If you're looking for casinos that accept Trustly then we suggest visiting

Friday, 18 March 2011

Using Past Performance Data in Football

Football is big business in the gambling world and you have only to read a few blogs to notice they are part of portfolios or dedicated tipsters plying their trade. Well, I found this article, which makes for interesting reading.

Ninety per cent of what happens on the football field can be explained by the ability of the players, as revealed in the results they have achieved. I do not clutch this figure out of the air, by the way, as I shall explain in a moment.

The all-important question, of course, is how we interpret these past performances and convert them into odds that express what is likely to happen in the next contest.

I will explain how I evaluate past performances in my ratings. I will also tell you a couple of ways in which you can develop your own sets of match odds intuitively, without a computer - or even, for that matter, a paper and pen.

I will try to give you ideas on how you do things yourself, in your own way. The ratings I compile are used to produce estimates of the likelihood of different results occurring in games from more than 20 countries across Europe, not only in England and Scotland but also Spain, Italy and Germany and elsewhere from Turkey to Denmark and Portugal to the Ukraine.

I constantly compare my pre-match forecasts with post-match results, using standard statistical tests to check how accurate they are. And what these tests tell me is that around 90 per cent of the post-match results are explained by the pre-match forecasts.

I do not say this so that you will think my ratings are wonderful. They are not, really. All they do is offer an estimate of how well a team are likely to perform today, given how well they have performed in the past. I say this to convince you that the single most important influence on the result of a football match is the ability of the players.

In my computer ratings process I assess past performances according to the score, the quality of the opposition, the venue and the date. I do not think you will be surprised by this. On the contrary, I am sure almost everyone will agree that these are the variables which matter. The bigger the margin of victory, the more creditable the performance. The better the opposition, the more creditable the performance. A home win is less notable than an away win, because nearly all teams record superior results on their own ground. The last performance is more significant than the one before, and so on.

Do not get too hung up, though, on individual performances in isolation, or even small numbers of performances in isolation.

In Euro 2004, Italy and Spain were knocked out at the group stage. I do not accept that Italy were a worse team than Sweden and Denmark, and I am speaking here as someone who advised a sell of Italy on the group indices. Nor do I accept that Spain were a worse team than Portugal and Greece.

In a low-scoring game like football, the best team win less often than in other sports. In a three-match mini-league, it is comparatively easy for a good team to finish behind inferior opponents. In snooker, the best player wins less often over a first-to-five-frames format than he does over the first to 17.

In my ratings for club football I include the last 32 performances. I have found this to be the optimum number, having tested lots of others. Thirty-two games is almost a whole season in the Premiership and Scottish Premier League, and more than two-thirds of a season in the Football League.

The last performance, as I mentioned, is more significant than the one before, but it is not that much more significant. In my ratings, the most recent performance is given only three times as much weight as the one that occurred 32 games ago. In general, large samples of results are more revealing than small ones. I promised earlier that I would describe a couple of ways in which you can assess teams intuitively.

The first involves trying to relate teams to others in their division. If they played for a whole season how you think they are playing now, where do you think they would finish? Would it be in the middle? Higher or lower? How much higher or lower? At the top? At the bottom?

In both the Premiership and the Football League, teams who finish a quarter of the way down a table - fifth in the Premiership, eighth in any division of the Football League - usually have an average goal difference per game of +0.3. That is to say, if you subtract the number of goals they concede from the number of goals they score, and divide the result by the number of games they play, you will usually end up with a figure of around +0.3.

Teams who finish three-quarters of the way down a table - 15th in the Premiership, 18th in the Football League - usually have an average goal difference per game of -0.3. Teams who finish at the top normally have an average goal difference per game of +0.8, those at the bottom -0.8 . In the Premiership, the extremes are even further from the centre - teams who win the Premiership usually have an average goal difference per game of around + 1.1, those who finish last around -1.1.

In theory, a team with an average goal difference per game of +1.1 will win 62 per cent of their games, draw 23 per cent and lose 15 per cent, gaining an average of 2.1 points per game (80 over a 38-game Premiership season). The total number of goals in their games will average 2.8. In practice, these figures will often vary a bit.

You will have noticed that so far I have discussed only games between teams from the same division. What, you might ask, about the FA Cup or the League Cup?

I say: assess a team first within the context of their own division, then allow for the differences between divisions. A Premiership team would normally have an average goal difference per game 1.0 higher if they played in the Championship, 1.7 higher if they played in League One and 2.4 higher if they played in League Two.