So I’ll defer to you and maybe I’m too generous with my assumptions but you would think these would all modeled off PT translated to ABs or IP and then translated to all the categorical stats based on H%, BB%, K% type statistics (oversimplified here since I don’t feel like typing a ton ?)
Announcement
Collapse
No announcement yet.
Announcement
Collapse
No announcement yet.
best auction software
Collapse
X
-
Originally posted by Big Tymer View PostSo I’ll defer to you and maybe I’m too generous with my assumptions but you would think these would all modeled off PT translated to ABs or IP and then translated to all the categorical stats based on H%, BB%, K% type statistics (oversimplified here since I don’t feel like typing a ton ?)---------------------------------------------
Champagne for breakfast and a Sherman in my hand !
---------------------------------------------
The Party told you to reject the evidence of your eyes and ears. It was their final, most essential command.
George Orwell, 1984
Comment
-
It would indeed be an awesome study. I can hypothesize several possible approaches but the first that comes to mind would control for a population where expected playing time closely matched actual playing time (to control for player specific variance caused by injury). If say you isolated a subpopulation of just the players who hit their expected playing time projections (give or take) then you have somewhat a clean sample to then use for comparison purposes against all the main projections available. At that point, you could model for accuracy across sites probably as an index.
Not perfect but you would need to do something to just isolate a rep sample I think since there is no way to project outliers since by definition that outlier outcome would be outside the range of expected possibilities. I’d personally be more curious to see who is most accurate projecting when “noise” is eliminated basically.
Comment
-
Originally posted by Big Tymer View PostIt would indeed be an awesome study. I can hypothesize several possible approaches but the first that comes to mind would control for a population where expected playing time closely matched actual playing time (to control for player specific variance caused by injury). If say you isolated a subpopulation of just the players who hit their expected playing time projections (give or take) then you have somewhat a clean sample to then use for comparison purposes against all the main projections available. At that point, you could model for accuracy across sites probably as an index.
som
Not perfect but you would need to do something to just isolate a rep sample I think since there is no way to project outliers since by definition that outlier outcome would be outside the range of expected possibilities. I’d personally be more curious to see who is most accurate projecting when “noise” is eliminated basically.---------------------------------------------
Champagne for breakfast and a Sherman in my hand !
---------------------------------------------
The Party told you to reject the evidence of your eyes and ears. It was their final, most essential command.
George Orwell, 1984
- Likes 1
Comment
Comment