Picking the Bracket: How To Beat Your Typical Office Pool

By Matthew Greenberg

It’s official. March is here. As a sports junkie with a penchant for anything that can lead to bragging rights and a little more cash in my pocket, nothing beats March Madness; it is simply the best time of the year. With non-stop college basketball and Cinderella stories ruling the airwaves for the next three weeks, the action is enthralling. However, there is a flipside to this monster: Being the loser. Every year there is that guy, who thinks they know everything there is to know about college basketball, watches all of the coverage and thinks this year is his year; then when the thirteen seed he was so sure was going all the way loses in the first round, his bracket turns to junk. I’ve been there. I know how tough it is. Luckily for all you concerned Jay Bilas wannabes, with a little statistical analysis, I can help you insure that you will never be that guy again, and maybe win you some cash in the process.

In the past, my strategy was simple: Pick the teams you believe are most likely to win, but try as much as you can to go against the consensus. In any given year, it is much more likely that the best team in the tournament will get upset at some point than that they will win six straight games against some very good basketball teams. Therefore, it is in your benefit to pick the team with the highest chance of winning, that no one else is picking. For example, if 50% of people are picking the favorite, who has a 25% chance of winning, it might be beneficial for you to take the team with only a 15% chance of winning but only 5% of people are picking; even though you are picking the less likely winner, if that team does win you will be at a significant advantage over the field. If you were to pick the same champion as everyone else, you would also have to outperform the masses in the rest of the bracket, and that is significantly more difficult. So, I created a metric to estimate your points gained on the field with any given pick, based on the teams odds of reaching a certain round and what percentage of people picked them to get to that round. The formula is simple:

Expected Points Gained = Expected Value of Your Pick – Expected Value of the field

EV of your pick = (% Chance team advances)*(Points for an advance)
EV of the field= (% chance your team advances)*(Points for an advance)*
(% of field selecting your team) + sum for all other teams (Points for other team advancing)(Points for an advance)(% of field selecting that team)

However, this year is different. Instead of just competing against the field, you are also competing against yourself, and perfection. This year, Warren Buffett has created the biggest free office pool in history, with a $1 Billion prize on the line for anyone that can pick a perfect bracket. Therefore, this year, I will be breaking down the bracket from two statistical points of view:

1) To win your typical office pool
2) To win the perfect bracket challenge.

Winning Your Typical Office Pool

At this point in my life, I have come to terms with the fact that the so-called “experts,” while seemingly always wrong, still know exceptionally more than I do. So, instead of trying to predict the odds of a team advancing to any given round, I let them do the work for me. This year, Five Thirty Eight Sports’ Matthew Conlen and Nate Silver created a model that gives you the exact odds of any team reaching any round (it can be found at this link http://fivethirtyeight.com/interactives/march-madness-predictions/). Using their odds, along with ESPN’s measurement of percentage of people choosing a given team to reach any given round, I was able to calculate the teams with the highest EPG for every matchup, and choose the bracket with the statistically highest probability of winning your office pool. However, be aware, this is solely based off of ESPN’s percentage picked by the public, and the odds will change if the ESPN sample is significantly different from your office pool’s makeup. For example, if you are aware that 100% of the people in your office are Florida fans, it’s probably a bad idea to pick Louisville to win, even if that’s what my projections are saying is the optimal choice.
So, without further ado, here’s what the model says:

First Round Upsets:

1. South
a. Pitt def Colorado
2. East
a. None
3. West
a. Oklahoma St def Gonzaga
4. Midwest
a. Iowa/Tenn def UMass
b. Arizona St def Texas

My Analysis

This is actually exactly what I had in my bracket without the model. Since Colorado lost Spencer Dinwiddle, they simply have not been the same team, going 9-9; I fully expect that Pitt will take advantage, and move on.
Oklahoma State is by far the scariest 9 seed in this tournament, having their seed significantly hindered as a result of the losses they had during Marcus Smart’s suspension. Since his return, Oklahoma State has looked very good, and they could be a threat to Arizona in the next round should they get passed Gonzaga, as I expect they should.
The Iowa/Tennessee matchup is one that features two teams that could, and should, knock off UMass, despite UMass star point guard Chaz Williams. Ultimately, these two teams have two much talent, and should be able to pull off the 11 over 6 “upset” (still an upset despite that the odds makers favoring the 11 seeds).
Arizona St Vs Texas is actually a very interesting game. Texas was really solid early in the season, but since have been one of the coldest teams in college basketball, going 1-7 in their most recent 8 non-home games. On the other side, Arizona State has a stacked backcourt, but little depth, and they have been inconsistent all season long. Ultimately, this is a toss-up, and that’s why I’m picking the underdog.

Round of 32 Upsets
None. The model predicts chalk all the way up and down the bracket here. The top seeds this year are significantly stronger than the rest of the field, and barring a 5 over a 4, or possibly Kentucky over Wichita State, it seems unlikely that we will see any serious upsets.

The Model’s Elite 8

  1. Florida
  2. Kansas
  3. Michigan State
  4. Villanova
  5. Arizona
  6. Wisconsin
  7. Louisville
  8. Duke

The upsets here are Louisville over Wichita State, Michigan State over Virginia and Duke over Michigan. This is a product of unfortunate seeding, and a seriously bad draw for Wichita State. Louisville is the hottest team in the country, and playing in any other major division, they would have been a one seed. Michigan State is finally healthy again, and extremely dangerous; they could be the best team in the country when fully healthy. The Michigan-Duke game is scary, and as a Michigan fan I am inclined to believe Nik Stauskas will carry the Wolverines, but the model seems to think Jabari Parker will be too much for my beloved Blue.

The Final Four

  1. Florida
  2. Michigan State
  3. Arizona
  4. Louisville

Two four seeds in the Final Four? Again, this is a result of poor seeding, but the model sees right through it. Michigan State and Louisville are both extremely dangerous teams, and Florida is the best team in the country. While Arizona is not my favorite, they got a relatively easy draw, and even without Brandon Ashley, should be able to knock off Wisconsin

The Finals and the Champion: Florida beats Arizona

Oddly enough, the model says that the best pick to win the entire tournament is actually Louisville, not Florida. However, because so few people are picking Arizona to advance to the Final, the model believes your best chance to win your bracket is achieved by picking Arizona over Louisville in the Final Four. As a result, the model predicts that Florida will win the whole thing.

Winning The Perfect Bracket Challenge

Unlike winning your typical office pool, the optimal choice when attempting to pick a perfect bracket is not to worry about what other people are doing, but rather only to pick the most likely outcome. In this scenario, Louisville takes the crown, but everything else remains the same. Just remember, even playing the odds perfectly, your chances of winning are still something like 1 in 500 million, so keep your expectations tampered.


March Madness is a lot of fun, even when you have nothing on the line, but it is exponentially more fun when you are playing taking part in a pool and winning. Following this model will give you the highest probability of doing just that.
Good luck out there! Maybe one of you will be $1 Billion richer at the end of this whole thing.

Check out the Perfect Bracket Here: ESPN - Tournament Challenge - Model Prediction

Leave a Reply