The Sports Page

The Pool Desk

A football pool is the best statistics classroom anybody ever built by accident. It has a real prize, a real opponent, and a scoreboard that settles arguments every Sunday. It is also almost universally misunderstood, which is what makes it worth a desk of its own.

We run an entry in a family pool. We publish what we pick before kickoff, and we grade it afterwards whether it worked or not. So far it has mostly not worked, and saying so is the point — a strategy page written by somebody who hides their results is worth nothing.

Nobody here is named. Everyone in the pool is a letter. The interesting thing is the shape of what happens, never who it happened to.

Start here

The Introduction

Football Pool 101

Where your score actually comes from, why being good at picking buys you less than you think, and the one free move that nearly doubles your chance of winning. No math required. Read it →

Going Deeper

The Strategy Notes

The simulations behind the introduction, with the assumptions written out and the limits stated. For anyone who wants to check the arithmetic rather than take our word for it.

What we have learned so far

We cannot pick. Tested against 2,278 college games, the statistics we trust went 50.2% against the spread. Break-even is 52.4%. That is not a disappointing result, it is the normal one, and every strategy we publish assumes it.

The ranking is worth nothing to your score and a great deal to your odds. Over 120,000 simulated weeks, where you put your big numbers never moved the average score by more than a tenth of a point — and moved the chance of winning from 14.5% to 22.4%.

We got beaten by a hat. Our week-three entry scored at the 6th percentile of what randomly shuffling the same picks would have managed. We put our big numbers on losers. A rule we built afterwards to fix it also lost to random, and its reasoning turned out to be false in the same data that produced it.

How we keep ourselves honest

Anybody can write a rationale after a result and sincerely believe they thought it beforehand. That is memory working normally, which is exactly why a promise not to do it is worth nothing.

So our reasoning is sealed before kickoff: written, moved out of the repository, and published only as a fingerprint of its exact bytes. After the games the text goes up and the fingerprint is recomputed. Change a single character and it will not match. You do not have to trust us; you can check.

The Sports Page — one question per issue, answered with numbers.
Back to the paper · see also the Odds Board.