MBB: Bracketology 2019-20

[Kedsy]

But... One seeds are better teams and are expected to advance regardless of their path. If two seeds were beating one seeds regularly in the Elite Eight, it would be very surprising.

Since 1985, 1-seeds have played 2-seeds 47 times. The #1 has won 23 and lost 24 of those games. So perhaps not regularly, but we're certainly not seeing #1s clobber #2s if they're so much better (in fact we're not even seeing them win more than they lose).

[HereBeforeCoachK]

Since 1985, 1-seeds have played 2-seeds 47 times. The #1 has won 23 and lost 24 of those games. So perhaps not regularly, but we're certainly not seeing #1s clobber #2s if they're so much better (in fact we're not even seeing them win more than they lose).

That's interesting that it's that close....and of course, in one region, when a 1 and a 2 play, theoretically - that's national #4 versus national #5 - so that's virtually the same seed.

[npdevil27]

Since 1985, 1-seeds have played 2-seeds 47 times. The #1 has won 23 and lost 24 of those games. So perhaps not regularly, but we're certainly not seeing #1s clobber #2s if they're so much better (in fact we're not even seeing them win more than they lose).

How many of those 1 vs 2 games were in the same region, and what is the record in those games? In other words, how much of the 2 to 1 difference of 1 seeds making it vs. 2 seeds is explained (or not explained) by this data point?

[uh_no]

Yeah, but it’s impossible to quantify how much of that is due to 1 seeds having an easier path (which is all that matters when you’re talking about the same team getting a 1 vs a 2 seed) and how much is due to 1 seeds generally being better teams than 2 seeds, which is why they got a 1 in the first place.

i disagree that it's impossible to quantify. There are plenty of models and at least 20 years worth of objective data that would allow you to make relatively accurate estimations about probability of reaching the final four were 1 and 2 seed paths swapped.

It's something I've wanted to do for some time, but unlikely to have time to do in the near future.

In the end it's probably some of both, though I suspect given that there will be a tail on a distribution, your average 3 and 4 seed, and 7 and 8 seed will be closer in ability than your average 1 and 2 seed...then of course the 2 seed would be expected to lose to the 1 seed anyway...so you would want to look at the probability of reaching the elite 8.

https://sports.cbsimg.net/images/col...1seedguide.pdf

There is a huge amount of data incorporated here, the most important points:

* support for my above argument that there is a long tail at the top of the distribution
* 1v2 is 18-18, which is weak evidence that the two are equivalent, but it is not the only explanation. IF, for instance, we accept that there are 2 seeds which are equivalent to 1 seeds in quality, than those 2 seeds will be disproportionately expected to reach the E8 vs an average 2 seed, and also would be expected to perform better than an average 2 seed. The 2 seeds, oddly enough, perform WORSE than expected against all other seeds, which is somewhat contradictory.

In any case, I saw Kedsy's response, even though I started typing this before he committed it, and I think my "alternate scenario" in the second bullet point points out one of the flaws of using JUST rudimentary aggregated data, as there are clearly confounding variables that lead to seemingly contradictory data.

That being the case, short of actually simulating tournaments swapping 1 and 2 seeds and observing potential of reaching the elite 8, I'll be unconvinced of arguments that say much more than "the actual ability of the 1/2 seed and the average difficult of the path both likely play some role in determining likelihood to reach the elite 8," which I think is hardly a controversial statement.

[Phredd3]

That being the case, short of actually simulating tournaments swapping 1 and 2 seeds and observing potential of reaching the elite 8, I'll be unconvinced of arguments that say much more than "the actual ability of the 1/2 seed and the average difficult of the path both likely play some role in determining likelihood to reach the elite 8," which I think is hardly a controversial statement.

A lot of words there, but in the end I've got you down for, "It's theoretically possible to quantify, but nobody has actually successfully quantified it." Is that about right?

[SCMatt33]

i disagree that it's impossible to quantify. There are plenty of models and at least 20 years worth of objective data that would allow you to make relatively accurate estimations about probability of reaching the final four were 1 and 2 seed paths swapped.

It's something I've wanted to do for some time, but unlikely to have time to do in the near future.

In the end it's probably some of both, though I suspect given that there will be a tail on a distribution, your average 3 and 4 seed, and 7 and 8 seed will be closer in ability than your average 1 and 2 seed...then of course the 2 seed would be expected to lose to the 1 seed anyway...so you would want to look at the probability of reaching the elite 8.

https://sports.cbsimg.net/images/col...1seedguide.pdf

There is a huge amount of data incorporated here, the most important points:

* support for my above argument that there is a long tail at the top of the distribution
* 1v2 is 18-18, which is weak evidence that the two are equivalent, but it is not the only explanation. IF, for instance, we accept that there are 2 seeds which are equivalent to 1 seeds in quality, than those 2 seeds will be disproportionately expected to reach the E8 vs an average 2 seed, and also would be expected to perform better than an average 2 seed. The 2 seeds, oddly enough, perform WORSE than expected against all other seeds, which is somewhat contradictory.

In any case, I saw Kedsy's response, even though I started typing this before he committed it, and I think my "alternate scenario" in the second bullet point points out one of the flaws of using JUST rudimentary aggregated data, as there are clearly confounding variables that lead to seemingly contradictory data.

That being the case, short of actually simulating tournaments swapping 1 and 2 seeds and observing potential of reaching the elite 8, I'll be unconvinced of arguments that say much more than "the actual ability of the 1/2 seed and the average difficult of the path both likely play some role in determining likelihood to reach the elite 8," which I think is hardly a controversial statement.

You can come up with lots of models. KenPom has a very good log5 analysis that he runs every year that give predictions on a percentage basis for each round. Swapping a team from a 1 to a 2 seed would give you a good idea of how much it would theoretically help or hurt to move up or down. Those things don’t and can never factor in specific circumstances like how well you match up with a certain style of play, or what upsets happen to help or hurt your path to get there. There’s some amount of advantage in general to being a 1 seed for the same team than if they got a 2 seed.

What you can’t quantify with any reasonable accuracy is how much this particular group of players in this particular year will be helped or hurt by getting a 1 or a 2 seed. If you could, there’d be someone making crap tons of money every year on their brackets, but factors that aren’t known on selection Sunday play a huge role in the actual outcomes. Something like KenPom gives you a good rough estimate and would be much more accurate than simply citing the difference in how many 1 seeds vs 2 seeds have made the FF, but it’s not and can’t be perfect, as a future game of basketball played by 18-22 year old kids is not a deterministic event.

It’s silly to see people argue that you aren’t better off getting a 1 seed. You absolutely are, and every single time I’ve commented over the years I’ve always been on the side of 1 seed in a less preferred location. But it’s just as silly when year in and year out, the percentage of one seed/ making the final four vs the percentage of two seeds gets thrown out there as evidence of how much of an advantage it is whe Duke is bringing the same roster either way.

[uh_no]

A lot of words there, but in the end I've got you down for, "It's theoretically possible to quantify, but nobody has actually successfully quantified it." Is that about right?

I would say "The kind of people who are interested in quantifying it haven't done so."

[uh_no]

You can come up with lots of models. KenPom has a very good log5 analysis that he runs every year that give predictions on a percentage basis for each round. Swapping a team from a 1 to a 2 seed would give you a good idea of how much it would theoretically help or hurt to move up or down. Those things don’t and can never factor in specific circumstances like how well you match up with a certain style of play, or what upsets happen to help or hurt your path to get there. There’s some amount of advantage in general to being a 1 seed for the same team than if they got a 2 seed.

What you can’t quantify with any reasonable accuracy is how much this particular group of players in this particular year will be helped or hurt by getting a 1 or a 2 seed. If you could, there’d be someone making crap tons of money every year on their brackets, but factors that aren’t known on selection Sunday play a huge role in the actual outcomes. Something like KenPom gives you a good rough estimate and would be much more accurate than simply citing the difference in how many 1 seeds vs 2 seeds have made the FF, but it’s not and can’t be perfect, as a future game of basketball played by 18-22 year old kids is not a deterministic event.

It’s silly to see people argue that you aren’t better off getting a 1 seed. You absolutely are, and every single time I’ve commented over the years I’ve always been on the side of 1 seed in a less preferred location. But it’s just as silly when year in and year out, the percentage of one seed/ making the final four vs the percentage of two seeds gets thrown out there as evidence of how much of an advantage it is whe Duke is bringing the same roster either way.

I largely agree with all that.

I can't imagine arguing in general that one would prefer a 2 seed to a 1 seed. There may be specific years and specific brackets where you look at it and say that being a 1 vs 2 might be a wash, or preferable to a 2. But the degree to which I see it referred to as a death knell is far overstated when you consider the factors that have often lead to duke ending up with a 2 seed. We can go back looking:

2018: we made it to the elite 8, where we lost to kansas, who we presumably would have lost to anyway were we the 1 and they the 2. Kansas had a far harder road anyway, with #26 SHU, and #14 clemson, vs #52 URI and #41 syracuse.
2017: duke deserved every bit of the 2 seed, and lost to a USC team that went to the final four, to whom duke would have presumably lost to as the 1 seed 2 rounds later
2013: Duke lost in the elite 8 to #1 seed louisville, whom we would have presumably lost to anyway as the 1 seed. I would buy the argument that perhaps UL is more rested, and I'm sure more than a few of you would make that case, but it doesn't change the fact that the two seed didn't materially impact our making the elite 8, where we had our chance to beat a team we would have played anyway, and didn't.
2012: duke didn't play in the tournament as the 2 seed
2009: duke was a 2 seed that lost to a final-four villanova team
2008: UCLA played #21 TX A&M in the second round, and we played #18 UWV. There was ~1 point difference between the two, but certainly not enough for me to believe we would have been any more or less successful against the former vs the latter, especially given our performance against belmont. Sure it's not the 300+ rank that UCLA faced, but you're going to have a tough time convincing me that if a team has trouble beating a 100+ team in the first round that they have the 2 seed to blame for not making the final 4.


So of the 6 times we've had a 2 seed:
4 times we lost to the team that went to the final four from that bracket.
1 Of those times the more difficult road of reaching that team may have affected our chances (though i'm not sure that an easier path would have made quinn more able to beat the press, or us more able to defend the PNR...and we lost by 22)
2 times the team was not good enough for the matchup they ended up with, and could have reasonably gotten on the other side of the bracket.

So out of 6 times, only once is there a case that getting a 2 instead of a 1 might be something we could blame...and even that is tenuous at best.

[House P]

i disagree that it's impossible to quantify. There are plenty of models and at least 20 years worth of objective data that would allow you to make relatively accurate estimations about probability of reaching the final four were 1 and 2 seed paths swapped.

It's something I've wanted to do for some time, but unlikely to have time to do in the near future.

In the end it's probably some of both, though I suspect given that there will be a tail on a distribution, your average 3 and 4 seed, and 7 and 8 seed will be closer in ability than your average 1 and 2 seed...then of course the 2 seed would be expected to lose to the 1 seed anyway...so you would want to look at the probability of reaching the elite 8.

One way of looking at the benefit of getting a 1 seed vs a 2 seed is to use KenPom ratings to estimate a the likelihood of a team making the Final 4 based on the seed a team receives within a particular bracket.

To get a sense for this, I estimated the likelihood of a team with an Adjusted Efficiency Margin (AdjEM) of 27.64 making the Final 4 against an "average" bracket based on their seed. I chose an AdjEM of 27.64 because that is Duke's current KemPom rating.

First of all, here is the average pre-tourney AdjEM by seed for the past 10 years.

Seed
Avg AdjEM

1
29.2

2
26.0

3
23.0

4
22.4

5
20.4

6
18.2

7
17.8

8
16.9

9
15.7

10
15.6

11
14.9

12
13.0

13
9.1

14
7.2

15
3.0

16
-3.1

If you put a team with Duke's current AdjEM into this "average" bracket, their likelihood of making the Final 4 would drop would drop from 33.3% as a 1 seed to 28.0% as a 2 seed and 24.2% as a 3 seed. Further details regarding the likelihood of advancing to a particular round according to seed are provided in the table below.

Seed
Advance to R32
Advance to R16
Advance to R8
Advance to Final4

1
97.9%
76.7%
52.9%
33.3%

2
94.8%
72.2%
49.8%
28.0%

3
91.2%
69.5%
43.1%
24.2%

4
88.9%
65.2%
34.4%
21.7%

5
83.3%
57.3%
30.2%
19.1%

6
80.1%
53.4%
33.1%
18.6%

7
78.7%
45.2%
31.2%
17.5%

8
78.5%
36.8%
25.4%
16.0%

Of course, as others have pointed out, Duke isn't going to face an "average" bracket. The actual probabilities will vary a fair amount depending on the actual bracket. For example, there are certainly scenarios where it may be better to be a 2 seed in a bracket with relatively weak 1 and 3 seeds, than to be a 1 seed in a bracket with strong 2, 4, and/or 5 seeds. But, on average, it is better to be a 1 seed that a 2 seed.

[sagegrouse]

Since 1985, 1-seeds have played 2-seeds 47 times. The #1 has won 23 and lost 24 of those games. So perhaps not regularly, but we're certainly not seeing #1s clobber #2s if they're so much better (in fact we're not even seeing them win more than they lose).

Kedsy: I think I am coming into the middle of a discussion and may have missed some of the argument, but, even so, don't #1 seeds have a far higher probability of reaching the Final Four?

In 35 years, according to NCAA.com, 57 of 140 #1 seeds (41 percent) have reached the Final Four as compared to 29 #2 seeds (21).

Of the champions, 22 have been #1 seeds (63 percent), or 15.7 percent sample probability for each #1 of 140 seed. Five #2 seeds have become NCAA champs, including Duke in 1991.

Perhaps a supporting point to your argument could be -- Any team that wins three NCAA games and makes the Elite Eight is pretty darned good, so the #2's, although fewer making the Elite Eight, are pretty formidable.

[uh_no]

If you put a team with Duke's current AdjEM into this "average" bracket, their likelihood of making the Final 4 would drop would drop from 33.3% as a 1 seed to 28.0% as a 2 seed and 24.2% as a 3 seed.

I would have guessed slightly more than that, but it validates the thesis that it's not a death-knell and the intrinsic ability of the team is far more impactful.

[Kedsy]

Kedsy: I think I am coming into the middle of a discussion and may have missed some of the argument, but, even so, don't #1 seeds have a far higher probability of reaching the Final Four?

In 35 years, according to NCAA.com, 57 of 140 #1 seeds (41 percent) have reached the Final Four as compared to 29 #2 seeds (21).

Of the champions, 22 have been #1 seeds (63 percent), or 15.7 percent sample probability for each #1 of 140 seed. Five #2 seeds have become NCAA champs, including Duke in 1991.

Perhaps a supporting point to your argument could be -- Any team that wins three NCAA games and makes the Elite Eight is pretty darned good, so the #2's, although fewer making the Elite Eight, are pretty formidable.

Yes, you're coming into the middle. This conversation started with me saying #1 seeds make the Final Four twice as much as #2 seeds. Other people said that was largely because #1 seeds are better than #2 seeds and one person said, "If two seeds were beating one seeds regularly in the Elite Eight, it would be very surprising." And I responded with the post you quoted, saying that #2 seeds do in fact win about 50% (a little better) of their games with #1 seeds.

I largely agree with all that.

I can't imagine arguing in general that one would prefer a 2 seed to a 1 seed. There may be specific years and specific brackets where you look at it and say that being a 1 vs 2 might be a wash, or preferable to a 2. But the degree to which I see it referred to as a death knell is far overstated when you consider the factors that have often lead to duke ending up with a 2 seed. We can go back looking:

2018: we made it to the elite 8, where we lost to kansas, who we presumably would have lost to anyway were we the 1 and they the 2. Kansas had a far harder road anyway, with #26 SHU, and #14 clemson, vs #52 URI and #41 syracuse.
2017: duke deserved every bit of the 2 seed, and lost to a USC team that went to the final four, to whom duke would have presumably lost to as the 1 seed 2 rounds later
2013: Duke lost in the elite 8 to #1 seed louisville, whom we would have presumably lost to anyway as the 1 seed. I would buy the argument that perhaps UL is more rested, and I'm sure more than a few of you would make that case, but it doesn't change the fact that the two seed didn't materially impact our making the elite 8, where we had our chance to beat a team we would have played anyway, and didn't.
2012: duke didn't play in the tournament as the 2 seed
2009: duke was a 2 seed that lost to a final-four villanova team
2008: UCLA played #21 TX A&M in the second round, and we played #18 UWV. There was ~1 point difference between the two, but certainly not enough for me to believe we would have been any more or less successful against the former vs the latter, especially given our performance against belmont. Sure it's not the 300+ rank that UCLA faced, but you're going to have a tough time convincing me that if a team has trouble beating a 100+ team in the first round that they have the 2 seed to blame for not making the final 4.


So of the 6 times we've had a 2 seed:
4 times we lost to the team that went to the final four from that bracket.
1 Of those times the more difficult road of reaching that team may have affected our chances (though i'm not sure that an easier path would have made quinn more able to beat the press, or us more able to defend the PNR...and we lost by 22)
2 times the team was not good enough for the matchup they ended up with, and could have reasonably gotten on the other side of the bracket.

So out of 6 times, only once is there a case that getting a 2 instead of a 1 might be something we could blame...and even that is tenuous at best.

I mostly disagree with your logic here, though I disagree more or disagree less depending on the year:

2018: Here, I think you're probably right. If we were a #1 we probably would have flipped with Kansas and played the same Elite Eight game.

2017: It's possible you're right, but I think it's more likely that if Duke and South Carolina had met in the Elite Eight it would have been an entirely different game. For one thing, the game wouldn't have been played in South Carolina. For another, if Duke was in the Elite Eight, the team may have been less tired and more confident (and perhaps less overconfident) than they played in their Round of 32 game with South Carolina. Duke had just won the ACC tournament by winning four games in four days. We were winning by 7 at halftime against USC. Then we just ran out of gas. I guess it's possible the same thing happens in the Elite Eight, but more likely it would have been an entirely different game. Even likelier, South Carolina wouldn't have made the Elite Eight if not for the burst of confidence they felt after beating Duke.

2013: Totally disagree with you here. There was no way Louisville would have been a #2 that year. If Duke were a #1, it would have been at Kansas's expense. Kansas's #2 in 2013 was Georgetown, which lost in the 1st round. Kansas lost to #4 Michigan, which Duke may have also lost to, but there's no evidence that we would have lost to Michigan, so I think your logic falls apart for 2013.

2012: I don't know what you mean that we didn't play in the tournament as a 2-seed. We were a #2 and lost to a tough #15. If we were a #1 that season, it's very, very unlikely we lose to a #16. The issue then becomes how far the admittedly flawed team would have gone after that.

2009: This one's a tossup. Like 2017, if Duke was a #1 and reached the Elite Eight, there's no telling whether Villanova would have also been in the Elite Eight or whether Duke would have played such a flat game with the Final Four on the line. Obviously, no telling whether Duke would have come out equally flat in the Sweet 16 against #4 Xavier or #12 Arizona (teams that #1 Pittsburgh and #1 Louisville played in the Sweet 16 that year), or if Duke had replaced Louisville whether we could have beaten Michigan State (as Louisville didn't).


Overall, I don't agree you can say, "we lost to a team that went to the Final Four so we would have lost anyway." The tournament is just way more complicated than that.

[HereBeforeCoachK]

2017: It's possible you're right, but I think it's more likely that if Duke and South Carolina had met in the Elite Eight it would have been an entirely different game. For one thing, the game wouldn't have been played in South Carolina. For another, if Duke was in the Elite Eight, the team may have been less tired and more confident (and perhaps less overconfident) than they played in their Round of 32 game with South Carolina. Duke had just won the ACC tournament by winning four games in four days. We were winning by 7 at halftime against USC. Then we just ran out of gas. I guess it's possible the same thing happens in the Elite Eight, but more likely it would have been an entirely different game. Even likelier, South Carolina wouldn't have made the Elite Eight if not for the burst of confidence they felt after beating Duke.


Overall, I don't agree you can say, "we lost to a team that went to the Final Four so we would have lost anyway." The tournament is just way more complicated than that.

Everything in bold is so so so on the money.

[brevity]

Since 1985, 1-seeds have played 2-seeds 47 times. The #1 has won 23 and lost 24 of those games. So perhaps not regularly, but we're certainly not seeing #1s clobber #2s if they're so much better (in fact we're not even seeing them win more than they lose).

I got bored and went through all 35 tournament years (1985-2019) to see how often #1 seeds and #2 seeds made the Elite Eight.

Out of 140 Elite Eights:

47 times were a chalk 1 vs. 2 matchup
49 times the 1 seed made it but the 2 seed did not (1 vs. 3/6/7/10/11)
17 times the 2 seed made it but the 1 seed did not (2 vs. 4/5/8/9/12)
27 times neither the 1 seed nor 2 seed made it

Seems like 1 seeds make the Final Four a lot more than 2 seeds not because they beat them head-to-head, but because they get to the Elite Eight games a lot more.

[camion]

I got bored and went through all 35 tournament years (1985-2019) to see how often #1 seeds and #2 seeds made the Elite Eight.

Out of 140 Elite Eights:

47 times were a chalk 1 vs. 2 matchup
49 times the 1 seed made it but the 2 seed did not (1 vs. 3/6/7/10/11)
17 times the 2 seed made it but the 1 seed did not (2 vs. 4/5/8/9/12)
27 times neither the 1 seed nor 2 seed made it

Seems like 1 seeds make the Final Four a lot more than 2 seeds not because they beat them head-to-head, but because they get to the Elite Eight games a lot more.

Not all that surprising I guess. The 1 seeds are slightly stronger and play slightly weaker teams (in general) than 2 seeds in each round. In any one game or round that is isn't much, but over several rounds one would expect 2 seeds to be upset more than 1 seeds.

[npdevil27]

Specific to this year, a 2 in the East seems preferable to a 1 anywhere else, given that the 1 in the East will likely be SDSU. Would anyone disagree with that? Again, this is specific to this year - playing in MSG against a west coast, non major 1 seed.

[Steven43]

Specific to this year, a 2 in the East seems preferable to a 1 anywhere else, given that the 1 in the East will likely be SDSU. Would anyone disagree with that? Again, this is specific to this year - playing in MSG against a west coast, non major 1 seed.

Sign me up for that! Why not?

[MarkD83]

A lot of words there, but in the end I've got you down for, "It's theoretically possible to quantify, but nobody has actually successfully quantified it." Is that about right?

After some deep thinking about this the answer is 42

[Troublemaker]

Since 1985, 1-seeds have played 2-seeds 47 times. The #1 has won 23 and lost 24 of those games. So perhaps not regularly, but we're certainly not seeing #1s clobber #2s if they're so much better (in fact we're not even seeing them win more than they lose).

I got bored and went through all 35 tournament years (1985-2019) to see how often #1 seeds and #2 seeds made the Elite Eight.

Out of 140 Elite Eights:

47 times were a chalk 1 vs. 2 matchup
49 times the 1 seed made it but the 2 seed did not (1 vs. 3/6/7/10/11)
17 times the 2 seed made it but the 1 seed did not (2 vs. 4/5/8/9/12)
27 times neither the 1 seed nor 2 seed made it

Seems like 1 seeds make the Final Four a lot more than 2 seeds not because they beat them head-to-head, but because they get to the Elite Eight games a lot more.

Yeah, if the 1-seed vs 2-seed contests are virtually a tossup (with 2 seeds slightly ahead even), that heavily suggests path is very important and the main difference.

[npdevil27]

Yeah, if the 1-seed vs 2-seed contests are virtually a tossup (with 2 seeds slightly ahead even), that heavily suggests path is very important and the main difference.

Still too much noise in the data. It could just as easily suggest that the 2 seeds that didn’t make the Elite 8 were inferior teams. You’d have to look at each individual team to come to conclusions (using kenpom would be a good start).