Open any preview page and there it is at the top: five or six coloured boxes showing recent results. Green for a win, grey for a draw, red for a loss. It's the most prominent piece of information on the page and it's close to worthless.

Here's why, and what to look at instead.

The sample size problem

Six matches. Against six different opponents of varying quality, three or four at home, some in cup competitions with rotated squads, possibly spread across two months with an international break in the middle.

Compressing that into six letters throws away every piece of context that made the results what they were. A team that lost narrowly away to the league leaders and beat a struggling side at home reads exactly the same as one that did the reverse — L, W in both cases.

Statistically, six matches is nowhere near enough to distinguish a genuine change in a team's level from ordinary variance. In a sport where the better team loses 30 or 40% of the time, a run of three defeats is entirely consistent with nothing having changed at all.

The regression trap

This is the practically important bit. Teams on hot streaks tend to cool, and teams on bad runs tend to improve, and this happens more often than not.

The reason is that extreme short-term results are usually produced by a combination of genuine quality and good or bad fortune. Quality persists. Fortune doesn't. So the next six matches revert towards the team's actual level.

Which means that using recent form as a predictor often points you in exactly the wrong direction. The team that's won five in a row is the one most likely to underperform its recent record next.

I've seen this framed as a paradox and it isn't one. It's just what happens when you select on an extreme value.

What produced the form

The useful question isn't what the results were but what generated them.

A team that's won four in a row while creating very little and conceding a lot has been fortunate, and that will end. A team that's lost four while dominating the underlying numbers has been unlucky, and that will also end.

The way to check this is chance quality — how good were the shots created and conceded. If a side is producing 1.8 expected goals a match and scoring one, they'll start scoring more. If they're conceding 1.6 and shipping 0.8, they'll start conceding more.

None of this is complicated and it's the single most useful adjustment you can make to reading form.

The schedule matters more than the results

Six results against six opponents means the opponents are half the information and they're the half that gets dropped.

A run of four wins against teams in the bottom five is a different thing to four wins against the top five, and there's no reasonable way to compare a team's form to another team's form without accounting for who they played.

Some sites publish schedule-adjusted form, which is genuinely better. Most don't, because the coloured boxes are simpler and people are used to them.

When form is real

To be fair, there are situations where a change in results reflects a genuine change, and they're identifiable.

A tactical change. A manager switches to a back three and the team's underlying numbers move at the same time as the results. That's a real shift, not variance.

A key player returning or leaving. If a team's best creator has come back from injury, the improvement is causal.

A new manager, sustained over ten-plus matches. Not the initial bounce, which is mostly noise, but a durable change in how a team plays.

A squad depleted by injuries. A team missing five first-choice players is genuinely worse right now, and that's not going to revert until they're back.

The pattern is that real form changes have a mechanism you can point at. If you can't identify what changed, probably nothing did.

The psychology question

The obvious objection is that confidence is real, that teams on a good run play with belief, and that this is a genuine effect.

I'd accept that something like this exists, though it's very hard to isolate from the statistical regression that would happen anyway. The studies on momentum in football find weak effects at best, and much of what looks like momentum turns out to be quality reasserting itself.

My honest position is that confidence effects are probably real and probably much smaller than the way we talk about them suggests. Certainly smaller than the effect of playing at home, or of having your best striker available.

What I'd put at the top of a preview instead

If I were designing the page: chance quality created and conceded over the last fifteen matches, adjusted for opposition. Squad availability weighted by player importance. Rest days since the last fixture. And a note on what the team has actually been trying to do tactically.

That's four pieces of information, all of which predict better than the form guide, and none of which fit into coloured boxes. Which is presumably why we still have the coloured boxes.