Bugs

[SteveP]

And just to finish this up:

When you have scouts turned on, the AI is also subject to some method of "fog" which affect player evaluation and roster decisions. It is not clear to me, however, that this fog is active when the AI runs its lineup optimization. It's very possible that the AI is using actual ratings.

[Biggio509]

I think sabremetric lineups will give weight to the playing the splits. So the AI is more likely to platoon when this is on. I really have no idea either what the AI uses to set up the lineup but current stats isn't it. If it is using true ratings that is probably better. I suppose there can still be aberrations with true ratings be used. I had wrongly assumed AI evaluation had something to do with lineups but as you have pointed out it doesn't seem like it ever was intended to do that.

I must say things have improved though especially in the minors. In old versions it would play older guys who reached their potential over young promising guys in the minors lineups. At least for the lower minors now there is a preference to the up and coming over the has been there and will stay there. I am not sure if to some extent the Arizona screen shots I gave were showing this might be happening in the majors too. I accidentally deleted the league because I thought some files were corrupt and did not back up before a clean install. So I can't see their focus but the lineups might make more sense if they were in rebuilding mode.

To me this is not make or break, a sigh of disappointment at most not the I do not want to play this piece of trash I get with some games. It is minor but still a bit of a eye brow raiser sometimes seeing a guy who isn't hitting in the 3 hole or a catcher batting first with speed and stealing of maybe 10 combined. It isn't totally unrealistic. The Houston Astros this year are paying Lee and Berkman way too much to bench them and there are still hitting in the middle of the lineup because there isn't anyone better. So I can't say what I have seen in OOTP can't happen.

Maybe watching Scrap Iron and Cooper submit players in a different order each day for 5 years, until Mills took over this season made me forget how most managers set lineups.

[Questdog]

Originally Posted by SteveP

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?

Hmmmm.....

Then you have the silly situation where the computer signs the aforementioned 2-time defending MVP to a mega-contract to be a pinch hitter..........

And promotions are made (with ratings weight set to 0) of players with lousy stats. And players tearing the league to shreds get demoted. Why is this then?

[SteveP]

Originally Posted by Questdog

Hmmmm.....

Then you have the silly situation where the computer signs the aforementioned 2-time defending MVP to a mega-contract to be a pinch hitter..........

And promotions are made (with ratings weight set to 0) of players with lousy stats. And players tearing the league to shreds get demoted. Why is this then?

OK, so I might be confused by this thread. In any event, the AI signs, promotes and demotes players based on the AI Player Evaluation. Once they are on the Active Roster, the AI decides on lineups and depth charts based on a different programming to optimize the use of those players. That's the way the game works. Surprisingly enough, RL GMs and Managers do things that fans and sports journalists consider incomprehensible as well. It's tricky to figure out what is a bug and what is a good imitation of human behavior.

[Questdog]

Originally Posted by SteveP

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?

OK....this just happened in a league I am playing and I am incensed, though I think it proves that something is messed up with the AI weighting system.

A player was leading the league in homers at the All-Star beak, on a pace to break the league's all-time record. He was batting .294 in a league that bats .258 overall. And the computer sent him to AAA!!! I have the weights set to 50,33,16 for this year, last year....and 1 for ratings. He is 25 years old. His replacement from AAA was batting .247 in the minors with 8 homers. Since the call up the replacement has batted .185 with a .247 SLG in 80+ AB's while the homer champ is riding the pine in AAA, though he has a 1350 OPS in his pinch hitting opportunities. He still leads the majors in home runs, though he's missed a whole month......

Explain to me how this is possible?

[SteveP]

Originally Posted by Questdog

though he's missed a whole month......

I think Markus made an attempt in response to your ticket to explain why setting ratings to zero (or virtually zero) can produce some odd results, especially before a guy has a lot of ABs (or IPs). It's possible that you are seeing an example of that. Don't know. But maybe since he missed a month, his current year stats aren't large enough to carry much weight.

[Questdog]

Originally Posted by SteveP

I think Markus made an attempt in response to your ticket to explain why setting ratings to zero (or virtually zero) can produce some odd results, especially before a guy has a lot of ABs (or IPs). It's possible that you are seeing an example of that. Don't know. But maybe since he missed a month, his current year stats aren't large enough to carry much weight.

1) Where is this response of Markus

2) The player only missed the month because he got sent to AAA....it was July 8th when he was demoted...plenty of AB's....He was LEADING the league in homers, so even if his AB's were low, the weight should be pretty heavy.....

3) I've tried many variations on the settings.....100% last year until June, then 100% this year.....nothing seems to work......also 100% ratings doesn't jibe either....I tried a league for test and studied the actual ratings of players demoted/promoted and many times the player demoted had better ratings than the player promoted....but in most of those cases (maybe all) the player demoted was a youngster and the player promoted was a 30-35 year old vet who sucked....some kind of weight seems to be given to the players age, making it harder for youngsters to break into the bigs.....especially pitchers who tend to sit in the minors until they are 27-28

[SteveP]

Originally Posted by Questdog

1) Where is this response of Markus

Sorry. Someone posted a ticket in the support system which quoted the text from your OP here. I assumed it was you. Markus responded to that ticket. This is what he said:

The evaluatiuon AI works properly. Playing time is a factor, and missing playing time will be made up by using the ratings, no matter what the settings are. The AI will not base decisions on players having just 100 AB entirely on stats, for example.

Cheers,
Markus

[Questdog]

Originally Posted by SteveP

Sorry. Someone posted a ticket in the support system which quoted the text from your OP here. I assumed it was you. Markus responded to that ticket. This is what he said:

Ah, that WAS me....I need to check my email more often.....

[SteveP]

Originally Posted by Questdog

Ah, that WAS me....I need to check my email more often.....

Not a problem, I'm sure. Also, I have to admit that the answer Markus gave you is a bit ambiguous, since it implies that the AI will use ratings regardless of how you set the weightings. I don't think that's true, but I've certainly been wrong before.

[Questdog]

I just ran a test where I set the weights to 100% last year. Now, by my thinking, this should mean that no promotions or demotions should happen at all during the season, since the weights should be static....

However, just as many promotions/demotions happen as ever.....

[SteveP]

Originally Posted by Questdog

I just ran a test where I set the weights to 100% last year. Now, by my thinking, this should mean that no promotions or demotions should happen at all during the season, since the weights should be static....

However, just as many promotions/demotions happen as ever.....

All I can do is speculate, of course, but I would not make the same assumption. First, we know that the AI periodically (and possibly at regular intervals -- once a week perhaps??) evaluates rosters. We also know from lots of different things that the AI does, that it may look at the same situation and make two different decisions. The AI does not remember a decision it made before, and has no built in way to ensure consistency. We don't know to what extent the AI may have programming to decide whether a roster, overall, is satisfactory. It may make moves to satisfy that programming, and it may make different moves when it does this same evaluation at two different times, and those moves may be hard to explain when looking at what happens to just one player.

Along with that, I think it may be possible that the AI Evaluation Weightings only influence the AI -- that they may not be definitive. For example, in the case of your experiment, a rookie who has no prior year stats has to be evaluated in some reasonable fashion. If that is true, then Markus meant what he appeared to say: that the AI will consider ratings regardless.

[Questdog]

You would at least think that there would be fewer moves with my experiment, since whatever the computer is evaluating is static (supposedly). However the list of rosters moves from one season with this setting and without are of similar lengths........

The worst problem to me is that I can't make sense out of the moves it makes. Stats obviously have little or no bearing on the matter regardless of the settings. However, the ratings are not the only motivator in these moves either. The original example I gave of the HR leader being demoted is a case in point. His ratings were better in every area over the player who was promoted to take his job, except for 'Eye' rating in which the replacement player had a very large advantage. To me, I like the guy who can hit better than the guy who can walk, but it could be argued that the replacements ratings looked better overall than the HR king. However, the crux of the matter is that when the HR king went to AAA, he had to sit the bench because the starter there had MUCH better ratings than him in every area except 'Power', and his edge there was not tremendous. So, based on only ratings the AAA starter should have been the one promoted to the majors, not the bum who was. This guy was hitting .350 or so in AAA compared to .247 for the guy who go got the nod, so it couldn't have been the stats component that got the other guy the job. What then was it? Nothing that I can see, except the fact that the computer seems to like to keep players in AAA until they are 28..............

[SteveP]

Originally Posted by Questdog

The worst problem to me is that I can't make sense out of the moves it makes.

I think that sums it up, don't you? So here the way I look at it:

1. I'm fan of a couple of real life teams, and they do things that don't make sense to me. Even Billy Beane has done some very stupid trades.

2. The AI is stupid, and will always be stupid, because the AI in standard commercial games is stupid. This is not something that people who watch Sci-fi movies readily accept, but it's true.

3. The moves may not make sense when looked at one way, but may sort of make sense when looked at from a different direction (say, fielding)

4. Sometimes the AI makes moves that don't make sense to me, but work out pretty well in the end (not always, of course, but it doesn't make sense to focus on the ones that don't work out well.)

5. No one (including Markus) can explain a specific transaction. That is inherent in putting an AI in charge. You program an AI, and do a lot of wide field testing to see that it's acting in a way that isn't too extremely far off human behavior, and you put it on sale. If he had to create an AI that didn't make mistakes, he would find another line of work.

[Questdog]

Originally Posted by SteveP

Even Billy Beane has done some very stupid trades.

Billy Beane is a moron.....

However, this goes beyond moves that are questionable (like Dusty Baker's lineups) to things that just would NEVER happen in real life, like the 25 year old HR king being sent to AAA so a 32 year old never-was can hit .180 in his place. Or a player hitting .210 after two months in A ball getting promoted to AA, where he hits .350 for two months and then is demoted back to A ball.....

You talk like this is some sort of rocket science or better yet, some sort of unearthly magic that happens in the ether of space....

I LOVE this game and I am grateful that Markus has kept his commitment to improve it ever since I began playing it when I got my little blue CD in the mail some 10+ years ago.....

However, I have never understood why he has found it so difficult to make the computer GM's behave in a believable manner. It seems so frustratingly simple to me....

[SteveP]

Originally Posted by Questdog

Billy Beane is a moron.....

OK. I have no reason to get between you and Markus. May the best man win ...

[Questdog]

Originally Posted by SteveP

OK. I have no reason to get between you and Markus. May the best man win ...

I don't quite understand....is Billy Beane one of the gods Markus worships are something?......I don't want to ridicule a man's religion.....

[SteveP]

Originally Posted by Questdog

I don't quite understand....is Billy Beane one of the gods Markus worships are something?......I don't want to ridicule a man's religion.....

Not that I know of. I just thought the fact that you focused on that phrase (which was sort of off-the-cuff and semi-humorous) meant that you mostly wanted to have a debate with Markus. So I decided to bow out of the picture. Don't think I have anything more to contribute, anyway. Cheers.

[Questdog]

Originally Posted by SteveP

Not that I know of. I just thought the fact that you focused on that phrase (which was sort of off-the-cuff and semi-humorous) meant that you mostly wanted to have a debate with Markus. So I decided to bow out of the picture. Don't think I have anything more to contribute, anyway. Cheers.

I'm sorry, my comment was also meant to be in the same vein.....Billy Beane is irrelevant to the issue......

I don't want to debate, just help make the game better than it is....If every one is satisfied with the status quo, however, then I'll just have to keep dreaming of the day.....

By the way, I am anal retentive and obsessive compulsive, but none the less a very easy going guy who speaks with his tongue in his cheek most of the time and not offended by much and definitely don't mean to offend others.....

[SteveP]

OK, but for the sake of your own peace of mind, let me suggest that even Markus would not be able to tell you why the AI did what it did with respect to a particular player. To do that, he would have to have the game files in hand exactly before the key transaction took place. Even then, he'd be speculating about what happened, unless he took an inordinate amount of time to run tests and trace what was happening in the programming each time the AI made its decision.

OTOH, he could just say you shouldn't have set ratings weight to 1. I don't know. Either way, you aren't going to have accomplished any useful result. Markus pays attention to well-documented patterns that include lots of examples which are unambiguous. Even then, he can be a bit stubborn sometimes () but much of the time finally figures out where the bug is, if the pattern clearly indicates that there is a bug.

And wouldn't Markus be in software development heaven if everyone accepted the status quo ...