In-game Subs not based on Depth Chart

[jar2574]

Has anyone else noticed that in-game substitutions, (as opposed to ones made automatically before the game for fatigued players), are not determined by the depth chart? Instead, the in-game substitutions reflect what the manager would do if he set the depth chart. If your depth chart reflects the manager's choices, you're in luck. If you slot a different guy as the first back-up, you're not.

[SteveP]

Originally Posted by jar2574

Has anyone else noticed that in-game substitutions, (as opposed to ones made automatically before the game for fatigued players), are not determined by the depth chart?

Yes -- or, at least, what you do in the depth chart is no guarantee that the AI will adhere to it for substitutions during a game. To some extent the same thing happens with the decisions about relief pitchers. I've been puzzled for some time about why the game works this way, but it does.

It's a legitimate issue to raise, and see if we can find out from Markus (when he is back in action) why it works this way.

[RchW]

Yeah, this is a major frustration. Unless you play out each game the AI (or manager) will use subs the way he (it) wants to. It really affects relief pitching as the AI will continue to use the same relievers over and over, even if they are bad, while it will not use others having better years, unless the game is lost. I don't doubt that this could cost 3-8 wins per season, depending on the circumstances.

We should be able to force the issue by designating first out of bullpen (if not tired) second and so on.

[Pirate Fan]

One thing I do is check the box "never pinch hit for this player" for all of my position players. At least I don't end up with my backup catcher playing 2B with this approach.

[jar2574]

Originally Posted by Pirate Fan

One thing I do is check the box "never pinch hit for this player" for all of my position players. At least I don't end up with my backup catcher playing 2B with this approach.

I think I'm going to have to do that, and for pinch running as well. The AI keeps pinch running for the league's 2nd best hitter in the 7th inning, and then he's not in to bat in the 9th. The sad part is he has 12 speed, and 18 or 19 ability in steals and baserunning. It's not like he's lead footed.

[SteveP]

Originally Posted by jar2574

I think I'm going to have to do that, and for pinch running as well. The AI keeps pinch running for the league's 2nd best hitter in the 7th inning, and then he's not in to bat in the 9th. The sad part is he has 12 speed, and 18 or 19 ability in steals and baserunning. It's not like he's lead footed.

PRing is a lingering problem area as well. The only "fix" for this is an adjustment to the PR slider in the Team Strategy page. Moving toward "Never" reduces the tendency of the AI use PRs at all, which helps to reduce the likelihood of PRing for someone who has decent speed. Setting it to "Never" fixes the problem entirely (at least, I think "Never" mean never in OOTP10).

[Questdog]

Originally Posted by jar2574

I think I'm going to have to do that, and for pinch running as well. The AI keeps pinch running for the league's 2nd best hitter in the 7th inning, and then he's not in to bat in the 9th. The sad part is he has 12 speed, and 18 or 19 ability in steals and baserunning. It's not like he's lead footed.

You can also adjust your "strategy" sliders to good effect....

By the way, when I sim games, the manager seems to use my bullpen based on where I have them placed on the depth chart....

[Pirate Fan]

Originally Posted by Questdog

You can also adjust your "strategy" sliders to good effect....

By the way, when I sim games, the manager seems to use my bullpen based on where I have them placed on the depth chart....

Interesting. My experience has been more or less like that described by Rchw above. The manager will pick out one guy he really likes and that reliever will be the first out of the pen as long as he is not exhausted. The only exception is the pitcher I have in the closer role, who is generally used as RL closers are.

In the on-line league in which I play, I even traded away the reliever the manager liked since he wasn't doing very well. It was the only way I could get the manager to stop using him; no matter where I put him on the pitching chart he'd be in there as soon as the starter was out.

[satchel]

Originally Posted by SteveP

Setting it to "Never" fixes the problem entirely...

But causes another one that's just as bad.

AI pinch-running management was supposed to have been addressed in v10, and I'd assumed that it was. The situation described by jar2574 lacks some key details, such as his own pinch-running strategy settings, and other aspects of the game situation.

Originally Posted by Questdog

By the way, when I sim games, the manager seems to use my bullpen based on where I have them placed on the depth chart....

My bullpen settings definitely affect my AI manager's use of my pitchers. It's just that the game doesn't look at the bullpen setup and follow its structure absolutely, all of the time; the game factors in other things as well, such as game situation, the pitchers' fatigue, handedness, etc. So, my "Middle reliever" in the number two slot might have more appearances, over a whole season, than the man in the number one slot (same with my "Setup relievers"). But that's because of game circumstances. And the man in my number four "Middle reliever" slot will almost always be the manager's last choice out of the pen. So the AI manager's choice of pitchers will generally follow my role assignments, but the game doesn't slavishly go down my list from top to bottom, disregarding all other factors.

I do wish that we had the option of assigning bullpen roles in a more rigid way, such we can for starters by setting the rotation on "Strict order" (which will always follow my designated order, unless a pitcher is fatigued, in which case it will skip to the next slot).

[SteveP]

Originally Posted by satchel

AI pinch-running management was supposed to have been addressed in v10, and I'd assumed that it was. The situation described by jar2574 lacks some key details, such as his own pinch-running strategy settings, and other aspects of the game situation.

AFAIK, no one ever committed to that. What we did get were significant improvement in the frequency of PHing for position players. We also got some improved AI intelligence to avoid PHing/PRing for position players if there was no one suitable on the bench to take the fielding assignment ("suitable" still being somewhat in the eye of the beholder )

It's no good to complain that the AI doesn't make decisions the way I would -- on PRing or anything else. Markus will address something if he gets substantive evidence (including stats if possible) that show the AI doesn't do a good job of imitating RL managers. Making that case for PRing is very difficult -- especially if you take frequency out of the equation, and just focus on the relative baserunning skills for the player being take out and the player be put in. So this subject just couldn't make it up the priority list for OOTP10.

[RchW]

Originally Posted by satchel

But causes another one that's just as bad.

AI pinch-running management was supposed to have been addressed in v10, and I'd assumed that it was. The situation described by jar2574 lacks some key details, such as his own pinch-running strategy settings, and other aspects of the game situation.



My bullpen settings definitely affect my AI manager's use of my pitchers. It's just that the game doesn't look at the bullpen setup and follow its structure absolutely, all of the time; the game factors in other things as well, such as game situation, the pitchers' fatigue, handedness, etc. So, my "Middle reliever" in the number two slot might have more appearances, over a whole season, than the man in the number one slot (same with my "Setup relievers"). But that's because of game circumstances. And the man in my number four "Middle reliever" slot will almost always be the manager's last choice out of the pen. So the AI manager's choice of pitchers will generally follow my role assignments, but the game doesn't slavishly go down my list from top to bottom, disregarding all other factors.

I do wish that we had the option of assigning bullpen roles in a more rigid way, such we can for starters by setting the rotation on "Strict order" (which will always follow my designated order, unless a pitcher is fatigued, in which case it will skip to the next slot).

My experience is very different. Most frustrating is the AI usually disagrees with my choice of set up pitchers and uses them first out of the pen and uses the other guys as set up men. My AI eval is set on stats yet the AI insists on using very bad relievers who may have one high rating out of 3.

The one choice that seems to work as expected is the closer. This was not the case in v9.

In a recent season I tried everything to get a particular pitcher into games including putting him in every slot. He ended up in only 13 games, with a WHIP of 1.13. Another pitcher who performed much worse (WHIP of 1.67 ERA of 6+) appeared in 38 of the first 80 games. I finally DFA'd him to stem the bleeding.

We definitely need more control over bullpen usage.

[johnnyshaka]

Bumping this thread as this is the first one I found relating to the issue I'm concerned with...bad AI sub selections.

Here are a couple of dubious sub examples:

MLB Box Scores, Kansas City Royals at Oakland Athletics, 10-04-2010

For some reason Nady was subbed in at 2B to replace a PH but then was immediately replaced before taking the field. Wasted him for nothing.

MLB Box Scores, Kansas City Royals at Oakland Athletics, 10-02-2010

Everidge was subbed in at 3B to start the 8th after Cust pinch hit for the starting 3B in the 7th. Everidge is rated 1 at 3B and wouldn't have been listed anywhere other than 1B and DH. Pennington and Cardenas would've been listed as backups in the depth chart and at the time, neither one of them had been in the game.

MLB Box Scores, Texas Rangers at Oakland Athletics, 10-01-2010

Again, Nady played 2B after pinch hitting while Pennington and Valdez were sitting on the bench...two better options, defensively.

MLB Box Scores, Toronto Blue Jays at Oakland Athletics, 09-20-2010

Here's a beauty...a 1B and OF who don't have ratings at either 2B or SS end up playing there late in the game. Cardenas and Valdez, both more than qualified to play either of those spots, were on the bench.

I could go on and on...

My depth chart for 2B, 3B, and SS includes the starter and likely a combo of Pennington (solid defensively at all three spots), Cardenas (decent or better at all three), and Valdez (decent or better at all three).

Do I need to really adjust the sliders one way or another for all the strategies regarding "offense vs. defense"? I don't want to resort to not allowing players to be pinch hit for because there are times when it's warranted but if that keeps the AI from making bonehead subs then I just might have to pick my poison.

[SteveP]

Originally Posted by johnnyshaka

Bumping this thread as this is the first one I found relating to the issue I'm concerned with...bad AI sub selections.

Here are a couple of dubious sub examples:

MLB Box Scores, Kansas City Royals at Oakland Athletics, 10-04-2010

For some reason Nady was subbed in at 2B to replace a PH but then was immediately replaced before taking the field. Wasted him for nothing.

MLB Box Scores, Kansas City Royals at Oakland Athletics, 10-02-2010

Everidge was subbed in at 3B to start the 8th after Cust pinch hit for the starting 3B in the 7th. Everidge is rated 1 at 3B and wouldn't have been listed anywhere other than 1B and DH. Pennington and Cardenas would've been listed as backups in the depth chart and at the time, neither one of them had been in the game.

MLB Box Scores, Texas Rangers at Oakland Athletics, 10-01-2010

Again, Nady played 2B after pinch hitting while Pennington and Valdez were sitting on the bench...two better options, defensively.

MLB Box Scores, Toronto Blue Jays at Oakland Athletics, 09-20-2010

Here's a beauty...a 1B and OF who don't have ratings at either 2B or SS end up playing there late in the game. Cardenas and Valdez, both more than qualified to play either of those spots, were on the bench.

I could go on and on...

My depth chart for 2B, 3B, and SS includes the starter and likely a combo of Pennington (solid defensively at all three spots), Cardenas (decent or better at all three), and Valdez (decent or better at all three).

Do I need to really adjust the sliders one way or another for all the strategies regarding "offense vs. defense"? I don't want to resort to not allowing players to be pinch hit for because there are times when it's warranted but if that keeps the AI from making bonehead subs then I just might have to pick my poison.

Where to begin ...

Everyone should have PHing for position players set to "Very rarely". Markus made the frequency of PHing for position players better in OOTP10, but it is still too high. Don't know why ... but it may have to do with OOTPer's thinking that PHing happens much more often than it does IRL. There are various measures of what's realistic, but the easiest it this: PHing for more than one position player in a single inning is very, very rare IRL. In historical leagues, where there were more position players on the bench, the frequency is greater: ranging from infrequently to rare. If you see teams PHing more than they should, then you can adjust that team strategy slider accordingly.

This is important, because the tendency of the AI to make mistakes on substitutions after PHing is much, much greater when PHing is frequent. The first example you give is a bug. It has been reported to Markus. He has never seen it, so it is not on the schedule to be fixed.

The others are just examples of AI substitution quirks. In OOTP, the game is designed so that anyone can play any position. This is unrealistic, but it is expedient (because of the inability of the AI to plan ahead, which means that the AI can easily paint itself into a corner). The problem is that the AI has a tendency to take advantage of that fact, even when it isn't necessary. If, however, you have PHing at a realistic level, these unrealistic fielding assignments become very infrequent.

[johnnyshaka]

Originally Posted by SteveP

Where to begin ...

Everyone should have PHing for position players set to "Very rarely". Markus made the frequency of PHing for position players better in OOTP10, but it is still too high. Don't know why ... but it may have to do with OOTPer's thinking that PHing happens much more often than it does IRL. There are various measures of what's realistic, but the easiest it this: PHing for more than one position player in a single inning is very, very rare IRL. In historical leagues, where there were more position players on the bench, the frequency is greater: ranging from infrequently to rare. If you see teams PHing more than they should, then you can adjust that team strategy slider accordingly.

This is important, because the tendency of the AI to make mistakes on substitutions after PHing is much, much greater when PHing is frequent. The first example you give is a bug. It has been reported to Markus. He has never seen it, so it is not on the schedule to be fixed.

The others are just examples of AI substitution quirks. In OOTP, the game is designed so that anyone can play any position. This is unrealistic, but it is expedient (because of the inability of the AI to plan ahead, which means that the AI can easily paint itself into a corner). The problem is that the AI has a tendency to take advantage of that fact, even when it isn't necessary. If, however, you have PHing at a realistic level, these unrealistic fielding assignments become very infrequent.

Thanks for the reply, SteveP, very much appreciated.

Agreed, PHing is rather high if you don't dial it down manually...so that's an easy fix/workaround. Done.

Regarding your final point about the AI sub quirks, what about another option in the player strategy window that will let you specify, if you want to, the minimum position rating a player needs to have before being considered as a defensive sub at any position(s)? Maybe you can select each position manually that a player may be subbed in at. Or else, have the AI use the depth chart (I'm sure that's been suggested a million times) afterall, that's what it's there for, right?

[BMW]

Originally Posted by SteveP

Where to begin ...

Everyone should have PHing for position players set to "Very rarely". Markus made the frequency of PHing for position players better in OOTP10, but it is still too high. Don't know why ... but it may have to do with OOTPer's thinking that PHing happens much more often than it does IRL. There are various measures of what's realistic, but the easiest it this: PHing for more than one position player in a single inning is very, very rare IRL. In historical leagues, where there were more position players on the bench, the frequency is greater: ranging from infrequently to rare. If you see teams PHing more than they should, then you can adjust that team strategy slider accordingly.

This is important, because the tendency of the AI to make mistakes on substitutions after PHing is much, much greater when PHing is frequent. The first example you give is a bug. It has been reported to Markus. He has never seen it, so it is not on the schedule to be fixed.

The others are just examples of AI substitution quirks. In OOTP, the game is designed so that anyone can play any position. This is unrealistic, but it is expedient (because of the inability of the AI to plan ahead, which means that the AI can easily paint itself into a corner). The problem is that the AI has a tendency to take advantage of that fact, even when it isn't necessary. If, however, you have PHing at a realistic level, these unrealistic fielding assignments become very infrequent.

This could be the greatest post ever. Lets see if this helps the issue in my online league.

[thehatfield]

Originally Posted by RchW

Yeah, this is a major frustration. Unless you play out each game the AI (or manager) will use subs the way he (it) wants to. It really affects relief pitching as the AI will continue to use the same relievers over and over, even if they are bad, while it will not use others having better years, unless the game is lost. I don't doubt that this could cost 3-8 wins per season, depending on the circumstances.

We should be able to force the issue by designating first out of bullpen (if not tired) second and so on.

Agreed on that. No matter where I tried to hide a certain reliever (even several times as only being assigned as the mopup guy) in the online league I play in, the manager would ALWAYS use him in every and any situation. Ended the season with 99 innings from 78 appearances with an ERA near 6.

[endgame]

Originally Posted by johnnyshaka

what about another option in the player strategy window that will let you specify, if you want to, the minimum position rating a player needs to have before being considered as a defensive sub at any position(s)?

Well that's ten minutes of my life I won't get back. Thanks a lot, johnny. When I first read your post, I read another as "the other" and went scrounging around trying to find that option. I couldn't believe I didn't know about it. Well...doh! Good idea.

And yes, SteveP, thanks. You continue to contribute good strategic ideas.

[SteveP]

Originally Posted by johnnyshaka

Regarding your final point about the AI sub quirks, what about another option in the player strategy window that will let you specify, if you want to, the minimum position rating a player needs to have before being considered as a defensive sub at any position(s)? Maybe you can select each position manually that a player may be subbed in at. Or else, have the AI use the depth chart (I'm sure that's been suggested a million times) afterall, that's what it's there for, right?

In theory, the AI in OOTP10 is not supposed to take someone out for a PH or PR, unless there is a "suitable" fielding substitute available to replace that person (which could, of course, be the PHer/PRer). My sense is that Markus programmed this in a way which was fairly generous in the definition of "suitable". Moreover, the AI can still get itself in trouble, despite that feature.

Realistically, there is a limit to how far you can go in constraining the choices available to the AI. Any constraint you add will undoubtedly backfire in some other way. The AI is not smart. The AI can't plan ahead. The AI has trouble taking account of all the variables in a situation. IMHO, one just has to be willing to accept some questionable decisions by the AI (though, for the moment, I'm not willing to include the AI's refusal to follow the depth chart in that list ) Since I play solo, I let the AI manage my team in-game -- that way, my AI manager gets the license to be as dumb as the other AI managers in the league. But, if I were really annoyed by substitution decisions, I would take control of that myself.

[RchW]

This and your other post are excellent. It had not occurred to me before but the thousands of substitution decisions that may have to be made must have some flexibility or you would simply crash the game.

Part of the problem it seems is other flaws of the AI, especially in the minors where it doesn't follow the very simple roster logic of 2-3 catchers 5-6 outfielders and 5-7 infielders. Not having suitable subs to begin with creates a cascade effect IMO.

Another logic flaw is the tendency to ignore defensive ratings and start unsuitable (low def ratings but marginally better rated hitters) players in the infield or even the OF. The "Force Start" option helps but would be easier to use if it was a drop down on the lineup screen instead of buried in the individual player page.

I still find it bizarre that the game produces so many IF that can play all four positions well yet produces OF with no rating at one position.

[johnnyshaka]

Originally Posted by SteveP

Realistically, there is a limit to how far you can go in constraining the choices available to the AI. Any constraint you add will undoubtedly backfire in some other way. The AI is not smart. The AI can't plan ahead. The AI has trouble taking account of all the variables in a situation. IMHO, one just has to be willing to accept some questionable decisions by the AI (though, for the moment, I'm not willing to include the AI's refusal to follow the depth chart in that list ) Since I play solo, I let the AI manage my team in-game -- that way, my AI manager gets the license to be as dumb as the other AI managers in the league. But, if I were really annoyed by substitution decisions, I would take control of that myself.

I don't know how telling the AI not to pinch hit for somebody in the 9th inning if there isn't a defensive replacement of "x" rating available to sub in is all that difficult. I'm not asking the AI to be "smart" or "plan ahead"...if you want to pinch hit for somebody, look down the bench to see if you have the criteria necessary before making the change. Sounds pretty foolproof to me.