Quote:
Originally Posted by Huckleberry
Just one way to do it.
|
Wouldn't it be simple to just project a PECOTA-like weighted mean projection using the last 3 years of stats. As the plate appearances accumulate in the current season more weight is moved to the current year stats and weight is removed from the projection.
Project 250/330/425
Stats through 300 PA 300/400/500
At 300 PA the stats are weighted 60/40 and the AI is valuing a player somewhere around 280/365/470 (guessing as I'm lazy to actually calculate).
Now that I think about it projections might fix many of the games issues. How to value stats versus ratings. How to value players for trade/waiver. Have an internal rolling 5 year WARP projection for each player and that might make it much easier for the AI to properly value prospects and players with big contract.
Another possible way to fix the release of the Cy Young type player is to tie popularity to actually being on the field and not just in the organization. If Award winners become highly popular and you need them to be on the field to actually drive attendance/revenue then that gives the AI another reason to not release an all-star pitcher the next season after a quick ratings crash.