The Dark Statistician Guide to the Office March Madness Pool
contrarian vampirism for expected value $20
Today I will use my undergraduate statistical knowledge for evil. Because I’m writing with the heart of an undergraduate I will skip the hard math and do only what I need to get credit. This is a guide to the minimum knowledge needed to succeed in an office March Madness pool.
I morally object to sports gambling not because gambling itself is evil. I deeply honor the human urge to tempt fate with a roll of dice, and believe Dmitri Karamazov is just as important as Alyosha and Ivan. But gambling today is ruled by vampiric bureaucracies who tempt warm flesh into their perfectly-calculated lairs with shining lights on screens.
The office March Madness pool is a safe place where gambling remains pure. Coworkers bet totally uninformed but stake real money, to talk about how their bracket got busted as they pass each other on the way to the lunchroom. They base bets on eternal American allegiances to colleges, memories of past March Madness performance, and random psychological factors like bracket shape or school name coolness. It’s very sweet. For them, losing the money is worth the conversations. But someone gets to take the money.
When uninformed bettors systemically bet against reality, you can employ vampiric casino math to return to reality and bet better than them. I make no promise of results in any given year, but you can very easily be better than average.
Ethically, this is minor vampirism of a childhood friend at a sleepover. Yes you're sucking their blood but you guys had a good time together and when they wake up they probably won't even notice the blood is gone. Other people are doing this too. Optimization culture always finds every corner. It’s up to you whether you want to use this guide, but I always say a small prayer for those other humans I siphon money from. I will eat the Chipotle bowl I expect to earn with this dark application of magic with great gratitude.
Your bracket is an expected value problem, in which you want to maximize your odds of scoring more points than other brackets in your pool. You don’t need any college basketball knowledge. In fact, college basketball knowledge is actively unhelpful because it leads you to think you know more than you do. When you’re making your bracket think of it as sixty-three roulette wheels spinning that will soon reveal their secrets. Each roulette wheel is one of the sixty-three games that you are picking. In the first round the roulette wheels have two colors, in the second they have four, in the third they have eight, and so on. You want to be positioned to make the most of any bounce they take.
The roulette wheel’s odds are public information. I use FiveThirtyEight, which is not perfect, but the user interface is really good and I can very easily pull numbers.
Using 538’s odds, you want to pick not the strictly most likely events, but events that are likely and are being underestimated. Picking the most likely events will give you a consistently above average bracket, but someone else will predict more of the randomness than you. Bracket pools are typically winner take-all or payout for the top 3, so being consistently above average does not matter.
The goal is to be contrarian and pick events that other people are not picking. Your expected value total points will be lower than strictly picking favorites, but your odds of winning will be higher because you’re going against the crowd into underpicked slots of the roulette wheel. For census data on what uninformed bettors are picking, I use Yahoo’s Pick Distribution table.
Examples of strong underpicked events in the first round:
Utah State over Missouri (64% odds by 538, 35% picked by Yahoo)
Boise State over Northwestern (59% odds by 538, 38% picked by Yahoo)
Florida Atlantic over Memphis (39% odds by 538, 26% picked by Yahoo)
Furman over Virginia (28% odds by 538, 12% picked by Yahoo)
Creighton over NC State (78% odds by 538, 64% picked by Yahoo)
Mississippi State/Pitt over Iowa State (42% odds by 538, 20% picked by Yahoo)
Strong underpicked events for second round:
Creighton to Sweet 16 (46% odds by 538, 25% picked by Yahoo)
Tennessee to Sweet 16 (46% odds by 538, 34% picked by Yahoo)
San Diego State to Sweet 16 (39% odds by 538, 25% picked by Yahoo)
Kentucky to Sweet 16 (45% odds by 538, 35% picked by Yahoo)
Strong underpicked events for third round:
Connecticut to Elite 8 (32% odds by 538, 15% picked by Yahoo)
Creighton to Elite 8 (26% odds by 538, 8% picked by Yahoo)
These are meant to be illustrative examples, not a list of picks. I recommend pulling up the two spreadsheets side-by-side yourself looking through the numbers with this attitude.
I only bet on events that have a greater than 10% chance of happening. This goes back to the thesis that you want to maximize your chances of winning the pool, not maximize your total number of expected points scored. I adjust to 5% for the championship because winning is pretty unlikely for any given team. I also check casino odds here because casinos are usually a bit more precise than 538..
Houston (25% odds to win by 538 | 15% odds by casinos | picked by 11% of Yahoo)
Alabama (15% odds to win by 538 | 12% odds by casinos | picked by 20% of Yahoo)
Texas (8% odds to win by 538 | 6% by casinos | picked by 6% of Yahoo)
Purdue (5% odds to win by 538 | 8% by casinos | picked by 8% of Yahoo)
Kansas (5% odds to win by 538 | 11% by casinos | picked by 13% of Yahoo)
I wouldn't pick any other teams than these, besides maybe UCLA who 538 has at 3% but casinos have at 8%. Houston is a really tempting pick, heavily favored by casinos and even more by 538 but not as heavily picked by the crowd.
Other events can happen though. These teams only represent half of the final roulette wheel, but the other half is spread over a ton of teams, which makes it really hard for you or anyone to pick the correct winner out of the rest of the field. Riding a safe favorite for champion is defensive, because it prevents a scenario in which lots of people in your pool to score a ton of points on you when a favorite wins.
This strategy will give you a single bracket with a better chance of winning than average, but the chances of that individual bracket winning is not a guarantee because it’s hard to place at the top of a large field with scoring so dependent on picking the correct national champion. To increase your chances of winning money in any given year, run multiple brackets and decentralize your picks. Pick different champions and different Final Four in each bracket. The scenarios in which one favorite does not win very often contain the second or third favorites winning. If you’re running a solid contrarian strategy for the early rounds the wider variety you pick in the later rounds means you dramatically increase the odds of one of your well-structured brackets panning out. Even though Alabama is overpicked, picked by 20% of the fields with only a 12% chance to win, 12% chance to win is pretty good and I’ll probably run an Alabama bracket, because I want to have a chance in that 12% scenario.
This is really all that I do. Bet specifically against the uninformed field and decentralize across my different brackets. I run these two strategies in combination, picking different but likely and underpicked upsets across my brackets. Keep in mind scoring and pool size when you structure your picks. Smaller pools call for less aggressive picks, bigger pools call for more aggressive picks. Some pools use alternate scoring that gives less points for the champion, which means be more aggressive in your champion pick but more cautious in your early picks. You can combine the differences in pool structure with the fact that you already want to decentralize your picks, and run different mixes well-suited to each pool that maximize your personal odds for the year to win money.
Advanced optional strategy is to try to account for your geographic area’s unique skew. Yahoo data is nationwide, but different areas of the country will typically pick their own teams more. Betting against the popular colleges of your area typically gives you an advantage because other people in your pool will be heavily picking them.
Note that these strategies are optimized for pools between 10 and 100 participants. If you have under 10 participants, just pick 538 favorites, no need to get fancy about it. If you get over 100, playing the slot machines on an underpicked national champion with 3% odds to win is a good strategy.
Though I’ve written a guide to taking the soul out of your bracket, I encourage you to still feel the soul if you run these strategies. Check the score of the games during work, tease your coworkers who made worse picks. Because I’m not running vampirism on the scale of casinos, what I find is that I’ll need one or two games at the end of the tournament to go my way. My vampirism allows me to end up in that situation almost every year, while normal people may get there just once every few years. Needing one game to win the office pool I feel the human soul of gambling, the tense drama as the roulette wheel slows, the feeling Uncut Gems is all about. The moment where the roulette wheel has yet to decide its final resting place is special. You should enjoy your free spin.
This post is financial advice.