Predicting goals – Part one

To get some context into this post read the goal ratings thread…

When you’re looking at a football match, goals are obviously of great importance. Every strategy you deploy on a football match is reliant to some extent on whether a goal is scored or not.

Starting point

The first place to start when you trading in football is to work out if you think there will be a goal, obviously.

Low scoring rates

High quality matches between even teams, where the away team is slightly stronger, tends to produce draws. But high quality teams tend to possess quality all round and therefore a goal is likely, subject to the defence of the opposing team. However low quality matches between two poor teams can produce the same results and tends to produce fewer goals. But the common denominator here is between equally matched teams, where the away team slightly stronger.

My range of prediction

When I look a football match the first thing I do is try and predict how many goals I think there will be. My prediction is centred on the long-term historical average for goals. That long-term average is somewhere between 2.5 and 2.6 goals, depending upon the league or competition. When I first started doing this in the 80’s nobody was attempting it, now there are quite a few sources and opinions. But I still make the effort to run through mine, even if they do or don’t agree with the market.

Now it seems odd that you would forecast 2.64 goals, but what this forecast is telling you is if we played this match 100 times how many goals would be scored. In the case of a forecast of 2.64 we are saying there would be somewhere between 26 and 27 goals scored in this match over 100 games.

Short term inaccuracy

We all know that they can’t be .64 of a goal in a game so this forecast is based around finding the average number of goals. This is where football becomes tricky, because you will probably end up with two or three goals and therefore you will always be out to some extent on a specific set of results. But football is a sport that plays out over a long period of time, so all my positions in football are based around being accurate in the long term forecast but most likely being inaccurate during a set of matches. I’m not fussed about this because I know it will average out over time.

Competitive vs. uncompetitive matches

If I get a very competitive match between two top teams I will very likely to nudge my forecasts toward two goals. If it is an uncompetitive match then I will nudge the forecast towards three goals and in a very uncompetitive scenario you may find me lift the forecast to 4 goals. Again, I’m not saying that there will be this precise number of goals within the match but I’m saying if this match was played repeatedly over a long period of time this is where the average number of goals would sit.

Group forecasting

This is why often group forecasts together, by doing this you can see the goal forecast average out fairly quickly. Last night I predicted 23 goals in total across all matches and we end up very close to that number.

I’ve spent a ball achingly large amount of time working on the average number of goals and over the course of the season it tends to be very accurate. Last year I was 0.07 goals out in the scheme of things, on average. Generally though this isn’t because of some amazing foresight, though I do have a complex model that runs on each scenario, it’s more to do with the variability within football.

Coping with variability

I could forecast three goals, but one could be a dodgy offside decision, one a blatant handball and the other a missed penalty. In the models terms, a value is assigned to each of these scenarios so it can reasonably discount what the “true value” of the actual match was, regardless of whether it actually occurred or not. That’s the way football works! But thanks to this variability, you can often look a right fool with some forecasts.

Summary

But the first stage of doing something in a football match is to come up with a good estimate of how many goals there will be. There were some really good sites that had information on this value, but incredibly most of them have vanished. I’m pondering introducing one myself, if I can find the time.

Next up, how the number of goals influences the time of a goal.

Tags: football, forecast, goals

Category: bookmark, Football trading strategies, Trading strategies