Here’s my presentation on Game States at the 2015 OptaPro forum. During questions, Danny Altman asked about the effect of minutes played. The model I discussed at the forum contained a main effect for minutes played, and as expected, showed that shooting intensity increased later in the game. But it assumed that the effect of Game State was constant throughout the match.
Following Danny’s question I had a look at this, and the answer is that the effect of Game State does vary throughout the match. Teams that are ahead early in the game don’t tend to reduce their shooting, but when they are ahead later in the game they do. That certainly confirms our intuitions.
For now, I’m posting the original results without this effect. I think the results still stand as far as they go, but I’ll develop an updated model as time permits.
Goals Change Games