The AI is releasing all the amateur draftees!

[CharlesC]

I am playing a league that began in 2007, using historical players from 1965. After the first season, I drafted fictional players rather than historical ones. The problem is, a couple of months later, the AI released virtually all of the newly drafted players. Granted, they all had very poor current skills, but many of them had excellent potential ratings. I have set the AI’s weights for evaluating players at 10% potential and 30% each for: this year, last year and 2 years ago. I did so because (1) I want to emphasize past records over future potential and (2) I was under the impression that if no stats existed from, say, 2 years ago, the AI would place more weight on the player’s potential.

Is the problem my weighting system of 10%/30%/30%/30%? If so, what should I change the weights to in order to maximize the value that the AI places on past performance? If my weighting system isn’t the reason why the AI released all these new players, then what do I do in order to prevent this from happening again?

Related questions: if the season starts on April 15, 2007, on April 15 I presume that the AI considers 2007 to be "this year". But on April, 14, does the AI consider 2006 to be this year? How about on December 31, 2006?

Thank you,
Charles