Bugs

[Questdog]

I don't know where we're supposed to report bugs with this new version so I'm sticking it here.

Two things: The League Modifiers for feeder leagues do not work. For sure, HR, BB, and SO are broken, but not certain of others. I also suspect the SH and SF modifier for all the leagues have no effect.

The weights for AI evaluation have no discernible effect on the AI's decisions. I particularly would like to have the ratings have NO weight in the AI's evaluation of a player, with the exception of fielding ratings and base running ratings. I set the rating weight to 1 (in case a "0" is mishandled by the algorithm), the stat ones at 54,30,15. However, the 2-time defending league MVP, a second baseman who hit .367 in each the last two years was not in the opening day lineup ( a career .240 hitter with more homer power took his place, though the MVP had been a doubles machine and solid defensively). For the season, the MVP only got 98 AB's, though he wasn't ever injured and he hit .327 in the few chances he got.

I have truly believed, Markus, since time immemorial that 90% of OOTP's silly transactions could be fixed by having the game evaluate players by their stats rather than ratings. Whatever algorithms are used are clearly not working properly. The computer consistently announces that a player hitting .190 in AA is ready for AAA and the fellow hitting .350 in AAA is possibly over-matched. This is silly, even if correct....

The computer will also give a player with an ERA of 10.00 30 starts.....

The game would be much more realistic if they computer could bench players in slumps and give the kid tearing up AAA a shot. The way it is now, the kid will get his shot when his ratings seem better to the computer than the incumbent's. Never mind whether the kid's hitting .180 in AAA and the incumbent is hitting .325 in the majors....

Or he's the two time defending MVP.....

[RandyTiger]

Bug reports and support: OOTP Developments - Kayako eSupport Help Desk Software

[SteveP]

Originally Posted by Questdog

The weights for AI evaluation have no discernible effect on the AI's decisions. I particularly would like to have the ratings have NO weight in the AI's evaluation of a player, with the exception of fielding ratings and base running ratings.

I am a bit unclear on this one. I see two possibilities:

1. You believe that a guy's future performance in OOTP is best indicated by his recent stats and not by his ratings; or

2. You don't care if the ratings might be a better guide. You just want the AI to use the same method of evaluation that you prefer to use yourself.

Is there a third possibility?

[Questdog]

Originally Posted by SteveP

I am a bit unclear on this one. I see two possibilities:

1. You believe that a guy's future performance in OOTP is best indicated by his recent stats and not by his ratings; or

2. You don't care if the ratings might be a better guide. You just want the AI to use the same method of evaluation that you prefer to use yourself.

Is there a third possibility?

Of course, the ratings are the best guide to future performance, however......

1) I don't think the formula used to evaluate those ratings is sound and

2) even if it was, in real life talent is only discernible through performance

In real life Stephen Strasburg would not have been promoted to the majors if his AAA ERA was 6.00, even though the Nationals would still believe he had immense talent.

My main point is that I want my OOTP league to feel like real life and with the CPU using ratings to evaluate the talent it does not. I argued this several years ago and Markus gave us the option to weight the AI's evaluation, but it appears to have no effect.

All I'm asking for is for the feature to work properly, not to try to convince anyone that my point of view is the proper one.....

[fintach]

I could be mistaken, because I haven't messed with this myself, but isn't there a setting for how the AI evaluates talent, based on ratings/current year/last year/last two years?

[Questdog]

Originally Posted by fintach

I could be mistaken, because I haven't messed with this myself, but isn't there a setting for how the AI evaluates talent, based on ratings/current year/last year/last two years?

Yes....and my point is that it doesn't work.........

[Biggio509]

Originally Posted by SteveP

I am a bit unclear on this one. I see two possibilities:

1. You believe that a guy's future performance in OOTP is best indicated by his recent stats and not by his ratings; or

2. You don't care if the ratings might be a better guide. You just want the AI to use the same method of evaluation that you prefer to use yourself.

Is there a third possibility?

Can't speak for QD but it is frustrating at times to see the AI put a guy in the top 3 spots because he has good ratings but he is barely hitting his weight. Ratings should play in whether a guy is cut, traded, or the team wants to take time to develop him. However, a guy who your scout thinks has 17 CH and 17PH out of 20 who is hitting .220 with 5 HRs should not be in the 3 spot day in and day out and this can happen if ratings are rated too heavily.

The third possibility is that an idea that heavily weighting current stats should have an effect as to whether a guy is benched or at least moved down the lineup if he isn't hitting.

In short let the scouts have a role in determining if the guy just needs playing time or if it is time to give up on him. Let the manager use stats (empirical data) to determine where he hits in the lineup and whether or not he is benched rather than ignore what the guy is doing and just listen to the scout tell him how good the guy is. That includes past stats too so if you rate last year and 2 years ago high enough the guy might stay in the lineup if struggling but has two good seasons. A guy who didn't hit the last 2 years and still isn't hitting should be benched. This does not seem to happen if your scout thinks he has good ratings and playing with stats vs. ratings % in the AI evaluation in my experience hasn't change this much.

[SteveP]

I trust everyone here realizes that these players are digital fictions and that their performance is a function of the ratings. Scouting provide some fog with respect to what those true ratings are. Stats provide a way to cut thru that fog. But, in the end, it's the ratings that produce the results. Each time a guy steps up to bat in OOTP, what he did in the recent past has little impact on what he is going to do in that AB. The ratings have an enormous impact.

[Questdog]

Originally Posted by SteveP

I trust everyone here realizes that these players are digital fictions and that their performance is a function of the ratings. Scouting provide some fog with respect to what those true ratings are. Stats provide a way to cut thru that fog. But, in the end, it's the ratings that produce the results. Each time a guy steps up to bat in OOTP, what he did in the recent past has little impact on what he is going to do in that AB. The ratings have an enormous impact.

Aha! You are correct!...except I believe the stats offer a better summation of those ratings than the AI does with its formulae........

If the AI evaluated those ratings properly, then the player's stats wouldn't stink.........

And anyway, we're supposed to have a feature to allow this, but it does not work.........

[SteveP]

Originally Posted by Questdog

Aha! You are correct!...except I believe the stats offer a better summation of those ratings than the AI does with its formulae........

If the AI evaluated those ratings properly, then the player's stats wouldn't stink.........

And anyway, we're supposed to have a feature to allow this, but it does not work.........

OK. I understand. I'll leave it to you to gather the evidence for Markus. I assume you already know that just saying something doesn't work probably isn't going to result in anything from him.

[Questdog]

Originally Posted by SteveP

OK. I understand. I'll leave it to you to gather the evidence for Markus. I assume you already know that just saying something doesn't work probably isn't going to result in anything from him.

You can see for yourself by changing the weights and then running the computer manager on the league. Lineups will be the same regardless of the settings.....

[Biggio509]

Originally Posted by SteveP

I trust everyone here realizes that these players are digital fictions and that their performance is a function of the ratings. Scouting provide some fog with respect to what those true ratings are. Stats provide a way to cut thru that fog. But, in the end, it's the ratings that produce the results. Each time a guy steps up to bat in OOTP, what he did in the recent past has little impact on what he is going to do in that AB. The ratings have an enormous impact.

Whereas I can't argue with this, the problem is the fog of scouting is not just some fog it is a sometimes a blizzard. The AI should be using stats to cut through the fog as I do when I put a heavier emphasis on stats in AI evaluation. This seems to not be happening which seems to indicate something is not working right with AI evaluation settings. If I set AI evaluation to 0% ratings I get no change in the lineups. There should be some change here because the AI should be evaluating very differently.

I understand ratings drive stats but still if I set the AI to use stats as a basis to determine ratings when scouts may be way off the AI should use them. The problem here is not so much that I disagree with what you are saying it is the settings do not work. Even if I think stats should not be used and accept the game for what it is that is purely driven by a RNG, the game should change behavior when I "program" to change the evaluation algorithm. It does not seem to do this.

[Biggio509]

Here are some lineups with default AI evaluation.

[Biggio509]

60% current stats 30% last year 10% 2 years ago. No difference.

[Biggio509]

last year stats

[Biggio509]

2 yrs. ago stats

[Biggio509]

Note. It is a little strange that Freddy Roberson does not make the cut without a DH. He has little fielding with a DH Roberson actually plays RF where he has a 6 rating. Given his stats any manager would be questioned as to why he isn't an everyday player. Ryan Hughes does not much off an edge in fielding 10 vs. 8 and the ratings are similar. So stats determining lineup should have Roberson over Hughes or Jamison when there is no DH. Jamison has good power ratings but no stats to speak of. Therefore, the AI must still be using ratings to determine the lineup although ratings are set to 0%.

This is just an extreme case to show the point. I wouldn't want 0% ratings but if ratings are still a heavy influence when I supposedly turned them off then there is a problem.

I guess what I am getting at here is that the AI use of stats should be a lie detector for scouting reports. If stats are not working properly in the evaluation then the AI believes what ever the scout tells him and bases decisions on the information and when shown it is likely false. I would like to have a little heavier role of stats so the manager at least gets a good laugh from the scouting report every now and then.

[SteveP]

Originally Posted by Questdog

You can see for yourself by changing the weights and then running the computer manager on the league. Lineups will be the same regardless of the settings.....

Ah ... well, that is true. The AI Player Evaluation has no effect on lineup decisions. Never has. It only affects how the AI manages rosters: trades, promotions/demotions, FA offers, etc. Once a guy is on the Active Roster, how the AI uses him in the lineup and in the depth chart is based on programming that we have no ability to influence.

Or is that what you are referring to?

[Biggio509]

Yep that was it.

[SteveP]

I was incorrect on one point. There is one option in the game that has an influence on how the AI does lineups. That is the option of "traditional" vs "sabermetric". On some AI teams, that option has virtually no impact. On others, it does have a noticeable impact.

To clarify another aspect of this: I agree that the AI does not appear to consider a guy's recent stats in making lineup decisions. So, whatever "sabermetric" means, I have to assume that it does not mean, for example, the guy's recent OBP. I assume the AI is using only ratings in this programming.

The lineup/depth chart is likely a very complex optimization program -- perhaps the most difficult programming in the game. It is clear to me that Markus is very reluctant to mess around with this programming. For example, if he added a weighting for stats (assuming he agreed it was a good idea, first of all), then the game would have to get a lot of testing and bug fixing before it would be useable again. That's a risky thing to do.

I also think he would be even more reluctant to let users mess around with the parameters (weightings) used in this programming. That would just be an invitation for users to break their own games. Which would be OK, but users don't seem to react to that by saying, "oops, I broke my game." They tend to say things that include the title of this thread.