What bracket scoring do you prefer in your NCAA pool?

[mktmaker]

We have a fairly big office pool (70 - 80 people).

Most don't give it much of a thought and want the traditional scoring (1-2-4-8-16-32).

Though I have been lucky and have won the pool more times (2000, 2004, 2008, 2012, 2014) than any others, I find that scoring objectionable. And this year more than others because so many casual fans will pick Kentucky. If UK wins then most of the deciding factors will simply be tie-breakers.

I like progressive scoring, such as 1-2-3-4-6-8 or 2-4-5-8-9-10.

I also like +seed or x-seed.

IMHO the best system by far is a Calcutta (auction) but we've only been able to get that organized once.

I welcome your thoughts. Thank you.

This post was edited on 3/16 10:58 AM by mktmaker

 

[mkasten25]

Originally posted by mktmaker:

We have a fairly big office pool (70 - 80 people).

Most don't give it much of a thought and want the traditional scoring (1-2-4-8-16-32).

Though I have been lucky and have won the pool more times (2000, 2004, 2008, 2012, 2015) than any others, I find that scoring objectionable. And this year more than others because so many casual fans will pick Kentucky. If UK wins then most of the deciding factors will simply be tie-breakers.

I like progressive scoring, such as 1-2-3-4-6-8 or 2-4-5-8-9-10.

I also like +seed or x-seed.

IMHO the best system by far is a Calcutta (auction) but we've only been able to get that organized once.

I welcome your thoughts. Thank you.

You won 2015? Mind letting me know if I should get my hopes up? Have the sports almanac from BTTF?

 

[RockyMountainCat]

I work in Cincinnati and most of my coworkers are UC, OSU, or ND fans. They don't like UK so no one but me will be picking the Cats. I expect to be the sole winner in the pool.

 

[mktmaker]

Originally posted by mkasten25:
You won 2015? Mind letting me know if I should get my hopes up? Have the sports almanac from BTTF?

Sorry...I meant 2014. I rode the Cats...like many of you.

 

[far_away_fan]

TLDR: This is a neat way for academic types to run a pool, but is of no use for a typical office pool (unless you work at Google or something).

It's impractical for a pool of "normal" people who aren't into computing, but my favorite way of doing brackets is as follows:

Have everyone submit a (very large) file that lists their probability that team A will beat team B for every possible pair of teams in the pool. (There are 2278 such pairs so it's not likely people want to generate those probabilities by hand and type out the file!)

Then as the tournament proceeds, there will be 67 games played. You can then score each submission by looking at the 67 probabilities they list for these 67 games (ignoring the other 2278-67=2211 probabilities they predict).

Your actual score is given by the average over the 67 games of the log of the probability you assigned to the winning team for each matchup that was actually played in the event. The winner is the person with the highest score. (All scores will be negative, so it will be the person whose score is closest to zero.)

So for example, in a small 4 team event, suppose you had probabilities like:
P(1 beats 2) = .7
P(1 beats 3) = .8
P(1 beats 4) = .6
P(2 beats 3) = .6
P(2 beats 4) = .4
P(3 beats 4) = .35

Then in the actual tourney, team 2 beats team 1 and team 3 beats team 4 in the first round. Then team 2 beats team 3 in the second round. Your score would be:
1/3*(log .3 + log .35 + log .6) = -0.9215

If instead, the tourney had gone: team 1 beats team 2 and team 4 beats team 3 in the first round with team 1 then beating team 4 in the second round, the score would be:
1/3*(log .7 + log .65 + log .6) = -0.4328

(By way of comparison, if you made every game a 50%/50% toss-up, your score would be -0.6931 no matter what happened in the games. So we see that in the second sample tourney scenario, your score would be better than just guessing 50/50, but in the first tourney scenario, it would be worse. If you were a gambler and assigned a probability of one or zero to every possible matchup, then your score would be zero - the best possible score - if you got every game correct. However, it would be negative infinity if you missed even one game, so it's best not to use 0/1 probabilities.)

If you want to know who can really predict games, this system is the best. But your friends would all have to predict 2278 games by hand or know how to design a computer model to spit out those probabilities.

 

[bluelifer]

Originally posted by far_away_fan:
TLDR:

Correct.