Quote:
Originally Posted by Questdog
Your points are valid, except for one thing: OOTP does not work like you are expecting it to. Allowing the A.I. to only evaluate players on ratings would seem to be the logical way to get it to field the absolute best team at any given moment. But this is not how it works in practice. You need to have this year's stats included in the formula and to a pretty high degree, in order to compensate for it's inadequate evaluation of the ratings.
As for this year's stats throwing a monkey wrench into the equation because of small sample size, this is not a problem in OOTP, because this year's stats mean nothing until an adequate sample size has been achieved.
I do not like this. A better solution, to me,would be to count all results from the last 3 seasons, regardless of sample size and just weigh each plate appearance in the seasons differently. For instance, this year's stats each count 5 times more than last years, which count 5 times more than the year before. If a player had 500 AB last year, then last year's results will prove to be more important on decisions until he has 100 AB this year, at which point the two seasons will be equal in importance and as the season progresses, this year's stats will continue to grow in importance.
But that is not how OOTP works. It is all or nothing. Not enough stats and they have no value regardless of the settings.
|
that's not quite what i said... i said it is more strongly correlate than any 3 year percentage of stats (results).
i do use "this years" stats, just for different reasons... it's basically the same as the others 25 / 20 / 5. light on the third year, heavy on recent 2.
this years stats will not be an adequate sample size, that's the problem. the % error on many things will be quite large even in July when you are making really important roster decisions....
your idea on the 3year thing... 5x more than ... etc.. that's the same as doing your percentages of each year in a ratio like that. 1 : 5 : 25... then:
(1/31) * total Percentage allocated for stats for 2 years ago
5/31 * total percentaget allocated for stats for 1 year ago
25/31 * total % of allocated stats for current year.
That's already in the game, if you want to do that.
So, if it's too small to use for most of the year, realize the hefty chunk you want to be used isn't even being used for a good portion of the year... and when it is used it's not that accurate, because even an entire season is too small of a sample for many things, let alone a portion of one even if it ignore early portions.
will it play the "best" player all the time, no. but the greater amount you use ratings the more likely it is to happen. what do you think those stats are directly derived from in the video game? why wouldn't results be better predictors of success when you have the actual basis of them? they are exactly how it works, regardles of how well some bum was hitting in april - july... think chris shelton or brennan bosch.. fooled by randomness of results of a small sample size. in the case of brennan bosch, it even made his entire season look good (another det guy: creaig monroe, king of half-seasons of luck).
i don't have 100% clarity on how ootp works, but i'd bet the farm on this, lol.