Tournament by the numbers

[tmh8286]

Pomeroy has got a nice look at the Tournament from a statistical point of view.  Here's how he figures our bracket:

Here?s the East Region:

                2nd Rd Sweet16  Elite8  Final4  Finals  Champs

1  N. Carolina  99.26%  81.43%  67.94%  48.69%  32.61%  20.90%

2  Georgetown  96.59%  81.33%  67.30%  32.43%  17.66%  9.09%

4  Texas        88.43%  55.85%  14.54%    6.11%    2.14%  0.69%

9  Mich. St.    60.71%  12.72%    7.24%    3.09%    1.10%  0.36%

3  Wash. St.    81.51%  52.09%  14.57%    3.27%    0.83%  0.19%

12  Arkansas    55.18%  24.33%    4.61%    1.52%    0.40%  0.09%

7  Boston Coll. 64.90%  13.48%    7.16%    1.44%    0.33%  0.07%

8  Marquette    39.29%  5.82%    2.70%    0.91%    0.24%  0.06%

6  Vanderbilt  71.57%  34.34%    7.68%    1.35%    0.27%  0.05%

5  USC          44.82%  17.69%    2.86%    0.83%    0.19%  0.04%

10  Texas Tech  35.10%  4.62%    1.79%    0.22%    0.03%  0.00%

13  New Mex. St. 11.57%  2.13%    0.11%    0.01%    0.00%  0.00%

11  GW          28.43%  8.04%    0.90%    0.08%    0.01%  0.00%

14  Oral Roberts 18.49%  5.52%    0.52%    0.04%    0.00%  0.00%

15  Belmont      3.41%  0.57%    0.09%    0.00%    0.00%  0.00%

16  E. Kentucky  0.74%  0.03%    0.00%    0.00%    0.00%  0.00%

He's giving us about a one in five chance of winning today - I'll take that!

[ORUTerry]

Not a lot of love for PAC-10 members Washington State (#3) or USC (#5). How about those Hogs?

[tmh8286]

Other 14 seeds:

14  Pennsylvania  4.93%  0.72%    0.08%    0.01%    0.00%  0.00%

14  Miami (OH)  13.98%  2.31%    0.21%    0.02%    0.00%  0.00%

14  Wright St.  10.29%  1.22%    0.12%    0.01%    0.00%  0.00%

Statistically, we have the best chance of winning the first round of any 14 seed.

[tmh8286]

Lots of good tournament info on Ken's blog:

Pomeroy Blog

Did I ever say how much I love this guy's website???

[Bogus Smith]

Makes me feel like Western Illinois...no chance whatsoever to win the Tournament!  

Kinda ironic that 17% of #14-seeds have won and Pomeroy has us at 18%.  I guess he's playing the odds.

[tmh8286]

Not quite that simple, Bogus (as I know you know ).  Ken's using log5 analysis to generate those numbers.   Here's how it's done:

Log5 method

From Wiki Gonzalez

This is the name Bill James gave for a simple(?) derivation of what is the probability that a given team A with AWin% would beat another team B with BWin%.

ProbABeatsB = (AWin% * (1 - BWin%)) / ((AWin% * (1 - BWin%)) + ((1 - AWin%) * BWin%))

            = (AWin% - (BWin% * AWin%)) / (AWin% + BWin% - (2 * AWin% * BWin%))

The simple reasoning here is you can imagine AWin and BWin as two independent random variables that get a "win" with AWin% and BWin% respectively and a "loss" with 1 - AWin% and 1 - BWin% respectively. A game in essence is comparing randomly these two random variables and saying A wins only if A picked "win" and B picked "loss" but saying A losses only when A picked "loss" and B picked "win". If A and B picked the same run again.

Retrieved from "http://digamma.net/btfwiki/Log5_method"

[Bogus Smith]

Not quite that simple, Bogus (as I know you know ).  Ken's using log5 analysis to generate those numbers.   Here's how it's done:

Log5 method

From Wiki Gonzalez

This is the name Bill James gave for a simple(?) derivation of what is the probability that a given team A with AWin% would beat another team B with BWin%.

ProbABeatsB = (AWin% * (1 - BWin%)) / ((AWin% * (1 - BWin%)) + ((1 - AWin%) * BWin%))

            = (AWin% - (BWin% * AWin%)) / (AWin% + BWin% - (2 * AWin% * BWin%))

The simple reasoning here is you can imagine AWin and BWin as two independent random variables that get a "win" with AWin% and BWin% respectively and a "loss" with 1 - AWin% and 1 - BWin% respectively. A game in essence is comparing randomly these two random variables and saying A wins only if A picked "win" and B picked "loss" but saying A losses only when A picked "loss" and B picked "win". If A and B picked the same run again.

Retrieved from "http://digamma.net/btfwiki/Log5_method"

This page has been accessed 749 times. This page was last modified 11:16, 25 Nov 2004. Content is available under GNU Free Documentation License 1.2.

TMI 

[Texasgrip]

Not quite that simple, Bogus (as I know you know ).  Ken's using log5 analysis to generate those numbers.   Here's how it's done:

Log5 method

From Wiki Gonzalez

This is the name Bill James gave for a simple(?) derivation of what is the probability that a given team A with AWin% would beat another team B with BWin%.

ProbABeatsB = (AWin% * (1 - BWin%)) / ((AWin% * (1 - BWin%)) + ((1 - AWin%) * BWin%))

            = (AWin% - (BWin% * AWin%)) / (AWin% + BWin% - (2 * AWin% * BWin%))

The simple reasoning here is you can imagine AWin and BWin as two independent random variables that get a "win" with AWin% and BWin% respectively and a "loss" with 1 - AWin% and 1 - BWin% respectively. A game in essence is comparing randomly these two random variables and saying A wins only if A picked "win" and B picked "loss" but saying A losses only when A picked "loss" and B picked "win". If A and B picked the same run again.

Retrieved from "http://digamma.net/btfwiki/Log5_method"

This page has been accessed 749 times. This page was last modified 11:16, 25 Nov 2004. Content is available under GNU Free Documentation License 1.2.

Don't ever post that again...

[ORUTerry]

Okay... now I have a headache.

[ManiacVP]

I'm in math and society for a reason... and we're having a pi day party today thats right... in honor of yesterday 3/14... we are having a pi day (3.14) party. Dont throw numbers at me when im looking at a good day.