Gameweek 8 Preview 2.0
I am away this weekend so I'm not going to have time to write up how the below numbers have been generated, but it's essentially the format prescribed in the recent goals and assist pieces, only limited to teams for now. I'm posting the data now so we will have a real test set of data from this weekend to work with, which will hopefully highlight a couple of areas where the model could be improved (and by that I mean where the process, rather than the result, is wrong. If Stoke hold United to no goals on 27 shots, the model isn't necessarily broken, it was just a bad day at the office for van Persie and co).
xG Shots - expected goals scored based on the shot data they have registered and the shots given up by their opponents, regressed to a team average conversion rate (not a league average).
xG Chances - expected goals scored based on the chances created by a given team and the chances surrendered by the opposition, regressed on a team average conversion rate (not a league average).
Finally, apologies for the lack of reader questions this week - normal service will be resumed for GW9 so get them in early on Twitter, Facebook.
xG Shots - expected goals scored based on the shot data they have registered and the shots given up by their opponents, regressed to a team average conversion rate (not a league average).
xG Chances - expected goals scored based on the chances created by a given team and the chances surrendered by the opposition, regressed on a team average conversion rate (not a league average).
Finally, apologies for the lack of reader questions this week - normal service will be resumed for GW9 so get them in early on Twitter, Facebook.
Comments