Here's the problem. Football pundits go on and on about form, and it is true that teams have runs of wins and losses and also true that, over time, some teams get better and others get worse.

Yet there are copious numbers of academic papers that show that form is a myth. Winning one game makes you neither more nor less likely to win the next. If a striker scores in one match, that does not make him more likely to score in the next. The only sport that shows evidence of a “hot hand” effect is tenpin bowling, and even that disappears almost before you notice it.

So, how can both of these things be true? The Fink Tank detective agency, magnifying glass at the ready, has been poring over the crime scene. Making use of the chess rating system designed by Arpad Elo, the Hungarian-American mathematician, and Mark Glickman's work on parameter estimation, the Fink Tank model was re-engineered by Dr Henry Stott and Dr Ian Graham. This allowed us to update ratings for each team only by knowing their own result, rather than, as in our full model, knowing the result of every team in Europe for the past five years.

It immediately became apparent how slowly each team's ratings changed. Even a 3-0 victory for Middlesbrough against Bolton Wanderers would shift Middlesbrough's attack ratings by only 3 per cent. It is only over a number of matches that the ratings change significantly.

Taking Middlesbrough again, for each game in which they fail to score you might think a little less of their attack. After one match their attack effectiveness rating is down by 3 per cent, two matches 5 per cent and so on. After ten matches without scoring, it is down by 19 per cent.

However, alongside this you have the problem of uncertainty. A match in which Middlesbrough fail to score might just be down to a bit of bad luck. The goalless outcome might be telling you nothing at all. Football games vary a great deal. Even if we expected the home team to score 1.5 goals, they might score, say, four goals just through serendipity. That happens 7 per cent of the time.

And on top of this, teams vary in consistency. A team with a given standard of attack will play a little bit differently every time. Look at the graphic. It shows you that Reading's attack varies by 22 per cent. Given the quality of Steve Coppell's team, the Reading you encounter in any one match may have an attack up to 22 per cent better or 22 per cent worse than their average standard.

Right. Now we can pull everything together and say something about form. If a team changes from one match to the next, you are probably seeing luck and inconsistency, not a real change in quality. This so-called form is really just a trick of the eye.

But, after a few matches, you can begin to believe that the change is real. How many matches? An example might help. After six goalless games, Middlesbrough's attack rating would have gone down by 13 per cent. But we know that by luck alone it varies from match to match by up to 14 per cent. So only after seven games, when the Middlesbrough attack has gone down by 15 per cent, should we even start to think that we are seeing a real downturn, not just a bit of week-to-week luck.

Seven whole games without scoring. You have to wait a long time before you begin to talk about form as if it were telling you something.