Rasnell's Bugs - AI Favors Ratings over Stats

[battists]

Rasnell's bug from another thread:

Originally Posted by rasnell

By mid- to late-season, or some sort of minimum at bat and IP, the AI is going to have to be tweaked to begin considering actual stats over the ratings. Exceptional players with average ratings should obviously be in the lineup. We're talking .380 hitters after 300 at bats who are #3 in the depth charts because someone out-ranks them in the ratings. Remember when we had this problem as far back as OOTP 3 and Markus gave us the ability to give a weighted average to ratings vs. actual stats. I think we need this option again.

Same problem with on-base vs. power for leadoff hitters. Markus gave us weighted averages that we could set and tweak and this game has the same flaws.

[pstrickert]

I've noticed that the AI will leave a starter in the rotation all season, even when that starter has lost 20 games or more. If the AI looked at stats rather than ratings, this pitcher would be replaced in the rotation (or so you would think!) sometime during the year.

[mrbill]

Maroth!!!

[pstrickert]

Originally Posted by mrbill

Maroth!!!

LOL

I've simmed historical seasons and seen several pitchers with 20+ losses every time (or almost every time). I can understand a 20-loss season every now and then. But multiple players in the same season? Sometimes with as many as 24 or 25 losses?

[mrbill]

Yeah, there's definitely some validity to adding weighting of stats to the evaluation function.

I think its dicey because you know the only thing that goes into determining the outcome of *this play* is the ratings of all the parties involved, at least no one's shown "streaking" to matter (IRL or in OOTP).

But, if its added, at least it'll err on the side of "what a RL manager would do". Of course, if its a customizable option, it can be tweaked to give the right balance of realism and best management.

[tysok]

Originally Posted by mrbill

Yeah, there's definitely some validity to adding weighting of stats to the evaluation function.

I think its dicey because you know the only thing that goes into determining the outcome of *this play* is the ratings of all the parties involved, at least no one's shown "streaking" to matter (IRL or in OOTP).

But, if its added, at least it'll err on the side of "what a RL manager would do". Of course, if its a customizable option, it can be tweaked to give the right balance of realism and best management.

No one may have proved streaking, but there's plenty of proof of a player playing above his ratings, both in game and in in real life. If a guy sucks after 300 ABs he sucks, if a guy is destroying everything after 300 ABs he's a destroyer.
I've had too many instances of a rookie coming into my lineup early in the year and destroying everything thrown at him all year.... then ends up hitting .150 the next season. Kind of annoying sometimes really... but still quite realistic.

[rwd59]

Joe Charbonneau. (Probably spelled wrong)

[battists]

Can anyone provide some more specific instances of the AI clearly favoring a player's ratings, despite stats that show otherwise?

BTW, in terms of feature requests, we are definitely considering the return of a mix between ratings and stats like in OOTP 6.5. So, leaving that aside for a moment, I'd just like to see if there's any glaring examples of this problem that Markus might be willing to tweak right now.

Thanks,

Steve

[tysok]

Originally Posted by battists

Can anyone provide some more specific instances of the AI clearly favoring a player's ratings, despite stats that show otherwise?

BTW, in terms of feature requests, we are definitely considering the return of a mix between ratings and stats like in OOTP 6.5. So, leaving that aside for a moment, I'd just like to see if there's any glaring examples of this problem that Markus might be willing to tweak right now.

Thanks,

Steve

Bah... if only I could find it.
I had a lead off hitter it wouldn't move at all, even to the bench, that ended up hitting .205 for the season. Another solid candidate for the job ended up with 250 ABs for the year and hit .345. This was a while back, at least before the WIP patch, maybe before the first though. The next season when he was sucking I manually moved him to the bench and he was ALWAYS the first pinch hitter to come in.
I know I had a guy hit .230 for the year with 25 HRs that the computer kept hitting in the number 4 slot, number 5 and 6 hitter hit over .300 and over 40 HRs, that was before the WIP patch.

That's not counting contract problems, which a 2 time Cy Young winer and 3 straight 20 win seasons asking for 6 mil a year. That may fall into the low endurance starter problem though.

I'll keep an eye out for this.

[Another Mike D]

Originally Posted by battists

Can anyone provide some more specific instances of the AI clearly favoring a player's ratings, despite stats that show otherwise?

BTW, in terms of feature requests, we are definitely considering the return of a mix between ratings and stats like in OOTP 6.5. So, leaving that aside for a moment, I'd just like to see if there's any glaring examples of this problem that Markus might be willing to tweak right now.

Thanks,

Steve

I'll try a search, but there were some a while back that talked about pitchers with so-so ratings being 19-4 and the likes, and then getting demoted or bumped for guys with higher ratings from the minors.

[OakDragon]

Originally Posted by pstrickert

I've noticed that the AI will leave a starter in the rotation all season, even when that starter has lost 20 games or more. If the AI looked at stats rather than ratings, this pitcher would be replaced in the rotation (or so you would think!) sometime during the year.

My league's record is currently 28 losses in a season. Before that it was 27.

[tysok]

This thread is about some screwy AI taking ratings over stats... from a while back: http://www.ootpdevelopments.com/boar...d.php?t=124874

Right now I don't have anything else without taking a hard look at every team... my team is currently filled with superstars so it doesn't matter what the AI does with my lineup, I still win the world series.

If I see anything I'll post it.

[battists]

Guys, I'm going to close this out for now. The issue of ratings vs stats is one that probably cannot be tackled adequately in a patch, so for now we're going to table this discussion. It's definitely on the list of things to discuss for 2007, though.

Thanks,

Steve