotten that a lot of those teams still had one big non-conference game left on the slate. As a Duke fan, how do you root in that Challenge? Considering Tennessee would need to hit some major stumbling blocks to fall out of the running for a No. 1 seed (possible, given they haven't played their toughest SEC opponents yet, but it would take a lot at this point), I'm thinking that anything further devaluing the Big 12 (and further minimizing the likelihood Kansas gets itself back into the No. 1 seed discussion) is the best outcome for us. Unfortunately that might (blech) include (blech) rooting (blech) for (blech) Kentucky against KU...
    Scott Rich on the front page

    Trinity BS 2012; University of Michigan PhD 2018
    Duke Chronicle, Sports Online Editor: 2010-2012
    K-Ville Blue Tenting 2009-2012

    Unofficial Brian Zoubek Biographer
    If you have questions about Michigan Basketball/Football, I'm your man!

  10. #10
    Join Date
    Oct 2009
    Location
    Durham
    Quote Originally Posted by scottdude8 View Post
    I thought it might be interesting to see how the NET is matching up with the AP Poll (maybe I'll wait to do these write-ups until Mondays in the future, haha). The "top 6" are consistent between the two rankings, as we would expect. From there things get a little wonky. For example:
    • The NET has Nevada at No. 22, while the AP has them at No. 7.
    • The NET has Kansas at No. 19, while the AP has them at No. 9.
    • The NET has Nebraska at No. 13, while they're unranked in the AP.
    • The NET has LSU at No. 14, while the AP has them at No. 25.
    • The NET has Purdue at No. 15, while they're unranked in the AP.

    All things considered those aren't complete dealbreakers, especially compared to the mistakes the RPI would make. But if the NET continues to overvalue B1G and SEC teams while undervaluing Big 12 teams that could lead to some interesting conversations come March (importantly for us, it could keep the Big 12 out of the one-seed conversation entirely, which is what it looks like things are trending towards, but also could help the argument for two No. 1 seeds out of the B1G as opposed to ACC).
    comparing to kp

    NET AP KP
    nevada 22 7 19
    kansas 19 9 10
    nebraska 13 nr 12
    LSU 14 25 24
    Purdue 15 nr 9


    it's clear that some of the wonkies you point out actually aren't that wonky. nevada, nebraska and purdue are outliers of AP wrt KP, and kansas and LSU are outliers for NET.

    Either way, I'm not sure the relative ranking of teams in this range will have much impact on which of the top 6 or so teams gets a 1 seed.


    "If you don't address the things you're not doing well when you're winning the winning will eventually stop."

    -David Cutcliffe

  11. #11
    scottdude8's Avatar
    scottdude8 is offline Contributor, Zoubek disciple, and resident Wolverine
    Join Date
    Jan 2010
    Location
    Toronto
    Quote Originally Posted by uh_no View Post
    comparing to kp

    NET AP KP
    nevada 22 7 19
    kansas 19 9 10
    nebraska 13 nr 12
    LSU 14 25 24
    Purdue 15 nr 9


    it's clear that some of the wonkies you point out actually aren't that wonky. nevada, nebraska and purdue are outliers of AP wrt KP, and kansas and LSU are outliers for NET.

    Either way, I'm not sure the relative ranking of teams in this range will have much impact on which of the top 6 or so teams gets a 1 seed.
    Thanks for the KenPom insight! Considering that KP is a factor in the NET rankings it's not surprising that they correspond more with NET than the AP rankings do... so then the question becomes do we trust the analytics more than the media when it comes to ranking teams? I think most of us would lean towards the analytics considering how accurate KenPom has been over the past few years, but it's an interesting debate nonetheless.

    I do disagree with you on your final point though, and I think last year's Kansas team is the perfect example as to why. Despite having 7 losses, some of them objectively bad ones (i.e. at home or to inferior competition, or both!), Kansas still earned a No. 1 seed based largely on the quadrant system and their number of "Q1" wins. So if there ends up being a team with a high NET ranking that seems to be an outlier compared to the consensus opinion of the experts (coaches, media, etc.), that could end up helping a team that defeated said outlier come tourney time (and the reverse could obviously also occur). Perceived conference strength does seem to factor into seeding, whether the NCAA intends it to or not, so from that perspective these outliers could end up mattering come Selection Sunday. Obviously that's speculation, but I think it's informed speculation based on my research and recent history.
    Scott Rich on the front page

    Trinity BS 2012; University of Michigan PhD 2018
    Duke Chronicle, Sports Online Editor: 2010-2012
    K-Ville Blue Tenting 2009-2012

    Unofficial Brian Zoubek Biographer
    If you have questions about Michigan Basketball/Football, I'm your man!

  12. #12
    Quote Originally Posted by scottdude8 View Post
    Considering Tennessee would need to hit some major stumbling blocks to fall out of the running for a No. 1 seed (possible, given they haven't played their toughest SEC opponents yet, but it would take a lot at this point)...
    Every year at this time people around here say stuff like the above. I feel obligated to point out that teams that look like #1 seeds in January often don't end up that way.

    My favorite example is 2010, when many people on this board were arguing that Duke had no chance for a top seed because all four #1s were locked in. Especially Texas, the #1 team on January 21, 2010, which based on its schedule had almost no chance to drop from a #1. Or so they said. By tournament time, Texas was a #8-seed that lost in the first round to Wake Forest.

    In 2011 at this point, Syracuse was considered a shoo-in for a #1-seed. They ended up a #3 and lost to a #11-seed in the 2nd round. In 2012, people wondered whether Duke could beat out Baylor for a #1-seed. Neither of them made it, with Baylor ending up a #3. In 2014 at this time, people again considered Syracuse a lock for a #1, and again they ended up a #3 (and again they lost to a #11 in the 2nd round). Anyway, you probably get the pattern. Maybe Tennessee gets a #1 and maybe they don't. But it won't take "a lot" to knock them out, just the typical rough-and-tumble progression of a major conference schedule.

  13. #13
    Quote Originally Posted by scottdude8 View Post
    I think most of us would lean towards the analytics considering how accurate KenPom has been over the past few years...
    How do you measure "accuracy" in this context? I'm genuinely interested.

    Are you saying KenPom has accurately predicted seeding? Because a quick count told me (pre-tournament) KenPom was off by at least two seeds on 15 different teams in just the last two seasons, including stuff like saying Wichita State was the 8th best team in the country in 2017 (and thus should have been a 2-seed) when the Shockers got stuck as a #10.

    Or are you saying KenPom's top teams usually perform well in the tournament? Because (again) in the last two seasons, a full half (50%) of Pomeroy's top 8 teams (I chose top 8 because in theory that would be the teams deserving of #1 or #2 seeds) lost in the Round of 32 (and another three lost in the Sweet 16, meaning 69% of KenPom's top 8 teams lost before the Elite Eight in the past two seasons). For comparison, of the AP poll's top 8 teams (last poll before the NCAAT), only five teams lost in the first two rounds in the past two seasons (31.3%) and three more lost in the Sweet 16 (meaning 50% of AP's top 8 teams lost before the Elite Eight in the past two years).

    Or is it something else? Because if it's either of the above, I don't think KenPom is any more accurate than AP (and it appears less accurate).


    (and, sure, two years isn't that robust a sample; but KenPom's performance was so inaccurate in the two most recent years that I didn't feel the need to go back further)

  14. #14
    Join Date
    Oct 2009
    Location
    Durham
    Quote Originally Posted by Kedsy View Post
    How do you measure "accuracy" in this context? I'm genuinely interested.

    Are you saying KenPom has accurately predicted seeding? Because a quick count told me (pre-tournament) KenPom was off by at least two seeds on 15 different teams in just the last two seasons, including stuff like saying Wichita State was the 8th best team in the country in 2017 (and thus should have been a 2-seed) when the Shockers got stuck as a #10.

    Or are you saying KenPom's top teams usually perform well in the tournament? Because (again) in the last two seasons, a full half (50%) of Pomeroy's top 8 teams (I chose top 8 because in theory that would be the teams deserving of #1 or #2 seeds) lost in the Round of 32 (and another three lost in the Sweet 16, meaning 69% of KenPom's top 8 teams lost before the Elite Eight in the past two seasons). For comparison, of the AP poll's top 8 teams (last poll before the NCAAT), only five teams lost in the first two rounds in the past two seasons (31.3%) and three more lost in the Sweet 16 (meaning 50% of AP's top 8 teams lost before the Elite Eight in the past two years).

    Or is it something else? Because if it's either of the above, I don't think KenPom is any more accurate than AP (and it appears less accurate).


    (and, sure, two years isn't that robust a sample; but KenPom's performance was so inaccurate in the two most recent years that I didn't feel the need to go back further)
    That's a REALLY bad sampling error, especially when discarding things like the final margin and how close the respective poll/metrics said it should have been. Say one team is #1 in the poll, and the other is unranked. Say KP had the unranked team winning by 1. Now suppose the actual outcome was the unranked team lost by 1....who was closer? KP who was only off by a couple points in the final margin? Or the AP poll which vastly mis-represented at least one of the two teams? Obviously it depends on the context...but if your goal is to gain confidence in the predictor itself, the one that was off by 2 points is going to get my vote over the one that was off by 20 (?), even if in this case the victor was correct...since the former was far more accurate at correctly evaluating the relative strength of the two teams.


    Anyway, the point I was trying to make wasn't that we should kow-tow to KP as the be all end all of a proper ranking system, but that NET's results are not all that wonky, and seem to align in most cases with at least some other reasonable metrics, which might be AP or KP or whatever. KP was just a proxy for "some other metric that is largely reasonable."

    This is in contrast to RPI which seemed to sometimes produce results which were misaligned with every other metric under the sun...sometimes by huuuge margins.


    As I've mentioned before, I'd love to run some regressions on stuff like polls, seeds, KP, and "sum of KP rankings" to see how they actually stack up over time...but alas, time. Maybe if I get bored over the summer.


    "If you don't address the things you're not doing well when you're winning the winning will eventually stop."

    -David Cutcliffe

  15. #15
    Quote Originally Posted by uh_no View Post
    That's a REALLY bad sampling error, especially when discarding things like the final margin and how close the respective poll/metrics said it should have been. Say one team is #1 in the poll, and the other is unranked. Say KP had the unranked team winning by 1. Now suppose the actual outcome was the unranked team lost by 1...who was closer? KP who was only off by a couple points in the final margin? Or the AP poll which vastly mis-represented at least one of the two teams? Obviously it depends on the context...but if your goal is to gain confidence in the predictor itself, the one that was off by 2 points is going to get my vote over the one that was off by 20 (?), even if in this case the victor was correct...since the former was far more accurate at correctly evaluating the relative strength of the two teams.


    Anyway, the point I was trying to make wasn't that we should kow-tow to KP as the be all end all of a proper ranking system, but that NET's results are not all that wonky, and seem to align in most cases with at least some other reasonable metrics, which might be AP or KP or whatever. KP was just a proxy for "some other metric that is largely reasonable."

    This is in contrast to RPI which seemed to sometimes produce results which were misaligned with every other metric under the sun...sometimes by huuuge margins.


    As I've mentioned before, I'd love to run some regressions on stuff like polls, seeds, KP, and "sum of KP rankings" to see how they actually stack up over time...but alas, time. Maybe if I get bored over the summer.
    I understand the value of analytics in evaluating team performance. You know I do. I wasn't responding to your analysis, which I understood and which made sense to me.

    The question (by another poster, not you) appeared to be when it comes to ranking the teams for the purposes of seeding, which has been more "accurate," in relation to NCAAT performance? Obviously, it's hard to answer that question without knowing the definition of "accurate." If it's evaluating predicted scores vs. actual scores (as you suggest) there might be a different answer than in evaluating future wins vs. losses (which appears to me to be all anyone really cares about in the NCAAT). And that's before considering the interdependence of seeding with winning (i.e., do better seeds win because they're better teams or because they're playing worse teams?).

    Ultimately, systems like KenPom seem like a reasonable way to evaluate and compare the relative merit of college basketball teams. That's why I refer to KenPom often when we have conversations about evaluating and comparing NCAA teams. But the NCAA tournament often has little to do with which teams are objectively better. No system does that good a job in predicting who will win, lose, underperform, or overperform expectations in the one-and-done tourney.

    And that was the point I was trying to make. Sorry if I didn't make it so well.

  16. #16
    Join Date
    Oct 2009
    Location
    Durham
    Quote Originally Posted by Kedsy View Post
    I understand the value of analytics in evaluating team performance. You know I do. I wasn't responding to your analysis, which I understood and which made sense to me.

    The question (by another poster, not you) appeared to be when it comes to ranking the teams for the purposes of seeding, which has been more "accurate," in relation to NCAAT performance? Obviously, it's hard to answer that question without knowing the definition of "accurate." If it's evaluating predicted scores vs. actual scores (as you suggest) there might be a different answer than in evaluating future wins vs. losses (which appears to me to be all anyone really cares about in the NCAAT). And that's before considering the interdependence of seeding with winning (i.e., do better seeds win because they're better teams or because they're playing worse teams?).

    Ultimately, systems like KenPom seem like a reasonable way to evaluate and compare the relative merit of college basketball teams. That's why I refer to KenPom often when we have conversations about evaluating and comparing NCAA teams. But the NCAA tournament often has little to do with which teams are objectively better. No system does that good a job in predicting who will win, lose, underperform, or overperform expectations in the one-and-done tourney.

    And that was the point I was trying to make. Sorry if I didn't make it so well.
    I'm sure I'm on the bad end of the misinterpretation I usually am.


    "If you don't address the things you're not doing well when you're winning the winning will eventually stop."

    -David Cutcliffe

  17. #17
    scottdude8's Avatar
    scottdude8 is offline Contributor, Zoubek disciple, and resident Wolverine
    Join Date
    Jan 2010
    Location
    Toronto
    Quote Originally Posted by Kedsy View Post
    How do you measure "accuracy" in this context? I'm genuinely interested.

    Are you saying KenPom has accurately predicted seeding? Because a quick count told me (pre-tournament) KenPom was off by at least two seeds on 15 different teams in just the last two seasons, including stuff like saying Wichita State was the 8th best team in the country in 2017 (and thus should have been a 2-seed) when the Shockers got stuck as a #10.

    Or are you saying KenPom's top teams usually perform well in the tournament? Because (again) in the last two seasons, a full half (50%) of Pomeroy's top 8 teams (I chose top 8 because in theory that would be the teams deserving of #1 or #2 seeds) lost in the Round of 32 (and another three lost in the Sweet 16, meaning 69% of KenPom's top 8 teams lost before the Elite Eight in the past two seasons). For comparison, of the AP poll's top 8 teams (last poll before the NCAAT), only five teams lost in the first two rounds in the past two seasons (31.3%) and three more lost in the Sweet 16 (meaning 50% of AP's top 8 teams lost before the Elite Eight in the past two years).

    Or is it something else? Because if it's either of the above, I don't think KenPom is any more accurate than AP (and it appears less accurate).


    (and, sure, two years isn't that robust a sample; but KenPom's performance was so inaccurate in the two most recent years that I didn't feel the need to go back further)
    Good catch, because I definitely wasn't clear in my point here! What I was going at wasn't about seeding accuracy but more about what do we think is the most accurate rankings as far as overall team quality. I was going a bit from memory, but I remember in that in the early years of KenPom there was something along the lines of the National Champion ended up as No. 1 or No. 2 in his final rankings a disproportionate amount of the time (again, I'm going from memory here, I'm sure someone can correct me with what the actual trend was), and that was how KenPom sort of made his name.

    So I was probably reaching a bit with that statement, but the general thing I was going for was that analytics like KenPom have done a better job in the recent past at identifying the real NC contenders than things like the poll which are skewed by the eye test, the names of the schools, etc. Again, I could be completely wrong on that front because I'm operating from memory, haha.
    Scott Rich on the front page

    Trinity BS 2012; University of Michigan PhD 2018
    Duke Chronicle, Sports Online Editor: 2010-2012
    K-Ville Blue Tenting 2009-2012

    Unofficial Brian Zoubek Biographer
    If you have questions about Michigan Basketball/Football, I'm your man!

  18. #18
    Quote Originally Posted by scottdude8 View Post
    Good catch, because I definitely wasn't clear in my point here! What I was going at wasn't about seeding accuracy but more about what do we think is the most accurate rankings as far as overall team quality. I was going a bit from memory, but I remember in that in the early years of KenPom there was something along the lines of the National Champion ended up as No. 1 or No. 2 in his final rankings a disproportionate amount of the time (again, I'm going from memory here, I'm sure someone can correct me with what the actual trend was), and that was how KenPom sort of made his name.

    So I was probably reaching a bit with that statement, but the general thing I was going for was that analytics like KenPom have done a better job in the recent past at identifying the real NC contenders than things like the poll which are skewed by the eye test, the names of the schools, etc. Again, I could be completely wrong on that front because I'm operating from memory, haha.
    Your memory is not wrong (though the people who said that stuff about Pomeroy were). For years, people have claimed such things about KenPom, but when they did they almost always were talking about post-tournament rankings.

    Here's a table of the champions during the 17 years of KenPom, with post-tourney and pre-tourney Pomeroy, along with pre-tourney AP:

    Code:
    Year	Champion	PostKP	PreKP	PreAP	Seed
    2018	Villanova	1	2	2	1
    2017	UNC		3	3	6	1
    2016	Villanova	1	5	3	2
    2015	Duke		3	6	2	1
    2014	UConn		15	25	21	7
    2013	Louisville	1	2	4	1
    2012	Kentucky	1	1	1	1
    2011	UConn		10	15	21	3
    2010	Duke		1	2	4	1
    2009	UNC		1	3	1	1
    2008	Kansas		1	1	4	1
    2007	Florida		2	3	6	1
    2006	Florida		1	6	11	3
    2005	UNC		1	2	2	1
    2004	UConn		2	5	7	2
    2003	Syracuse	8	20	13	3
    2002	Maryland	3	4	4	1
    You probably read what you did around 2010 (in those days the 2002 and 2003 data was not available on KenPom's site, i.e., every year from 2004 to 2010 the national champ was #1 or #2 in KP's post-tourney rankings). Over the 17 years of (now) available KenPom stats, pre-tournament, the eventual national champion was #1 or #2 in only six of the 17 seasons (35%). For comparison, in five of the 17 seasons (29%) the eventual champion was #1 or #2 in the final (pre-T) AP rankings. So Pomeroy is a little more "accurate" but not much.

    If you look at how far away from #1 the pre-T rankings were (e.g., in 2010 Duke was #4, and that's 3 away from #1), the KenPom pre-T rankings are again a little more accurate than AP but not much (a total of 7 closer, over 17 years). Like I said in an earlier post, no system is very good at predicting the results of the NCAA tournament.

    Note also that 10 of the champions were in KenPom's top 4 (59%), and 10 were in the AP's top 4 (also 59%), but 11 of the champs (65%) were #1 seeds.


    .
    Last edited by Kedsy; 01-21-2019 at 04:47 PM. Reason: Added the seeding stuff

  19. #19
    Quote Originally Posted by Kedsy View Post
    Every year at this time people around here say stuff like the above. I feel obligated to point out that teams that look like #1 seeds in January often don't end up that way.

    My favorite example is 2010, when many people on this board were arguing that Duke had no chance for a top seed because all four #1s were locked in. Especially Texas, the #1 team on January 21, 2010, which based on its schedule had almost no chance to drop from a #1. Or so they said. By tournament time, Texas was a #8-seed that lost in the first round to Wake Forest.

    In 2011 at this point, Syracuse was considered a shoo-in for a #1-seed. They ended up a #3 and lost to a #11-seed in the 2nd round. In 2012, people wondered whether Duke could beat out Baylor for a #1-seed. Neither of them made it, with Baylor ending up a #3. In 2014 at this time, people again considered Syracuse a lock for a #1, and again they ended up a #3 (and again they lost to a #11 in the 2nd round). Anyway, you probably get the pattern. Maybe Tennessee gets a #1 and maybe they don't. But it won't take "a lot" to knock them out, just the typical rough-and-tumble progression of a major conference schedule.
    It's too early for my yearly "definition of 'lock'" rant.

  20. #20
    Join Date
    Jul 2008
    Location
    Nashville
    Quote Originally Posted by Kedsy View Post

    Note also that 10 of the champions were in KenPom's top 4 (59%), and 10 were in the AP's top 4 (also 59%), but 11 of the champs (65%) were #1 seeds.


    .
    This is the most relavent point, to me. We want a number 1 seed. Number 1 seeds have the easiest path to the Final Four. Falling to a 2 seed definitely hurt us in 2013 (when perhaps we could have avoided Louisville until the FF) and in 2017 (when we could’ve avoided a semi-away game against SC in the second round). It’s also why I view games like tonight’s matchup against Pitt as a must-win. It certainly doesn’t take us out of the running for a top seed if we stumble, but our path is a lot easier if we take care of business.

Similar Threads

  1. Alright..so is the RBC going to filled with the Duke Fans
    By jimrowe0 in forum Elizabeth King Forum
    Replies: 15
    Last Post: 01-20-2010, 04:39 PM
  2. Reggie: ABC News Person of the Week
    By jhole in forum Elizabeth King Forum
    Replies: 14
    Last Post: 01-25-2009, 12:18 PM
  3. Replies: 20
    Last Post: 11-03-2007, 05:56 AM
  4. I Am Filled With Anger!
    By EarlJam in forum Off Topic
    Replies: 11
    Last Post: 08-27-2007, 05:40 PM
  5. My bracket is finally all filled out
    By Channing in forum Elizabeth King Forum
    Replies: 2
    Last Post: 03-15-2007, 02:53 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •