Closed hkenawi closed 3 months ago
xPts is the sum of the following 3 categories:
- xG x PPG (Points per goal)
- xA x 3
- xCleanie x PPC (Points per clean)
To properly predict xG, I believe the model must contain match data for every player in the premier league and have the following features:
Opponent Fixture difficulty Home/away Average shots per game for the szn Avg distance of shot Shot % on target Conversion rate Opposing team goalie save % Opposing team clean sheet % Opposing team xGC xG(Y var for training)
xPts is the sum of the following 3 categories:
- xG x PPG (Points per goal)
- xA x 3
- xCleanie x PPC (Points per clean)
To properly predict xG, I believe the model must contain match data for every player in the premier league and have the following features:
Opponent Fixture difficulty Home/away Average passes per game for the szn Final 3rd passes per 90 Chances created per 90 Big chances created per 90 Passes into penalty area per 90 Shot creating actions per 90 Average conversion rate of starting forwards Opposing team goalie save % Opposing team clean sheet % Opposing team xGC xA(Y var for training)
xPts is the sum of the following 3 categories:
- xG x PPG (Points per goal)
- xA x 3
- xCleanie x PPC (Points per clean)
For xCleanie, I believe the best solution would be to leverage ready made resources such as bookie odds. Those models are already trained on the highest level of data obtainable and would lead to a more efficient project workflow. No need to recreate the wheel.
With the following 3 components of xPts laid out, this issue can be closed.
xPts is the sum of the following 3 categories: