NCAA Seed Performance Distribution

[wrestleknownothing]

WTT is in the rear view mirror and Final X is weeks away, so I dusted off some NCAA data to keep myself entertained in the between times.

The Seed (Round) rows group the seeds into the rounds they would finish if everything went chalk (or at least if their individual results went chalk).

The Round columns are where they actually finished. So the top 8 seeds finish as AA's 74% of the time, finish in the blood round 16% of the time, etc.

The data is from the 33 seed era only (2019 - 2025). The averages sum horizontally, not vertically.

• For the gamblers in the audience it looks like the 9-12 seeds are the value play as they finish on the podium more often than the blood round. The opposite should be expected.
  
  
• On the flip side, avoid they 13-16 seeds who finish one round short of expectations more often than on seed-based expectations.

 

[psumacw]

• For the gamblers in the audience it looks like the 9-12 seeds are the value play as they finish on the podium more often than the blood round. The opposite should be expected.

great stuff as always but i'm not sure i totally agree with this statement due to the size of the groups (8 places for AA but only 4 for R12). not exactly sure how to resolve that, maybe split AA into 1-4 and 5-8, and do the same with seeds 1-8? you're still left with 8 places for R16, R24, and R32, but those results seem pretty intuitive so maybe that's ok.

 

[Pennstate1985]

great stuff as always but i'm not sure i totally agree with this statement due to the size of the groups (8 places for AA but only 4 for R12). not exactly sure how to resolve that, maybe split AA into 1-4 and 5-8, and do the same with seeds 1-8? you're still left with 8 places for R16, R24, and R32, but those results seem pretty intuitive so maybe that's ok.

I had the same thought. Hard to compare an 8 person bracket to two 4 person brackets. I too would be interested in seeing 1-4 and 5-8.

 

[wrestleknownothing]

great stuff as always but i'm not sure i totally agree with this statement due to the size of the groups (8 places for AA but only 4 for R12). not exactly sure how to resolve that, maybe split AA into 1-4 and 5-8, and do the same with seeds 1-8? you're still left with 8 places for R16, R24, and R32, but those results seem pretty intuitive so maybe that's ok.

I had the same thought. Hard to compare an 8 person bracket to two 4 person brackets. I too would be interested in seeing 1-4 and 5-8.

You made me think about this for a bit, but I think it is OK to treat it the way I did because these are categorical rather than discrete units and that matches the format of the tournament. At first 9 are eliminated as a group with no separation among them, then eight are eliminated, then four, and four again.

Where I get onto thin ice is when I attempt to assign numerical values to the degree of under/over performance. If someone loses in the blood round should I assign a finish of 9? Or 12? Or 10.5 (the mid-point)? Every answer is wrong. So, I just choose one (usually the mid-point).

That said, I will play around with your ideas. Maybe the better way to do it is to aggregate round of 12 with round of 16 to make a group of eight out of the two groups of four, but even that is problematic. The wrestlers who make round of 12 had to win more than those who made round of 16, so you unfairly give the round of 16 wrestlers too much credit, or the round of 12 not enough.

 

[Pennstate1985]

You made me think about this for a bit, but I think it is OK to treat it the way I did because these are categorical rather than discrete units and that matches the format of the tournament. At first 9 are eliminated as a group with no separation among them, then eight are eliminated, then four, and four again.

Where I get onto thin ice is when I attempt to assign numerical values to the degree of under/over performance. If someone loses in the blood round should I assign a finish of 9? Or 12? Or 10.5 (the mid-point)? Every answer is wrong. So, I just choose one (usually the mid-point).

That said, I will play around with your ideas. Maybe the better way to do it is to aggregate round of 12 with round of 16 to make a group of eight out of the two groups of four, but even that is problematic.

Love your work so please don't take anything as a criticism.

If you are mimicking the data to how the tournament is run, wouldn't 9-12 be assigned a "9"? In reality, aren't all the losers in R12 tied for 9th? That said, I don't have an issue with 10.5 either.

 

[wrestleknownothing]

Love your work so please don't take anything as a criticism.

If you are mimicking the data to how the tournament is run, wouldn't 9-12 be assigned a "9"? In reality, aren't all the losers in R12 tied for 9th? That said, I don't have an issue with 10.5 either.

I take it as conversation. The good kind of criticism.

I should be clearer about the mid-point. If a wrestler seeded 8th finishes in the round of 12 I can compare them to either the top of the range, the bottom of the range or the mid-point. But if a wrestler seeded 9th through 12th finishes in the round of 12 they get a 0 relative performance.

Now that I type that it gives me a new thought. Anyone seeded 13-16 who finishes in the round of 12 should default to +4 (and vice versa) rather than seed minus a fixed number. And the same for other categorical moves. But that still doesn't solve the problem of anyone moving from the discrete category (1-8) to the grouped categories (round of 12, etc).

 

[Psalm 1 guy]

All of you statistical types are showing the argumentative types the proper way to work through differences of opinion. Kudos!

Much better than the usual path some of these threads take

 

[Pennstate1985]

All of you statistical types are showing the argumentative types the proper way to work through differences of opinion. Kudos!

Much better than the usual path some of these threads take

Maybe but Napoleon Dynomite was a great movie.

 

[Psalm 1 guy]

Maybe but Napoleon Dynomite was a great movie.

or did you mean this Dynomite!