How to Play This Thing

A short guide, with the arithmetic done. The short version: stop agonising over who wins, and start thinking about how different your sheet is from everybody else’s.

The format

Fifteen games. Fifteen picks. Points fifteen down to one. You play every game on the card — there is nothing to drop and nothing to choose.

And critically: picks are against the spread, not straight up. The rules say it plainly — you must pick the team that beats the winning margin. Taking a 28-point favorite does not mean taking a near-certainty. It means betting they win by 29.

Your weighting cannot help you

This is the part that surprises people, so here is the arithmetic.

Every game against the spread is close to a coin flip. That is not an opinion — we measured it across 2,140 college games and got 51.0%, with a margin of error that comfortably includes 50.

You always spend the same fifteen point-values. So whatever order you put them in:

Your score distribution is identical no matter how you assign the points. Ordering cannot help you and cannot hurt you.

What does matter: being different

If everyone scores about the same, you cannot win by being more accurate. You can only win by being different and right on the same games.

The charts below assume eight people play, which is what happened in week three. Ten are registered and turnout has varied. A bigger field lowers everybody’s numbers — with ten playing, a fair share is 10% rather than 12.5% — but it does not change the advice. Copying the room is the worst option at every field size, and it gets worse as the field grows.

Copy the room and you cannot finish above it. Drag the slider.

Watch the left chart refuse to move. Your own score has the same shape at every setting — centred on 60, give or take 18. You cannot make yourself score better by arranging the points differently, and the chart says so more honestly than we could.

The right chart is the one that changes. At zero it bunches in the middle: you finish somewhere unremarkable, almost every week. Slide it right and it splits in two — more chance of winning, and just as much more chance of coming last. That is the trade, stated plainly. You are not buying a better result. You are buying a wider one.

The climb is steep and then it stops. Nought to three roughly triples your chance of first; five to fifteen adds almost nothing. Three or four genuine disagreements is most of the benefit there is.

So, four rules

1. Stop agonising over who wins.

Nobody in this pool picks above about 53%, and week three showed the angriest-mascot method matching the rating model exactly. The hours go into the wrong question.

2. Put your big points where you disagree with the room.

Not randomly — on the games where you actually see something. The model of this was the Virginia Tech pick in week three: a real reason nobody else had, priced at one point because it could not be proven. It landed.

3. Play the standings, not the games.

Behind in the season race? You need variance — differ hard, because finishing a close second is worth nothing. Ahead? Match the field and let your lead do the work.

4. Write down why.

The outcomes are noise at fifteen picks a week. The reasons are the only part that compounds, and by November we will know whose reasons were doing work and whose merely sounded good. Send them when the picks lock.

What this guide cannot tell you

Whether any of it wins. The differentiation numbers come from a simulation that assumes every game really is a coin flip and that the group clusters near a consensus — both roughly true, neither exactly true.

And “your ordering cannot help you” is exact only under that coin-flip assumption. It is a null model, not a law. If you genuinely know that one game is likelier than another, putting more points on the likelier one does raise your expected score — that is simple arithmetic. The reason we use the coin-flip model anyway is that nobody here has yet shown they can tell which game is which. Three weeks in, the evidence says the assumption holds. It is an assumption all the same, and the moment somebody demonstrates otherwise this page is wrong for them.

Audit of this page by Codex, 23 September, which caught both points above.

Arithmetic and simulation by The Sports Page. Coin-flip figure from Issue #174, measured over 2,140 games. Win-chance curve from 60,000 simulated weeks per setting, eight entrants, field clustered at 75% agreement with a consensus sheet. Not a betting guide — there is no money in this pool, which is precisely why it is fun.