Brain dead AI

[Cory S. Ridgway]

I understand that code intelligent systems is tricky, but the AI for OOTP 4.1 has some issues with even the most basic decisions.

Case in point. I run a small local league with just a few managers (6 active, 2 that do little but keep the computer AI from blowing teams up) and the rest of the league run by the computer. I recently had an opportunity to add a manager, and our rules are that you have to take a bad team. So the new manager's choices were Milwaukee and Tampa Bay. In looking over the roster he thought TB had the better prospects (a stud 1B, and decent C, and three first rate pitching prospects, plus youngster Crawford, Hamilton, and Cantu) with a decent pitching anchored by a strong bullpen. But he had a question - why did the computer have Brandon Puffer (a 2 ERA rating) and Delvin James (a 0) in the majors while trade deadline 2003 acquisition Eddie Oropesa (7) and 2004 free agent addition Scott Williamson (7) didn't pitch a single inning in the majors in 2003, 2004, or the beginning of 2005 for TB?

The AI traded Steve Cox for Oropesa and spent nearly $5M in salary on Oropesa and Williamson in 2004 but didn't use either (neither appeared in any box score in that year - so I don't think they were injured).

I think the majors/minors AI needs work.



Cory Ridgway

[Scott Vibert]

Haven't seen this ever before.. I'd assume the AI was basing the decision on potential over the current ratings more than it should have in this situation.

[Steve Kuffrey]

I really have no idea concerning the situation you mentioned but I just took a quick look at the ratings of 30-something team solo league that I play. I looked specfically at the "era" numerical rating for the pitchers at each level for each team and the highest number era rating for a pitcher in the minors was "6"(I found one). Every team had mostly 0-2's for ratings in that area for their minor league pitchers. In every case the teams had their best or most highly rated pitchers at the major league level. Mind you I only checked the computer teams but everyone was like that.

I don't recall seeing anything in any of the leagues that I have played that would "throw up a red flag" as far as how teams handle their pitchers. Sounds like a small incident and their is probably more to it that we can see here.

[Cory S. Ridgway]

Originally posted by ScottVib
Haven't seen this ever before.. I'd assume the AI was basing the decision on potential over the current ratings more than it should have in this situation.

Perhaps, but unlikely, as Delvin James' potentials had deteriorated to all Fairs and Poors while Williamson's potentials were all quite good. Plus they'd spent $3M+ to sign Williamson in 2004; so they must have thought he was good. There wasn't a single category in which James bettered Williamson and only one in which he matched him (both had C stamina in relief).

Beyond that, Colome was better than Oropesa and Williamson, while Yan might have been though better in some ways (better control, rated as a Closer). And in addition, the team had several other relievers better than James in the minors (some also better than Puffer - but not clearly better).

We use scouts, so I can buy Puffer being over-rated, but this team went out of its way to get Oropesa and Williamson and as far as I've been able to tell, scouts rarely (if ever) underrate players. Could the scout have been so bad as to think a pitcher with no ratings above 2 (James) was better than a pitcher with all ratings of at least 6 (Williamson)?



Cory

[Cory S. Ridgway]

Originally posted by Steve Kuffrey
I really have no idea concerning the situation you mentioned but I just took a quick look at the ratings of 30-something team solo league that I play. I looked specfically at the "era" numerical rating for the pitchers at each level for each team and the highest number era rating for a pitcher in the minors was "6"(I found one). Every team had mostly 0-2's for ratings in that area for their minor league pitchers. In every case the teams had their best or most highly rated pitchers at the major league level. Mind you I only checked the computer teams but everyone was like that.

I don't recall seeing anything in any of the leagues that I have played that would "throw up a red flag" as far as how teams handle their pitchers. Sounds like a small incident and their is probably more to it that we can see here.

Good, I'm glad it is no more than a very rare or isolated incident. It came as a big surprise to me. I'll keep an eye open for any more occurences. I suspect that somehow Oropesa and Williamson became "invisible" in the minor leagues as pitchers sometimes seem to do (you know, those occasions in which minor league pitchers simply don't get used).


Cory

[Steve Kuffrey]

Do you have alot of minor leaguers?

[Cory S. Ridgway]

Originally posted by Steve Kuffrey
Do you have alot of minor leaguers?

That team had 9 pitchers and 7 position players at AAA, 8 and 5 at AA, and 9 and 12 in A ball. They were carrying only 10 pitchers in the majors (Colome, Yan, Phelps, Puffer, and James in the pen).
I've seen teams with a lot more.


Cory

[Steve Kuffrey]

That could be the cause of the problem or atleast part of it. For some reason the game does not work well with teams with too many players. Might try reducing the number a bit and see if that helps.

We will be watching this issue.

[Cory S. Ridgway]

Originally posted by Steve Kuffrey
That could be the cause of the problem or atleast part of it. For some reason the game does not work well with teams with too many players. Might try reducing the number a bit and see if that helps.

We will be watching this issue.

You consider that to be a lot of players!?!? A normal minor league roster will have from 21-24 players. I hope Markus is taking a look at his roster management routines to see if there are problems related to roster size. I'd guess that there's a problem with the roster that shows up more often on large rosters merely because such rosters provide more opportunities for the problem to manifest itself, not because the game simply has problems with such rosters.


Cory

[Steve Kuffrey]

My point was that current "for some reason"(a bug) the game doesn't work well with filled minors or alot of players(I do not know a number but guys who have alot of minor leaguers at times can't see them all).

[Cory S. Ridgway]

Originally posted by Steve Kuffrey
My point was that current "for some reason"(a bug) the game doesn't work well with filled minors or alot of players(I do not know a number but guys who have alot of minor leaguers at times can't see them all).

I've never run into that one myself, and I've had as many as 29 guys at single A.

I think this was a very odd occurence, but it does suggest that the code is not as tight as it could be. Based upon the use of the MFC42 shared object library I'm assuming that this game was developed in Microsoft VC++ and is at least somewhat object oriented in design. In which case I'd expect the roster to be a list of player objects. What the lineups and staves are is another matter. They may well be arrays (as might the roster, of course). In which case a bug related to the size of the roster would be much more likely. The size of a linked list is generally not a source of problems, but arrays are a different story.

Anyway, I think it's a minor bug and wouldn't put it high on the list of priorities.


Cory

[fantom1979]

I have seen it before where the computer picks a era 3 guy over an era 7 guy.. i thought the computer was nuts... now 2 years later.. that 3era guy is a 9era guy and is one of the best pitchers in the game, while the 7era guy is retired.

[Cory S. Ridgway]

Originally posted by fantom1979
I have seen it before where the computer picks a era 3 guy over an era 7 guy.. i thought the computer was nuts... now 2 years later.. that 3era guy is a 9era guy and is one of the best pitchers in the game, while the 7era guy is retired.

I agree that this might be the case; which is why I titled the thread as I did. I can believe that Puffer was being carried as a prospect (James had imploded), but I don't believe the AI should ever leave players of the quality that we are talking about in the minors. Either the computer lost track of those players or it made a bad decision; the first is a bug, the second suggests that the AI needs work.

Specifically, I think it needs to use some actual performance measures in making its decisions instead of just the reports from its scouts. Or perhaps the scout's ratings of major leaguers need to get more accurate as the season progresses. At some point James and his 7.39 ERA should have given way to one of Oropesa or Williamson.

In any case, that was just an example. The AI often makes odd roster decisions having to do with position choices (such as playing a mediocre centerfielder and using Josh Hamilton as a DH instead of moving Jose Cruz Jr from RF to CF).


Cory