Blog

Geek Challenge: March Madness

Geek Challenge: March Madness

This month’s Geek Challenge was created in conjunction with Kevin Ferrigno. The challenge is to pick the winners of this year’s NCAA tournament, not by choosing teams, but by picking the winning formula. You can use this spreadsheet to practice with based on last year’s games.

We’ve simplified the process of picking a bracket from selecting the outcome of 63 games down to selecting five numbers.

We will create 10 randomly generated brackets based on your chosen numbers, and the winner will be the person whose brackets do the best based on this formula:

Total Score = Highest Bracket Score + Average Bracket Score

Our formula has 2 parts:

  1. Accurately generate a rating for each team
  2. Determine how much randomness to use when making picks for each game

The factors to use in generating a rating are:

  • A. Won-Loss Percentage: A team's Wins divided by Games Played.
  • B. Strength of Schedule: W-L record of a team’s opponents, and often the opponents of their opponents (a 20-10 team from the Big Ten or Big East Conferences is not likely equivalent to a 20-10 team from the Horizon Conference). We’ll use the readily available RPI version for our calculations.
  • C. Tournament History: W-L percentage of a team in the NCAA tournament over the last 3 years. Some teams (recently Butler University) just have a knack for doing well in the tournament. This factor may allow you to give an extra nudge to teams with a history of tournament success.
  • D. Tournament Seed: The historical W-L percentage of teams with a particular tournament seed. #1 seeds win 78% of the time over all rounds. #2 seeds win 70% of the time. A #16 has never won a game in the field of 64. (Historical performance of teams by Tournament Seed)

All factors have been pre-scaled based to the mean and standard deviation of the group of 64 teams. So the challenge in this part is to determine the relative weight of the 4 statistics and express these in the coefficients W, X, Y and Z. For instance, if your strategy is that Win-Loss Percentage and Strength of Schedule matter equally, Tournament History doesn’t matter at all, and that underdogs do slightly better, you might pick the coefficients X=1, Y=1, Z=0, and W=-0.2.

Your second challenge is to pick a coefficient representing the randomness. A randomness coefficient K=0 means no randomization, and all games are picked based on weighting. Higher coefficients offer more randomness. Since each participant will get 10 brackets calculated, and the objective is to maximize high score plus average score, it makes good sense to have some randomization.

We've developed a practice spreadsheet using last year’s tournament. To participate in this month’s Geek Challenge, please submit your 5 coefficients before the first game is played in the round of 64.

Please submit your responses to: geekchallenge@dmcinfo.com

Comments

Anonymous User
# Anonymous User
Geek Challenge Results: March Madness Part I
Kevin Ferrigno
# Kevin Ferrigno
The preliminary 2012 spreadsheet is ready. While you can't test your numbers against actual bracket, you can see how your chosen set of coefficients will rate the teams given their 2011-2012 season performance. You can also create sample brackets for this year's tournament to see if you like the results. Email geekchallenge@dmcinfo.com and we can send you a copy. The final version will be subject to the results the "First Four" games. Entrants will be provided with a spreadsheet containing their official 10 entries on Thursday.
Ken Brey
# Ken Brey
Please submit your numbers to geekchallenge@dmcinfo.com.

The brackets calcualted and published when the 64 teams are known (after the play-in games).

Post a comment

Name (required)

Email (required)

CAPTCHA image
Enter the code shown above: