Quote:
Originally Posted by PeteMoss
If you can show me any proof that using a smaller sample of goalie performance instead of the full season is a better predictor of future results than I will take this seriously. Or if it is that Jan 1st should be the cut-off instead of Dec 1st or Feb 1st.
Until then I will take larger full season sample over smaller 2.5 months sample.
|
Just spitballing here, but I think the best thing would be a metric that is weighted over time:
Last 10 games started: weigh at 100%, except for the best and worst game, which you weigh at 50%
Previous 10 games started: weigh at 75%, except for the best and worst game, which you weigh at 37.5%
Previous 10 games started: weigh at 50%, except for the best and worst game, which you weigh at 25%
Obviously you can adjust the number of games and weightings any way you'd like, but I think this would give the best sense of recent performance without completely discounting the past.
I wouldn't consider any statistics of goalies when they come in in relief (though it would be interesting to know how that impacts stats, whether it's in relief to injury or poor play)...unless you're planning to treat them like pitchers with mid-game switches, I'm not sure there's much point in knowing how good a goalie is without typical starting goalie game day preparation.