6.5a problem with AI starting bad player

Eckstein 4 Prez · 06-26-2005, 11:47 PM

I thought this was supposed to be fixed in 6.5a, but I've still got an AI team that's starting a far, far worse player than one of the other guys they have. If this is a scout issue, it's WAY more severe than any I've seen prior to 6.5.

sporr · 06-27-2005, 12:58 AM

Can you provide the example in this case? Not because I think the AI has a good reason, but to help track down what specific cases may cause the logic to be flawed.

Johnny99 · 06-27-2005, 11:46 AM

http://www.ootpdevelopments.com/boar...d.php?t=101943

He's right.

mtw · 06-27-2005, 01:39 PM

Probably related to this (and I'm speaking off the top of my head since I don't have access to my league right now)...

On the teams it controls in my league, the AI in my league had a few major league caliber players in the minors while pulling up 1 blue star players from A and AA at the same positions.

In one instance, the player was the best OF on the team (a 2.5 star "contribute a lot to the team" RF), and the team had 8 OF, 6 with 1 star, including 2 or 3 1 blue star A or AA players. The RF was a 4-year player.

In another instance, a similar player (2.5 star "contribute a lot to the team" RF) was the third best OF on the team, also carrying 8 OF, 4 with 1 star, and 1 A and 1 AA blue star OF. This player was an 8-year veteran making $6mil per season with 2 years left on his deal.

There were at least two other instances (one a SS and the other an OF), but I don't remember enough to describe the details of the situation.

I don't remeber seeing this before 6.5a.

Eckstein 4 Prez · 06-27-2005, 04:23 PM

Sure, although I don't have my solo league hosted on any websites, so I'll just have to tell you about it.

The Providence Grays are team6 in my league, and have Jimmy Wood, a 33-year-old second baseman who is rated 6 (out of 1-10) in defense with an .868 fielding percentage. His hitting ratings are 56-86-25-17-87. Wood was just signed this season as a free agent, and makes $4.7 million/year.

They also have 32-year-old Dan Patterson, who is rated 3 in defense with an .836 fielding percentage. His hitting ratings are 27-17-16-11-89. Patterson was signed to a minor league contract as a free agent this season, and makes $300,000/year. Both players' talents are pretty close to their actual ratings.

The game realizes that Wood is a better fielder than Patterson, as the AI depth chart has Wood as a defensive replacement. But it considers Patterson the better hitter when making the depth chart.

One note: this error does not extend to making lineups. The AI puts Patterson in the #8 spot against both lefties and righties. However, when Patterson was out injured for a couple of weeks, the game inserted Wood as the #4 hitter.

If there's anything you'd like me to zip up and send anywhere, just let me know.

Markus Heinsohn · 06-28-2005, 03:00 AM

Quote:

Originally Posted by Eckstein 4 Prez

Sure, although I don't have my solo league hosted on any websites, so I'll just have to tell you about it.

The Providence Grays are team6 in my league, and have Jimmy Wood, a 33-year-old second baseman who is rated 6 (out of 1-10) in defense with an .868 fielding percentage. His hitting ratings are 56-86-25-17-87. Wood was just signed this season as a free agent, and makes $4.7 million/year.

They also have 32-year-old Dan Patterson, who is rated 3 in defense with an .836 fielding percentage. His hitting ratings are 27-17-16-11-89. Patterson was signed to a minor league contract as a free agent this season, and makes $300,000/year. Both players' talents are pretty close to their actual ratings.

The game realizes that Wood is a better fielder than Patterson, as the AI depth chart has Wood as a defensive replacement. But it considers Patterson the better hitter when making the depth chart.

One note: this error does not extend to making lineups. The AI puts Patterson in the #8 spot against both lefties and righties. However, when Patterson was out injured for a couple of weeks, the game inserted Wood as the #4 hitter.

If there's anything you'd like me to zip up and send anywhere, just let me know.

Yeah, please zip the league files (*.dat) and send them to markus@ootpdevelopments.com

Thanks!!

Eckstein 4 Prez · 07-08-2005, 11:51 AM

I'm bumping this because I decided to continue going forward with my league in spite of this bug. However, after playing a couple games, I noticed another instance of it: the Milwaukee club is starting catcher Andrew Thompson instead of Rit Harrison. Neither of them is exactly Johnny Bench, but Harrison's ratings are far better than Thompson. Again, the AI knows this most of the time, as they bat Harrison #6 against lefties and righties if I insert him into the depth chart myself, while Thompson hits #8 against both.

Markus already has my league files, so he should be able to see this problem as well. That's two instances in my current league - and I only have six teams in the whole league. Not only that, this is really making me lose confidence in the AI's ability to manage its roster in every way.

f.montoya · 07-08-2005, 12:15 PM

In the meantime, use your imagination: The Manager is benching this big-headed all-star for various reasons like Jose Guillen was by Mike Scioscia last year.

Johnny99 · 07-08-2005, 01:20 PM

Here's the problem in a nutshell. Doing a historical sim. Joe Jackson is in his early 20's and starring for the A's. I check out the lineup and see that he is starting against righties but not lefties. The player starting ahead of him against lefties has far worse ratings in every category and, like Jackson, bats lefty.

So I edit the scrubs ratings to make him even worse. I then hit auto depth. Guess what - the scrub is now starting against both righties and lefties. I then edit the scrub again, this time giving him mediocre ratings, somewhat better than he originally had. Hit auto depth. Jackson is now starting against R and L.

This is with 6.5a, by the way.

jbmagic · 07-08-2005, 01:23 PM

Eckstein 4 Prez

do another follow up email to see where the status is on this.

thanks

Eckstein 4 Prez · 07-10-2005, 03:12 AM

Quote:

Originally Posted by Johnny99

Here's the problem in a nutshell. Doing a historical sim. Joe Jackson is in his early 20's and starring for the A's. I check out the lineup and see that he is starting against righties but not lefties. The player starting ahead of him against lefties has far worse ratings in every category and, like Jackson, bats lefty.

So I edit the scrubs ratings to make him even worse. I then hit auto depth. Guess what - the scrub is now starting against both righties and lefties. I then edit the scrub again, this time giving him mediocre ratings, somewhat better than he originally had. Hit auto depth. Jackson is now starting against R and L.

This is with 6.5a, by the way.

Yep. There definitely seems to be something with the AI that's causing really, really awful players to be considered good. So we've got the bizarre situation where the AI really does seem (to me, at least) to be a lot smarter at handling its depth chart when players aren't too far apart in ratings, but gets the absolute no-brainer decisions wrong.

mtw · 07-11-2005, 12:02 PM

I can say now I've seen a few of these as well. One of the premiere sluggers in my league (58 contact, 102 power on a 100 scale) was on the bench backing up a 34-yo scrub with an 11 contact, 24 power. I caught this after about two weeks, and cut the offending player from the AI team.

In a couple of other instances, I also found that if I increase the players ratings to 25 in each category, the player will be returned to the minors by the AI. 25 appears to be the magic number for me.

Also, I noted that this bug applies to the individual AI RH/LH depth chart, not the player. If a players rating was above 25 (I saw one guy who had a 24 contact vs LH, 37 vs RH), he would only start against the pitcher whose handedness would match him up below 25, and the best player would start against the other handed pitchers.

O2Brew · 07-12-2005, 02:10 AM

My experiences would seem to support mtw here. I've only encountered the issue once, but raising the abilities of the player to 20 something and recalcing the deptch charts fixed the issue. I didn't try and find the exact point at which the player would be removed from the depth charts, but it was definitely right around 25. I don't have the league in front of me right now so I can't look to see just what I had to set his abilities to.

Also, it seemed that his contact rating was the one responsible. The other ratings seemed to have little or no effect on whether he listed on the depth charts.

Eckstein 4 Prez · 07-12-2005, 11:36 AM

Interesting. Of the two cases of this I have in my solo league, one of the players is rated a 26 in contact against both lefties and righties, and the other is rated 12 against lefties, 11 against righties.

The odd thing is that not every player with ratings that low seems subject to the bug. For instance, the team that's starting the 12/11 guy has two other position players rated at below 25 in contact on the major league roster, and does not think it needs to start those guys. Those were the only two other guys on the major league rosters with ratings that low, though, so I guess there's a solid 50-50 shot the AI is going to mistakenly start a guy if his ratings are that bad.

Any update on how/whether work on fixing this is progressing?

mtw · 07-12-2005, 12:03 PM

It's not just the contact rating. At the least, the power rating also plays a role.

I found two more examples of a bad player starting, but these guys had a rating of 26 and 27 in contact. However, the power ratings were 11 and 9 respectively. I bumped these up to 25, recalc'ed the depth charts, and the players went back to 1% usage or off the depth chart entirely.

IMO this gives a some visibility into AI player evaluation.

In my one and only 6.5a season (16 teams, 14 CPU teams), so far I'm up to 7 instances of bad starting players.

I don't know if this matters, but the other thing about all these bad starting players was that they were all at least 30 years old. I need to check out the ratings of all the other players on the AI rosters, but I find this interesting.

Unfortunately, I'm now in the position of having to do a day to day check of all the AI rosters to make sure no funny business is going on.

Eckstein 4 Prez · 07-12-2005, 02:29 PM

The ages of the two awful major leaguers in my league who the AI doesn't start: 28 and 26.

The ages of the two that it does: 32 and 33. And the one with a 26 contact rating has gap-power-eye ratings of 17-13-11, respectively. I think you've figured out what it's doing: guys over 30 who are dreadful in contact and/or power are made into starters by the AI.

Now, the question is: is something in the works that will remedy this?

Johnny99 · 07-12-2005, 03:03 PM

Seems like a quick patch for a pretty big problem could have been released by now, but what the hell do I know.

Eckstein 4 Prez · 07-12-2005, 05:30 PM

Quote:

Originally Posted by Johnny99

Seems like a quick patch for a pretty big problem could have been released by now, but what the hell do I know.

Well, in the past just bumping my bug threads once a week or so has (eventually) worked. So I'm just going to keep on with that.

Johnny99 · 07-12-2005, 05:47 PM

I've just stumbled upon the fact that reducing the AI's reliance on ratings and increasing its reliance on stats seems to at least reduce the problem.

Eckstein 4 Prez · 07-12-2005, 08:54 PM

Quote:

Originally Posted by Johnny99

I've just stumbled upon the fact that reducing the AI's reliance on ratings and increasing its reliance on stats seems to at least reduce the problem.

My AI is already only basing 35% of its decisions on ratings. I'm not willing to go any lower than that.

06-26-2005, 11:47 PM	#1
Eckstein 4 Prez Hall Of Famer Join Date: Dec 2002 Location: The OC Posts: 6,358	6.5a problem with AI starting bad player I thought this was supposed to be fixed in 6.5a, but I've still got an AI team that's starting a far, far worse player than one of the other guys they have. If this is a scout issue, it's WAY more severe than any I've seen prior to 6.5. __________________ Looking for an insomnia cure? Check out my dynasty thread, The Dawn of American Professional Base Ball, 1871.

06-27-2005, 04:23 PM	#5
Eckstein 4 Prez Hall Of Famer Join Date: Dec 2002 Location: The OC Posts: 6,358	Sure, although I don't have my solo league hosted on any websites, so I'll just have to tell you about it. The Providence Grays are team6 in my league, and have Jimmy Wood, a 33-year-old second baseman who is rated 6 (out of 1-10) in defense with an .868 fielding percentage. His hitting ratings are 56-86-25-17-87. Wood was just signed this season as a free agent, and makes $4.7 million/year. They also have 32-year-old Dan Patterson, who is rated 3 in defense with an .836 fielding percentage. His hitting ratings are 27-17-16-11-89. Patterson was signed to a minor league contract as a free agent this season, and makes $300,000/year. Both players' talents are pretty close to their actual ratings. The game realizes that Wood is a better fielder than Patterson, as the AI depth chart has Wood as a defensive replacement. But it considers Patterson the better hitter when making the depth chart. One note: this error does not extend to making lineups. The AI puts Patterson in the #8 spot against both lefties and righties. However, when Patterson was out injured for a couple of weeks, the game inserted Wood as the #4 hitter. If there's anything you'd like me to zip up and send anywhere, just let me know. __________________ Looking for an insomnia cure? Check out my dynasty thread, The Dawn of American Professional Base Ball, 1871.

07-08-2005, 11:51 AM	#7
Eckstein 4 Prez Hall Of Famer Join Date: Dec 2002 Location: The OC Posts: 6,358	I'm bumping this because I decided to continue going forward with my league in spite of this bug. However, after playing a couple games, I noticed another instance of it: the Milwaukee club is starting catcher Andrew Thompson instead of Rit Harrison. Neither of them is exactly Johnny Bench, but Harrison's ratings are far better than Thompson. Again, the AI knows this most of the time, as they bat Harrison #6 against lefties and righties if I insert him into the depth chart myself, while Thompson hits #8 against both. Markus already has my league files, so he should be able to see this problem as well. That's two instances in my current league - and I only have six teams in the whole league. Not only that, this is really making me lose confidence in the AI's ability to manage its roster in every way. __________________ Looking for an insomnia cure? Check out my dynasty thread, The Dawn of American Professional Base Ball, 1871.

07-11-2005, 12:02 PM	#12
mtw All Star Reserve Join Date: Dec 2001 Posts: 653	I can say now I've seen a few of these as well. One of the premiere sluggers in my league (58 contact, 102 power on a 100 scale) was on the bench backing up a 34-yo scrub with an 11 contact, 24 power. I caught this after about two weeks, and cut the offending player from the AI team. In a couple of other instances, I also found that if I increase the players ratings to 25 in each category, the player will be returned to the minors by the AI. 25 appears to be the magic number for me. Also, I noted that this bug applies to the individual AI RH/LH depth chart, not the player. If a players rating was above 25 (I saw one guy who had a 24 contact vs LH, 37 vs RH), he would only start against the pitcher whose handedness would match him up below 25, and the best player would start against the other handed pitchers. __________________ Over-Zealous Apologist

07-12-2005, 11:36 AM	#14
Eckstein 4 Prez Hall Of Famer Join Date: Dec 2002 Location: The OC Posts: 6,358	Interesting. Of the two cases of this I have in my solo league, one of the players is rated a 26 in contact against both lefties and righties, and the other is rated 12 against lefties, 11 against righties. The odd thing is that not every player with ratings that low seems subject to the bug. For instance, the team that's starting the 12/11 guy has two other position players rated at below 25 in contact on the major league roster, and does not think it needs to start those guys. Those were the only two other guys on the major league rosters with ratings that low, though, so I guess there's a solid 50-50 shot the AI is going to mistakenly start a guy if his ratings are that bad. Any update on how/whether work on fixing this is progressing? __________________ Looking for an insomnia cure? Check out my dynasty thread, The Dawn of American Professional Base Ball, 1871.

06-27-2005, 12:58 AM	#2
sporr Global Moderator Join Date: Dec 2001 Location: Muscatine, IA Posts: 8,277	Can you provide the example in this case? Not because I think the AI has a good reason, but to help track down what specific cases may cause the logic to be flawed.

06-27-2005, 11:46 AM	#3
Johnny99 Minors (Triple A) Join Date: Sep 2004 Posts: 209	http://www.ootpdevelopments.com/boar...d.php?t=101943 He's right.

06-27-2005, 01:39 PM	#4
mtw All Star Reserve Join Date: Dec 2001 Posts: 653	Probably related to this (and I'm speaking off the top of my head since I don't have access to my league right now)... On the teams it controls in my league, the AI in my league had a few major league caliber players in the minors while pulling up 1 blue star players from A and AA at the same positions. In one instance, the player was the best OF on the team (a 2.5 star "contribute a lot to the team" RF), and the team had 8 OF, 6 with 1 star, including 2 or 3 1 blue star A or AA players. The RF was a 4-year player. In another instance, a similar player (2.5 star "contribute a lot to the team" RF) was the third best OF on the team, also carrying 8 OF, 4 with 1 star, and 1 A and 1 AA blue star OF. This player was an 8-year veteran making $6mil per season with 2 years left on his deal. There were at least two other instances (one a SS and the other an OF), but I don't remember enough to describe the details of the situation. I don't remeber seeing this before 6.5a. __________________ Over-Zealous Apologist

07-08-2005, 12:15 PM	#8
f.montoya Hall Of Famer Join Date: Nov 2004 Posts: 6,077	In the meantime, use your imagination: The Manager is benching this big-headed all-star for various reasons like Jose Guillen was by Mike Scioscia last year. __________________ Fidel Montoya Asahi2 Baseball League ex-Commissioner(Historical League Since 2004) Ex-Web Host Current Mod Maker??

07-08-2005, 01:20 PM	#9
Johnny99 Minors (Triple A) Join Date: Sep 2004 Posts: 209	Here's the problem in a nutshell. Doing a historical sim. Joe Jackson is in his early 20's and starring for the A's. I check out the lineup and see that he is starting against righties but not lefties. The player starting ahead of him against lefties has far worse ratings in every category and, like Jackson, bats lefty. So I edit the scrubs ratings to make him even worse. I then hit auto depth. Guess what - the scrub is now starting against both righties and lefties. I then edit the scrub again, this time giving him mediocre ratings, somewhat better than he originally had. Hit auto depth. Jackson is now starting against R and L. This is with 6.5a, by the way.

07-08-2005, 01:23 PM	#10
jbmagic Hall Of Famer Join Date: Sep 2003 Location: Los Angeles Posts: 3,417	Eckstein 4 Prez do another follow up email to see where the status is on this. thanks

07-12-2005, 02:10 AM	#13
O2Brew Bat Boy Join Date: Dec 2004 Posts: 16	My experiences would seem to support mtw here. I've only encountered the issue once, but raising the abilities of the player to 20 something and recalcing the deptch charts fixed the issue. I didn't try and find the exact point at which the player would be removed from the depth charts, but it was definitely right around 25. I don't have the league in front of me right now so I can't look to see just what I had to set his abilities to. Also, it seemed that his contact rating was the one responsible. The other ratings seemed to have little or no effect on whether he listed on the depth charts.

07-12-2005, 12:03 PM	#15
mtw All Star Reserve Join Date: Dec 2001 Posts: 653	It's not just the contact rating. At the least, the power rating also plays a role. I found two more examples of a bad player starting, but these guys had a rating of 26 and 27 in contact. However, the power ratings were 11 and 9 respectively. I bumped these up to 25, recalc'ed the depth charts, and the players went back to 1% usage or off the depth chart entirely. IMO this gives a some visibility into AI player evaluation. In my one and only 6.5a season (16 teams, 14 CPU teams), so far I'm up to 7 instances of bad starting players. I don't know if this matters, but the other thing about all these bad starting players was that they were all at least 30 years old. I need to check out the ratings of all the other players on the AI rosters, but I find this interesting. Unfortunately, I'm now in the position of having to do a day to day check of all the AI rosters to make sure no funny business is going on. __________________ Over-Zealous Apologist

07-12-2005, 02:29 PM	#16
Eckstein 4 Prez Hall Of Famer Join Date: Dec 2002 Location: The OC Posts: 6,358	The ages of the two awful major leaguers in my league who the AI doesn't start: 28 and 26. The ages of the two that it does: 32 and 33. And the one with a 26 contact rating has gap-power-eye ratings of 17-13-11, respectively. I think you've figured out what it's doing: guys over 30 who are dreadful in contact and/or power are made into starters by the AI. Now, the question is: is something in the works that will remedy this? __________________ Looking for an insomnia cure? Check out my dynasty thread, The Dawn of American Professional Base Ball, 1871.

07-12-2005, 03:03 PM	#17
Johnny99 Minors (Triple A) Join Date: Sep 2004 Posts: 209	Seems like a quick patch for a pretty big problem could have been released by now, but what the hell do I know.

07-12-2005, 05:47 PM	#19
Johnny99 Minors (Triple A) Join Date: Sep 2004 Posts: 209	I've just stumbled upon the fact that reducing the AI's reliance on ratings and increasing its reliance on stats seems to at least reduce the problem.