Football Pool 101

This page is the introduction. You do not need any math to read it, and you will finish it knowing more about how these things work than most people who have played one for twenty years.

1. The rules, in one paragraph

You get a slate of games — ours is fifteen. Each game has a point spread, so you are not picking who wins, you are picking who beats the spread. Then you rank your fifteen picks. Your most confident carries 15 points, your least confident carries 1. Get a pick right, collect its points. Highest total takes the week.

That ranking is the part people agonize over. Hold that thought.

2. Where your score actually comes from

Here is the number that reorganizes everything. Suppose you get 10 of 15 right. Those ten picks are carrying ten of the numbers between 1 and 15. On average, a number pulled from that pile is 8. So your expected score is simply:

how many you got right  ×  8

Ten right is about 80 points. Nine right is about 72. Your score is almost entirely decided by how many you get right, and barely at all by how you ranked them. Over a season the ranking gives back exactly what it costs.

3. So how much is being good at picking worth?

Less than you would like, because the spread is built to make these coin flips. A sportsbook sets the line precisely so that about half the money lands on each side. If the line were easy to beat, it would move until it was not.

We tested our own picks against two seasons of college football — 2,278 games — using the statistics that survive scrutiny. Result: 50.2% against the spread. You need 52.4% just to break even against the house edge. We are not good at this. Almost nobody is.

Here is what improving would actually buy you, in a ten-person pool:

Now hold that 50% row in your head, because the next section beats it for free.

4. How you actually win

You do not win by scoring well. You win by scoring more than everyone else.

Read that twice, because everything follows from it. If you and nine other people all pick sensibly, you will all score about the same, and which of you wins is decided by noise. Scoring well is necessary. It is not how you get separation.

Separation comes from being different and right at the same time. And here is where that ranking you agonized over finally earns its keep — not by raising your score, which it cannot do, but by deciding how much your differences are worth.

We simulated 120,000 weeks. Every game a true coin flip, you disagreeing with the crowd on eight of the fifteen. The only thing that changes is where your big numbers sit:

The mean score is identical to a tenth of a point — the ranking really does add nothing. The chance of winning moves from 14.5% to 22.4%.

Compare that against the previous table. Going from 19% to 22% by ranking well is worth about as much as improving your picking from a coin flip to break-even — and one of those two is free.

5. The three ways people lose

One: they copy the room.

It feels safe, and it is the worst thing you can do. If your sheet looks like everybody's sheet, you finish where everybody finishes, and only one person wins. In a ten-person pool, tracking the consensus closely drops your chance of winning a week to under 5%. Being averagely right is a way of losing quietly.

Two: they bury the only thing making them different.

This one is counterintuitive and it is the most common. People take their contrarian pick — the one game they see differently — and put 2 points on it, because it feels risky. That is backwards. The fewer games you differ on, the more it matters that those games carry weight:

With only four disagreements, weighting them properly more than doubles your chances. Hide them at the bottom of the sheet and you have thrown away the only edge you had.

Three: they protect a lead that is not there.

Somebody out in front starts playing safe. We checked it: a leader holding a 12-point advantage with ten weeks to play, who picks the consensus side on every game, drops from a 28% chance of finishing first to under 4%. A single week's score swings by about 17.6 points. Over ten weeks two players drift apart by around 79. A 12-point lead is a rounding error, and playing to protect it is how you lose it.

Read “playing safe” precisely, because it is easy to misread. It means siding with the crowd on every game — not copying one particular rival's whole sheet. If you truly duplicated another entrant, pick for pick and number for number, you would hold your lead over that one person for ever. You would also be helpless against the other eight.

6. Why people get this wrong

Because the scoreboard shows points, so people optimize points. It is the obvious thing to measure and it is not the thing you are trying to maximize.

You see the same confusion everywhere once you know the shape of it. A fund manager who trails the market buys the market, because being wrong the same way as everyone keeps his job while being wrong differently does not. A team up three scores runs into the line on third down — a play with terrible expected yards and excellent expected wins. A politician who is ahead stops saying anything.

None of them are playing badly. They are playing a different game to the one on the scoreboard, and so should you.

7. Our own pool, with the names taken off

We run an entry in a family pool. Everyone here is a letter, because the point is the shape and not who is who. Season totals through week three:

Two things worth noticing. The top three are separated by less than one good week, which is what "decided by noise" looks like in practice. And the biggest single factor in the bottom half is not bad picking — it is not filing an entry. Showing up beats strategy.

We publish our own picks before kickoff and grade them afterwards, including the bad ones. Our week-three ordering finished at the 6th percentile of what randomly shuffling the same picks would have scored. We put our big numbers on losers. That is in public because a strategy page written by someone who hides their results is worth nothing.

What this page does not claim

That any of it makes you better at picking games. It does not, and we have no evidence anyone in our pool can beat a coin. Every number here assumes you are guessing, because that is what our own testing says we are all doing.

This is how to lose slowly and win occasionally. If you want to be right more often, that is a different and much harder page, and we have not earned the right to write it yet.