PDA

View Full Version : Odds of winning NCAA tourney



duketaylor
03-15-2010, 11:07 AM
Interesting stuff, IMO.
http://waynewinston.com/wordpress/
Also shows the odds of advancing to each round.

DevilWolf
03-15-2010, 11:23 AM
Duke has a 12% chance to win 6 games but an 11% chance to win 5 games?

riverside6
03-15-2010, 11:30 AM
I'll see your Sagarin simulation of 5000 times and raise you to a Pomeroy simulation done 1 million times (pinky finger to the lips).

http://scacchoops.com/tt_NewsBreaker_External.asp?NB=3426

-bdbd
03-15-2010, 11:43 AM
It may be more intuitive if you look at the lower of the two charts.
At this point in time Sagarin would say about our expectations:
Duke has 99.1% odds of winning the first game. 76.7% odds of winning 2 games. 55.2% of winning three. 37.5% odds of winning 4 games. 24% odds of winning 5 games. And 12.1% odds of winning 6 games. IOW, Sagarin would expect us to make it to, but then lose in, the Elite 8 round.

I thought it interesting that Sagarin gives Cal about 54% odds of beating Louisville (hope that pans out).

Lastly, note that NOBODY has a "probability" of winning it all. The best team has less than a 32% expectation of winning it all and other than KA, Duke is 2nd highest of that expectation at barely 12%. So if anyone wants Kansas, and gives you "the field".... take it! (at 2:1 odds!) :)

Caveat: This is all dependent on the Sagarin rating system, and I don't think it factors in WHERE they are playing (such as Baylor playing in Hou), who is injured, who is playing hot, etc.

I think these stats are most useful in a relative sense more than anything -- comparing a couple of teams.

Team 1 2 3 4 5 6
Kansas 0.996 0.879 0.738 0.600 0.446 0.319
Duke 0.991 0.767 0.552 0.375 0.240 0.121
Syracuse 0.967 0.728 0.552 0.358 0.178 0.108
Kentucky 0.969 0.671 0.463 0.291 0.161 0.074
West Va. 0.956 0.653 0.461 0.259 0.146 0.067
Kan St 0.956 0.585 0.387 0.211 0.081 0.043
Villanova 0.971 0.662 0.402 0.191 0.097 0.037
Baylor 0.890 0.603 0.317 0.147 0.068 0.026
Georgetown 0.898 0.552 0.314 0.103 0.051 0.020
Ohio St 0.914 0.594 0.325 0.096 0.043 0.018
Purdue 0.733 0.436 0.176 0.092 0.042 0.015
Wisconsin 0.820 0.470 0.194 0.094 0.042 0.015
BYU 0.707 0.329 0.204 0.102 0.035 0.015
Temple 0.685 0.374 0.147 0.069 0.028 0.010
Maryland 0.858 0.523 0.126 0.067 0.026 0.010
Texas 0.661 0.251 0.134 0.068 0.030 0.009
Pitt 0.835 0.448 0.180 0.073 0.021 0.009
Texas A & M 0.611 0.312 0.118 0.054 0.021 0.008
Butler 0.557 0.318 0.115 0.050 0.017 0.008
New Mexico 0.826 0.440 0.171 0.064 0.024 0.007
Xavier 0.627 0.352 0.143 0.057 0.016 0.006
Marquette 0.579 0.318 0.123 0.047 0.019 0.005
Mich St 0.813 0.401 0.080 0.033 0.013 0.005
Tennesee 0.594 0.277 0.137 0.035 0.013 0.005
Vandy 0.667 0.342 0.118 0.048 0.013 0.005
Missouri 0.519 0.187 0.094 0.034 0.014 0.004
Georgia Tech 0.514 0.206 0.083 0.018 0.007 0.003
Clemson 0.481 0.157 0.084 0.033 0.013 0.003
Fla. St. 0.551 0.158 0.077 0.030 0.007 0.002
Saint Mary’s 0.531 0.183 0.075 0.023 0.007 0.002
Notre Dame 0.512 0.192 0.071 0.019 0.007 0.002
San Diego St 0.406 0.156 0.063 0.012 0.004 0.002
Cal 0.541 0.135 0.055 0.024 0.007 0.002
Okla St. 0.486 0.185 0.073 0.015 0.004 0.001
UTEP 0.443 0.230 0.065 0.024 0.007 0.001
Utah St 0.389 0.151 0.039 0.015 0.004 0.001
Richmond 0.469 0.154 0.060 0.019 0.005 0.001
UNLV 0.469 0.051 0.021 0.007 0.003 0.001
N Iowa 0.531 0.069 0.031 0.012 0.005 0.001
Gonzaga 0.449 0.110 0.051 0.017 0.004 0.001
Minn 0.373 0.166 0.047 0.014 0.003 0.001
Florida 0.293 0.082 0.034 0.010 0.003 0.001
Washington 0.421 0.205 0.063 0.021 0.006 0.001
Lousiville 0.459 0.097 0.039 0.014 0.005 0.001

InSpades
03-15-2010, 11:43 AM
Duke has a 12% chance to win 6 games but an 11% chance to win 5 games?

As the 2nd best team according to Sagarin... when we make the final we are more likely than not to win the final. Kansas is *way* more likely to win 6 games than to win 5 games.

Note: that we're talking about winning exactly the number, not that number or more.

blueprofessor
03-15-2010, 11:51 AM
Evaluation of every team from 1 round to the next.
http://vegaswatch.net/

Best regards--Blue Prof:)

superdave
03-15-2010, 01:31 PM
I'll see your Sagarin simulation of 5000 times and raise you to a Pomeroy simulation done 1 million times (pinky finger to the lips).

http://scacchoops.com/tt_NewsBreaker_External.asp?NB=3426

I just looked through last year's result and it had UNC at #5 in odds and they won all their games by 10+ points I believe. It had Duke at #6 in odds and we got stomped! (Villanova was 18th in likelhood to win 6 games).

J_C_Steel
03-15-2010, 01:33 PM
How does this work? Does it take into account how certain teams match up against others?

I've never been a huge fan of statistical prediction models in basketball.

riverside6
03-15-2010, 02:11 PM
I just looked through last year's result and it had UNC at #5 in odds and they won all their games by 10+ points I believe. It had Duke at #6 in odds and we got stomped! (Villanova was 18th in likelhood to win 6 games).

Show me the model that was exactly right, and I'll show you one that filled out by someone and backdated!

That said, you're exactly right, what this fails to do is take into account things like Duke playing significantly better at home, injuries like the one to Hummel, etc. Its not perfect, but it does present a way of showing teams that might be seeded far too low or far too high.

superdave
03-15-2010, 02:48 PM
Show me the model that was exactly right, and I'll show you one that filled out by someone and backdated!

That said, you're exactly right, what this fails to do is take into account things like Duke playing significantly better at home, injuries like the one to Hummel, etc. Its not perfect, but it does present a way of showing teams that might be seeded far too low or far too high.

It is definitely useful to see how a model can rate teams. I'm just suprised Unc wasnt the darling of the models last year.

The model this year seems to show Wisconsin has an upset or two up their sleeve, which was sort of what I was thinking. It seems to me that a team like Wisco could throw enough odd looks on both ends to upset a better team. But if you play them year after year in conference, those looks arent quite so odd anymore, hence their good but not great results.

InSpades
03-15-2010, 03:14 PM
Interesting stuff, IMO.
http://waynewinston.com/wordpress/
Also shows the odds of advancing to each round.

There's an office pool where I work that has you pick 10 teams, you get a number of points equal to that teams seed for each win they get. So if you pick... Maryland and they lose in the sweet 16 you would get 8 points (4 for each of their wins). Instinctively my thoughts went to picking teams in the 6 seed range (although 10 was also reasonable I guess). I decided to use the above #s to see which team is predicted to garner the most points and the results were a bit curious.

In order:
7 BYU
8 Texas
12 UTEP
11 Old Dominion
10 Missouri
10 Georgia Tech
10 Saint Mary’s
11 Washington
9 Fla. St.
6 Xavier
12 Utah St
11 San Diego St
5 Mich St
11 Minn
5 Temple
4 Wisconsin
6 Marquette
4 Maryland
6 Tennesee

BYU I could see picking but most of the rest I wouldn't have imagined that high. This obviously doesn't factor in that Purdue is a lot weaker than their results indicate. I'd expect Siena and Texas A&M to be 2 very highly picked teams. Obviously the 1 seeds are all awful picks (if they win it all it's as good as a 6 seed winning their opening round game).

Note: I'm assuming that the play-in game doesn't count as you'd almost definitely pick 1 of them (or both of them).

CDu
03-15-2010, 04:34 PM
Duke has a 12% chance to win 6 games but an 11% chance to win 5 games?

It's not a cumulative calculation (i.e., not the probability of winning at least 5 games, but rather exactly 5 games). The cumulative probability of winning 5 games would be 23%. So this calculation is saying that if we make the final game, we're more likely to win it than we are to lose it. And it seems to suggest we're twice as likely to win our Final Four game than to lose (if we make it that far).

Of course, it also has our most likely outcomes as losing in the second or third rounds (and we're definitely expected to lose by the second weekend).

But, it gives us the second-best overall odds, which is nice.

What I think is most useful about these projections is to illustrate just how unlikely it is to (a) win it all and (b) make it to the final four. Kansas, the #1 overall seed, is only considered 1 in 3 chance to win it all.

hurleyfor3
03-15-2010, 04:53 PM
There's an office pool where I work that has you pick 10 teams, you get a number of points equal to that teams seed for each win they get. So if you pick... Maryland and they lose in the sweet 16 you would get 8 points (4 for each of their wins). Instinctively my thoughts went to picking teams in the 6 seed range (although 10 was also reasonable I guess).

Hmmm... If you pick all the 7-seeds, all the 10-seeds and both the teams in one of the 8/9 games you're guaranteed 36 points. And usually one of the 7/10 seeds makes it to the final eight (a #2 almost always goes down in the second round). That's another 14 or 20 points if none of the other picks wins a single game. There are variations on this theme, but the idea is to "waste" picks that are impossible to all be correct.

Neals384
03-17-2010, 12:05 PM
Hmmm... If you pick all the 7-seeds, all the 10-seeds and both the teams in one of the 8/9 games you're guaranteed 36 points. And usually one of the 7/10 seeds makes it to the final eight (a #2 almost always goes down in the second round). That's another 14 or 20 points if none of the other picks wins a single game. There are variations on this theme, but the idea is to "waste" picks that are impossible to all be correct.

Well, then, I'm going to pick all the 8 and 9 seeds, plus one 7/10 matchup, guaranteeing me 39 points!

Jderf
03-17-2010, 12:28 PM
All very interesting. Personally I take the Pomeroy stats over Sagarin, which seems heavily skewed to the top.

riverside6
03-17-2010, 01:13 PM
One more less mathematical view of predictions, we had BC assistant coach Mo Cassara make his picks for the first 2 rounds of the tourney, you can find his predictions here...

West & Midwest Regions (http://scacchoops.com/tt_NewsBreaker_External.asp?NB=3434)
East & South Regions (http://www.scacchoops.com/tt_NewsBreaker_External.asp?NB=3444)

Reddevil
03-17-2010, 01:23 PM
NOBODY has a "probability" of winning it all. The best team has less than a 32% expectation of winning it all and other than KA, Duke is 2nd highest of that expectation at barely 12%. So if anyone wants Kansas, and gives you "the field".... take it! (at 2:1 odds!) :)



Always!.....unless it's the field against the UCONN women.:eek:

CDu
03-17-2010, 04:38 PM
Hmmm... If you pick all the 7-seeds, all the 10-seeds and both the teams in one of the 8/9 games you're guaranteed 36 points. And usually one of the 7/10 seeds makes it to the final eight (a #2 almost always goes down in the second round). That's another 14 or 20 points if none of the other picks wins a single game. There are variations on this theme, but the idea is to "waste" picks that are impossible to all be correct.

That sort of strategy can work in a small pool (because it maximizes your floor). But if the pool is of reasonable size, that strategy is bound to lose (because it minimizes your ceiling). In bigger pools there is less chance that nobody tops that strategy with a less "wasteful" strategy. In other words, in bigger pools, there will be more combinations selected. So there's less chance that the correct combination goes unselected.

InSpades
03-17-2010, 04:59 PM
That sort of strategy can work in a small pool (because it maximizes your floor). But if the pool is of reasonable size, that strategy is bound to lose (because it minimizes your ceiling). In bigger pools there is less chance that nobody tops that strategy with a less "wasteful" strategy. In other words, in bigger pools, there will be more combinations selected. So there's less chance that the correct combination goes unselected.

It's kind of like playing roulette at a casino with your friends. Everyone else picks a number (or set of #s) and you decide to bet on every number. You might do better than they do (meaning lose less, you are guaranteed to lose), but odds are one of them will get lucky and beat you just from simple variance.