Keep fanciful predictions to the bare minimum – goal averages point the way to go



Kevin Pullein

ONE of the reasons footballis so popular is that much of the time it provides what most sports fans crave – genuine contests.

In two out of every three games played in the Premier and Football Leagues, the final whistle blows with no more than one goal separating the teams. Theoutcome is in doubt right until the very end.

The commonest scores are the low ones – in other words, not the more fanciful ones that tend to be predicted by over-optimistic supporters on their way to a ground.

The six commonest scores in the Premier and Football Leagues are those in which the home team scores no more than two goals and the away team scores no more than one goal. The next three are those in which the home team scores either nought, one or two goals and the away team scores two goals. In nearly three-quarters of all games played in the Premier and Football Leagues, nobody scores more than two goals.

Is there a way, though, of estimating the chance of any individual score occurring in any individual game? I hinted at it last week when discussing the next goal market – the football betting market to which all others are related.

I gave an admittedly rough description of how you can work out how many goals are likely to be scored in a game and how those goals are likely to be split between the two teams. It was deliberately rough: I wanted to offer suggestions that might be helpful to people who have neither the time nor the will to spend hours sat in front of a powerful computer.

A word, though, on a term I am about to introduce: expectation. In a betting context, expectation means what we think would happen, on average, if the fixture could be replayed a number of times.

So let's imagine a game on which we have put a goals expectation of 2.5, with 60 per cent to be scoredby the favourites and 40 per cent by the outsiders. We have given the favourites a goals expectation of 1.5 (2.5 x 0.6 =1.5) and the outsiders a goals expectation of 1.0 (2.5 x 0.4 = 1.0). We are saying that if this contest could be repeated a large number of times we believe the average number of goals scored by the outsiders would be 1.0.

However, what we need, of course, is a way of being able to estimate the chance, in this particular match about to be played, of the outsiders scoring no goals, exactly one goal, exactly two goals and so on. The Poisson distribution provides this.

The Poisson distribution can tell us how likely it is that a thing will happen a specific number of times, provided we know how often that thing happens on average. Itcan tell us, therefore, how likely it is that a football team will score no goals, exactly one goal, exactly two goals and so on, provided we know how many goals that football team score on average. I have given the answers in the tables above.

You might guess that the prospects of a match finishing goalless would be the prospects of the favourites failing to score multiplied by the prospects of the outsiders failing to score, and you would be almost right – but here the difference between being almost right and completely right is an important one.

In football matches, draw scores occur slightly more often than we would have anticipated if we simply multiplied together the prospects of both teams scoring the same number of goals, and other scores occur slightly less often.

In the Premier and Football Leagues during the last ten seasons, for example, 23 per cent of home teams failed to score and 34 per cent of away teams. The proportion of games finishing 0-0, however, was not 7.8 per cent (which is what you get if you multiply 23 per cent by 34 per cent) but 8.3 per cent – equivalent to odds of 11-1 rather than 12-1.

And there were similar differences for 1-1, 2-2 and so on. In short, the above table will hopefully be of some use to you, but for it to be of some use you will need to be aware that draw scores will occur slightly more often than it suggests and other scores slightly less often. It could be worth knowing.

One of the most frequently-asked questions of our ever-popular sport is not so much “who'll win” but “what will the score be?”

HOW MANY GOALS WILL THE TEAM SCORE?
Percentage of games each number of goals is scored
Goal Average 0 1 2 3 4 5 6 7+
0.5 61 30 8 1 0 0 0 0
0.6 55 33 10 2 0 0 0 0
0.7 50 35 12 3 0 0 0 0
0.8 45 36 14 4 1 0 0 0
0.9 41 37 16 5 1 0 0 0
1.0 37 37 18 6 2 0 0 0
1.1 33 37 20 7 2 0 0 0
1.2 30 36 22 9 3 1 0 0
1.3 27 35 23 10 3 1 0 0
1.4 25 35 24 11 4 1 0 0
1.5 22 33 25 13 5 1 0 0
1.6 20 32 26 14 6 2 0 0
1.7 18 31 26 15 6 2 1 0
1.8 17 30 27 16 7 3 1 0
1.9 15 28 27 17 8 3 1 0
2.0 14 27 27 18 9 4 1 0
2.1 12 26 27 19 10 4 1 0
2.2 11 24 27 20 11 5 2 1
2.3 10 23 27 20 12 5 2 1
2.4 9 22 26 21 13 6 2 1
2.5 8 21 26 21 13 7 3 1
MOST COMMON SCORES
Score Percentage of games in which it occurred
1-1 12.8
1-0 11.1
2-1 9.4
0-0 8.3
0-1 8.2
2-0 8.0
1-2 6.4
2-2 5.3
0-2 4.4










Copyright @ 2007 Centurycomm Limited or its licensors. All rights reserved.

High scores are indeed rare. I liked the system they had in one prediction league I was in where you got one point for the correct outcome (HW/D/AW), an extra point for the correct score, a further point if the correct score involved more than three goals (4-0/3-1/2-2 or higher), and yet another point if the aggregate involved more than 5 goals (e.g. 2-5, 8-1).

It seems reasonable to me to consider it harder to predict a 3-1 win than a 1-0 win and harder still to predict a 3-3 draw.