I’ve been looking at ways to use previous scores to try and predict upcoming fixtures, to attempt to understand team form in a more scientific ways. There are some versions of this already being used, but I wanted to get into the detail as much as possible.
This should allow me to look in more detail at where goals are coming from, and who against, and therefore which players to target. After trying a couple of ways, I’ve decided to use the average goals scored and conceded by each team so far this season, home and away, based on the weekend fixtures.
avg home goals scored*home games played+avg away goals conceded*away games played/total number of games
avg away goals scored*away games played+avg home goals conceded*home games played/total number of games
The data isn’t perfect (of course), partly due to the variations not included (quality of opposition, in-game scenarios etc.) and party down to the small sample size. At this stage the sample has resulted in fewer clean sheets than I would estimate, for example.
Over time though this will become more accurate and my plan in another couple of weeks will be to use the most recent 5 weeks’ data as a ‘form marker’ against the overall season figures. I’ll also at this point likely remove the highest and lowest scores, to reduce the anomalies.
I may develop this and add in some additional factors, but I think this is a good starting point.