The next step in football analytics: Game State and Game Pathway

Posted on: April 15th, 2016 by Jonny Grossmark

If you have an interest in  football data and analytics then you may have read about Expected goal totals  ( ExpG) which is not the best predictor of future performance.

 We can look at the example of the game between Manchester United and Fulham in 2014 where United had 31 shots and 9 shots on target and failed to win ( FT 2-2 ) .

The ExpG for Manchester United against Fulham was around 4.4 goals which indeed is an irrelevance in terms of expectation of future performance.

It is known that expectation of a goal is dependent on the current score ( game state ) but what ExpG has failed to consider is the effect of game state added to any variables that may occur such as an early goal on accuracy as time decays in a game.

If you look at 0-1 half time game state added to the goal 0-20 minutes in the Premier League , a home win is not common even if the home team is the perceived stronger team .

It is game state analysis that looks at individual leagues and the major half time game states ( 1-0 0-1 and 1-1 ) to define expectation of accuracy as time decays in the second half.

Game Pathway 

If you look at the Premier League for example in games where the home team open the scoring 0-10 minutes , the pathway 2-0 is far more likely then 1-1 and in effect you can look at individual leagues and map out the expected pathway as time decays in a game by factoring the variables that may occur in a game .

The Second half fightback metric

I define the second half fightback metric as games that are 0-1 /1-0 at half time and the pathway goes to 1-1 with the second goal in the 46-70 minutes time band with the  aim of understanding how accuracy will be effected by the fightback metric in individual leagues .

Summary

If you are betting on football using a  predictive model based around shot data then my argument is that it will have severe limitations because shot data is not repeatable on a game by game basis as each game will have in general different variables that may occur which may impede or indeed accelerate goal production.