
In effect, it's the same argument as saying that not all of a player's previous games should be taken into account, but (say) the last 5. Not least because the SQL query is a bit tricky to write  can't think of a way to do it without at least one and possibly multiple intermediate temporary tables  I haven't tried it yet.
Whatever I try, I want to look back on previous results with the new picker criteria and see if I can get some metric which predicts the results better than the current one.


I have for some time taken Jason's view.


At the moment I gather the teams are picked based on the total team average, eg sum of all the players' points divided by the sum of all the players' games. Doesn't this mean that long term players swamp more recent players? The numerator and denomenator are going to be completely dominated by the large values from the players that have played for ages. It would be better if instead the team average was the average of the individual player averages.
