Correct score predictions: how often does anyone get one right?

Calling the exact score is the purest version of football prediction and the most humbling. Everyone remembers the 3:1 they nailed in October. Nobody remembers the eleven weeks around it.

So what is the real hit rate, and what should a prediction game do about it?

The honest number

There are roughly a dozen scorelines that account for most football matches, and the rest of the probability is spread across a long tail of results nobody sensibly predicts. Even a well-informed predictor who always picks a plausible scoreline lands the exact result somewhere in the region of one match in eight to one in twelve.

That is the ceiling for a good predictor over a long run, not a bad week. Anyone claiming a third of their exact scores over a season is either counting differently or remembering selectively.

Useful way to think about it: on a twelve match coupon, one or two exact scores is a normal week. Three is a good week. Four is the sort of week you tell people about for a year.

Why the hit rate is lower than it feels

• The favourite winning is not the same as the score. You can be right about everything that matters and still be wrong, because the third goal went in during added time.
• Draws break patterns. A game can be a comfortable win for eighty minutes and finish 1:1.
• Goals are lumpy. Football's scoring rate is low enough that one deflection changes your prediction from correct to nearly correct.

That last point is the whole design problem for anyone scoring a prediction league. If exact scores are the only thing worth points, most weeks are dead for most players.

How good scoring systems handle it

The fix is partial credit that still respects the exact call. Take one prediction and look at what it earns as the result drifts away from it:

Our scoring layers those, so a near miss is worth having and the exact call is still the prize. The exact score is where the big numbers come from, especially when it is a scoreline nobody else picked.

What to do with a one in ten event

Two habits separate people who enjoy correct score prediction from people who quit in November.

1. Judge yourself on the season, not the week. A one in ten event produces long empty runs. That is arithmetic, not a slump.
2. Care about your near misses. Getting outcome and margin right most weeks is a real skill and it wins prediction leagues. Chasing only the exact score is how people end up predicting 4:3 in March out of boredom.

The people at the top of a well-designed table are almost never the ones with the most exact scores. They are the ones who are rarely completely wrong, and who bank the big multipliers when they do land one.