Premier League Betting: Ratings can also predict goal-count

The market thinks Chelsea v Man City is more likely to produce three goals than two. The statistics say the market is wrong.

In his ongoing quest to perfect his Premier League ratings, Jack Houghton thinks he has stumbled across a way to predict the likely goals in a match...

A couple of months ago - earlier on in this journey to discover whether a football-know-nothing like me could produce a profitable Premier League punting model - I wrote that I'd been testing the veracity of my ratings by seeing how predictive they were of past football matches.

What I found was that, although the model was incredibly accurate when it came to predicting the results of matches where teams were more-closely matched, it became less accurate in lop-sided matches.

As so often happens when dealing with data, when you're looking for one thing, you often find something else interesting that had nothing to do with what you were looking for in the first place.

On this occasion, the thing I noticed was that the more lop-sided the matches, the more goals that were scored, but that not nearly as many were being scored as I would have expected.

What I decided to do, then, was to chart the average goals that were scored by teams in all matches, given a certain ratings' supremacy. Here's a simplified version of what I found:

A couple of things stand out in this analysis for me.

First, the frequency with which superior home teams score goals ramps up rapidly when compared with away teams, which suggests that away teams are tactically more likely to be cautious when taking the lead, even when they are obviously superior to the home team; and also suggests that the antithesis - that superior home teams become more aggressive as they score goals - is also true. This knowledge is useful to those punters betting in-play.

Second, it gives a very quick - although admittedly simplistic - way of assessing the likely total goals in any match. Assuming, of course, that the causal relationship between supremacy and goals is as strong as I think it is.

Take a look at the following. It shows the next round of Premier League matches, followed by the percentage chance of a home win according to my ELO ratings:

Hull City v Newcastle United 34%
Crystal Palace v Everton FC 35%
Liverpool FC v West Ham United 51%
Manchester United v Leicester City 73%
Stoke City v Queens Park Rangers 61%
Sunderland v Burnley 52%
West Brom v Tottenham Hotspur 20%
Chelsea v Manchester City 45%
Arsenal v Aston Villa 79%
Southampton v Swansea City 60%

Comparing these percentages to the likely-goal chart above helps to identify some interesting bets. I'll recommend a few below, but will leave it up to you to play further. Enjoy.

Recommended Bets

Back Over 2.5 Goals @ 2.47/5 in Hull v Newcastle
Back Over 2.5 Goals @ 2.285/4 in Crystal Palace v Everton
Back Manchester United to win 2-1 @ 10.09/1 against Leicester
Back Arsenal to win 2-1 @ 10.09/1 against Aston Villa
Back Under 2.5 Goals @ 2.0621/20 in Chelsea v Manchester City