11.25 Pinch-hit AI too conservative with pitchers in late innings

[HoustonGM]

In the 8th inning, if the AI is planning to bring the closer in, it should pinch hit for whichever pitcher is currently in the game.

It was the bottom of the 8th in a 3-0 game, and Derek Lowe had thrown all 8 innings. He was allowed to hit, but then was pulled for the 9th for closer Takashi Saito. The AI should've pinch hit for Lowe, rather than allowing the pitcher to hit.

[ohms_law]

Agreed

[jkh]

I was also wondering if something can be done to the game to allow pitchers to be used as pinch hitters, pinch runners, and emergency defensive replacements.

[HoustonGM]

I was also wondering if something can be done to the game to allow pitchers to be used as pinch hitters, pinch runners, and emergency defensive replacements.
See Pitchers as Pinch Hitters and Pinch Runners (http://forum.sportsmogul.com/showthread.php?t=177426)

[filihok]

As HoustonGM pointed out, the AI is not pinch hitting for the pitchers as it should.
But maybe it's just a Dodgers thing.

In my game Dodgers lead 3-2 top of the ninth. P Ray Searage strikes out to end the inning and Dave Walsh comes in to pitch the bottom of the ninth.

The .MOG is attached if anyone wants to look. The game in question is April 2nd.

[belial]

In the 8th inning, if the AI is planning to bring the closer in, it should pinch hit for whichever pitcher is currently in the game.

It was the bottom of the 8th in a 3-0 game, and Derek Lowe had thrown all 8 innings. He was allowed to hit, but then was pulled for the 9th for closer Takashi Saito. The AI should've pinch hit for Lowe, rather than allowing the pitcher to hit.

That may not be a high leverage situation depending on the runners and outs - maybe the AI is doing the correct thing and not burning a pinch hitter because it may need one in a better scoring situation. I don't think it is as easy as saying, 'The AI should have pinch hit because a new pitcher was coming in'.

[HoustonGM]

Find me some instances of where a pitcher was allowed to hit, and then was immediately removed for a reliever.

If the pitcher isn't staying in the game to pitch, a pinch hitter is used, unless the pitcher is a real good hitter, where the alternatives are worse than he is, but there's only a few pitchers in today's game where that applies.

[belial]

Find me some instances of where a pitcher was allowed to hit, and then was immediately removed for a reliever.



You didn't say please.

[Reade]

Find me some instances of where a pitcher was allowed to hit,

I know of it happening once this year, but the pitcher was a really good hitter(Zambrano) so that is not a good example.

[belial]

Well I think Zambrano has pinch hit before. :)

Do you want the AI to make intelligent decisions or just copy managers' behaivor? I'm not sure burning a pinch hitter that almost never would be likely to score in most situations is the best decision. And as far as typical manager decisions go, some managers throw out the book completely.

[filihok]

Well I think Zambrano has pinch hit before. :)

Do you want the AI to make intelligent decisions or just copy managers' behaivor? I'm not sure burning a pinch hitter that almost never would be likely to score in most situations is the best decision. And as far as typical manager decisions go, some managers throw out the book completely.

As in my instance. The visiting team leading 3 - 2 in the top of the ninth. Two out. Pitcher due up. No runners on. The majority of managers would pinch hit.

It's a one run game. Any chance to score a run must be taken.

The pitcher made out. And a new pitcher came in to pitch the bottom of the ninth.

[JeepGuy63]

Agreed ... should have pinch hit.

[HoustonGM]

Do you want the AI to make intelligent decisions or just copy managers' behaivor? I'm not sure burning a pinch hitter that almost never would be likely to score in most situations is the best decision. And as far as typical manager decisions go, some managers throw out the book completely.
For a simulation game, yes, I do think that it should copy real-life behavior as best as possible. You have a 3-0 lead in the bottom of the 8th inning. Your starting pitcher is up, and you're planning on bringing in the closer for the 9th. You have a full bench. It is smart to use a better hitter than your pitcher in order to increase your chances of extending the lead. With one inning left and a 3 run lead, the chances that you wasted a pinch hitter that you'll later need are incredibly slim.


It's a one run game. Any chance to score a run must be taken.
Exactly.

[belial]

It was a 3 run game.

[HoustonGM]

It was a 3 run game.
Exactly, which means the likelihood that using the pinch hitter would later result in a situation where you need him and can't use him is extremely, EXTREMELY slim, which means that using the pinch-hitter only serves to increase your chances of extending your lead. Since the pitcher is being removed anyway, it's silly to handicap yourself offensively with a near-automatic out, as well as increasing the chances that the pitcher gets hurt.

As a baseball simulation, it should model real-life as best as possible. There are many things that happen in real life that aren't the smartest things, and Mogul should replicate that, even if it's not the smartest move. In this specific case, I do think that using a pinch hitter to hit for a pitcher who is being taken out of the game the very next inning is the smart move, but even if it wasn't, since it's how the game is played often in real life, Mogul should simulate that.

The only times I can recall seeing the pitcher hit and then be taken out the very next inning are when

a) the pitcher is a good hitter, ie. the Carlos Zambrano example from Reade.
b) the pitcher was going to stay in, but got hurt while batting.
c) the pitcher was going to stay in as he's having a good game and the score is close, but during the inning that he hits, his team extends the lead a lot, and he's pulled to save his arm, as good pitching isn't as necessary in a blow-out game.

[beerchaser]

It's highly unlikely a pitcher would hit in that situation, but is it impossible? That's what y'all kept asking me on the "deep fly to CF" thread. Sorry, but turnabout is fair play. :p

[filihok]

as well as increasing the chances that the pitcher gets hurt.

Joe Morgan AND Hank Steinbrenner in the forum? Woo-Hoo!!!

[HoustonGM]

It's highly unlikely a pitcher would hit in that situation, but is it impossible? That's what y'all kept asking me on the "deep fly to CF" thread. Sorry, but turnabout is fair play. :p
No, it's not impossible. I already gave situations where it's possible. I'm saying the frequency needs to be turned down.

[belial]

Exactly, which means the likelihood that using the pinch hitter would later result in a situation where you need him and can't use him is extremely, EXTREMELY slim, which means that using the pinch-hitter only serves to increase your chances of extending your lead. Since the pitcher is being removed anyway, it's silly to handicap yourself offensively with a near-automatic out, as well as increasing the chances that the pitcher gets hurt.

As a baseball simulation, it should model real-life as best as possible. There are many things that happen in real life that aren't the smartest things, and Mogul should replicate that, even if it's not the smartest move. In this specific case, I do think that using a pinch hitter to hit for a pitcher who is being taken out of the game the very next inning is the smart move, but even if it wasn't, since it's how the game is played often in real life, Mogul should simulate that.

The only times I can recall seeing the pitcher hit and then be taken out the very next inning are when

a) the pitcher is a good hitter, ie. the Carlos Zambrano example from Reade.
b) the pitcher was going to stay in, but got hurt while batting.
c) the pitcher was going to stay in as he's having a good game and the score is close, but during the inning that he hits, his team extends the lead a lot, and he's pulled to save his arm, as good pitching isn't as necessary in a blow-out game.
Honestly, when the AI does something like this I imagine a situation like that. I make up a reason why such a thing happened, because these do happen in real life and mogul can't simulate all the oddities. I wouldn't pinch hit with a 3 run lead and 2 outs no one on because I feel losing a pinch hitter in a situation I would almost certainly not score is worth going for a run. The problem with the leverage charts is they don't factor in how much of a bench is left. If I did pinch hit, it wouldn't be my best hitter, and why burn a bench player to gain a minimal chance to score?

The problem though is probably not the pinch hit ai, but when the ai works on the decision to make a pitching change. I would almost bet it doesn't decide on pitching changes except before its defensive inning or in the middle of an inning.

[HoustonGM]

Honestly, when the AI does something like this I imagine a situation like that. I make up a reason why such a thing happened, because these do happen in real life and mogul can't simulate all the oddities.
Yes, I do this too, as I illustrated in the long flyball/putout thread. The situations I listed in the quoted post, though, are specific and can't be imagined, unless you want to imagine a pitcher is a good hitter even if he isn't, or imagine that the team scored a bunch of runs even if they didn't, which is fine if that's what somebody wants to imagine.


If I did pinch hit, it wouldn't be my best hitter, and why burn a bench player to gain a minimal chance to score?
If I'm managing, I want to maximize the chance of scoring no matter what, even if it's just a minimum gain. I also want to minimize the risk of injury, and letting a pitcher hit when you're going to take him out risks injury with no benefit at all.


The problem though is probably not the pinch hit ai, but when the ai works on the decision to make a pitching change. I would almost bet it doesn't decide on pitching changes except before its defensive inning or in the middle of an inning.
This is a very good point and probably true. The AI does need to "learn" to "look ahead."

[belial]

I want to maximize my chance to win the game, regardless of the score.

[HoustonGM]

I want to maximize my chance to win the game, regardless of the score.
Yes, so do I. With a 3 run lead, using a better hitter than a pitcher to hit, when you know you'll be taking that pitcher out of the game after the inning, is increasing the chance that you score a run that inning (even if it's just a slight increase), padding your lead, which increases your chance of winning the game.

When you're taking your pitcher out for a reliever, and you let him hit, which is usually a near automatic out, you're not maximizing your chance of winning.

[belial]

I'd like to see leverage charts with bench players left and the subtraction value of using a bench player so this can be proved one way or the other. I am of the opinion that bench players left is vastly underrated.

[filihok]

A reminder that THIS still needs a lil tweeking.

http://i496.photobucket.com/albums/rr321/NatsDynasty/2011/SEPTEMBER/PinchHitBug.jpg

The AI let the starting pitcher make the 3rd out in the 8th inning, then relieved him to start the ninth.

It's September, there are expanded rosters. The AI should have hit for him if it was going to take him out of the game afterwards. There were plenty of hitters and fielders left should the game get extended so "saving" a pinch hitter shouldn't have factored in.

This nugget probably shouldn't get lost in the shuffle.


The problem though is probably not the pinch hit ai, but when the ai works on the decision to make a pitching change. I would almost bet it doesn't decide on pitching changes except before its defensive inning or in the middle of an inning.

[Fish Troll]

If I'm managing, I'll have someone pinch hit just so my bench get work and increase their playing time. Its 3 more outs, I rather go for the win then worrying if the other team will tie it.