Why so unrealistic sometimes?

[Dyzalot]

Originally Posted by Rain King

Those were the 6th & 7th words of your post.

And the 3rd, 4th and 5th words were "one of the". Easy to take things out of context to make them seem more ridiculous than they actually were.

[Brad K]

Originally Posted by Dyzalot

Also note that if I was trying to "troll" the community, I wouldn't have answered the first reply within ten minutes.

Not necessarily. Some people like to watch people argue. Some people like to be involved.

[OutS|der]

This game is so unrealistic. Today the Braves beat the Marlin's 29-9. One pitcher even gave up 12 runs. The game is ruined
This would never happen in real life.

[thehef]

In the end, I think the issue brought up in the OP straddles the line between:

"This is a bit unrealistic, and possibly an area for the developers to tweak postseason AI logic,"

and

"Have you seen Dave Roberts manage in the postseason in real life?"

[Dyzalot]

Listen, if you guys think this is a remotely realistic box score, then produce one from the past 50 years in MLB that is even close. Show me a game where the team was down by 5+ early, starting pitcher removed by the 2nd inning, team lost by 8+ and relief pitchers got every PA available to the #9 spot. It can be playoffs or regular season. But the game has to matter so it can't be from a team that was already out of the playoff hunt.

I don't think this is "a bit unrealistic". I think this is complete fantasy.

Heck, I'd be surprised if anyone can find a single MLB game in the past 50 years regardless of score where relief pitchers got all three plate appearances available to the #9 spot. Real MLB managers just don't manage games that way.

[Syd Thrift]

Originally Posted by Dyzalot

Puzzles are fun!

This is literally what trolling is.

[Dyzalot]

Originally Posted by Syd Thrift

This is literally what trolling is.

I was being facetious. To me that box score was as much of a "puzzle" as asking someone which number comes next in this set:

1,2,3,4,?

I didn't think it would be that difficult for people to notice that relief pitchers got all the plate appearances in a game where they probably should have gotten none of them.

[Rain King]

Here is an example from 2019.

https://www.baseball-reference.com/b...01904160.shtml

Yes, these are rare and it is not realistic for this to happen often. I also don't think it really matters, which is why this thread is somewhat annoying. In the realm of tweaks this game could use in order to make the AI more competitive I don't imagine allowing relief pitchers to hit too many times during a blowout is anywhere near the top of the list.

The AI likely doesn't have a concept of letting a bench guy get a random at-bat to appease them or to get them some work or maybe to avoid sending relief pitchers up there to embarrass themselves.

So, yeah, I think they should put this on the list of stuff they need to fix...way down there somewhere. Thanks for pointing it out.

[Dyzalot]

Originally Posted by Rain King

Here is an example from 2019.

https://www.baseball-reference.com/b...01904160.shtml

Yes, these are rare and it is not realistic for this to happen often. I also don't think it really matters, which is why this thread is somewhat annoying. In the realm of tweaks this game could use in order to make the AI more competitive I don't imagine allowing relief pitchers to hit too many times during a blowout is anywhere near the top of the list.

The AI likely doesn't have a concept of letting a bench guy get a random at-bat to appease them or to get them some work or maybe to avoid sending relief pitchers up there to embarrass themselves.

So, yeah, I think they should put this on the list of stuff they need to fix...way down there somewhere. Thanks for pointing it out.

I agree that when it is actually a blowout, making it so the AI pinch hit for pitchers wouldn't help much. However, when the relief pitcher has already gone two innings and thrown 30+ pitches with the score 5-1, it might behoove the AI to pinch hit for him when leading the inning off instead of letting him hit so that he can face one more batter to "save the bulllpen". Also note that in your example, the guys on the bench at least still got used. Perhaps that should have been a qualifier since it does make sense to me that in a regular season game, it can make more sense to use the bench to rest your regulars in a blowout as opposed to using them to pinch hit. I wouldn't say that for the playoffs, but for a regular season game there is logic to it.

And really I would argue that fixing the AI so that it treats "must win" or playoff games differently from "run of the mill" regular season games differently is the fix I am looking for. It isn't so much about the lack of pinch hitting in the specific game I posted as it is the AI logic behind it being done in the playoffs. I've even seen this kind of behavior from the AI in Wild Card games. That's a part of strategy that could be "tweaked" that actually might make a measurable difference in how competitive the AI is.

One last note, the AI was controlling both teams in my example. Any fix in this department wouldn't actually affect my saves in terms of how competitive the AI is since I sim my games.

[Rain King]

Originally Posted by Dyzalot

And really I would argue that fixing the AI so that it treats "must win" or playoff games differently from "run of the mill" regular season games differently is the fix I am looking for. It isn't so much about the lack of pinch hitting in the specific game I posted as it is the AI logic behind it being done in the playoffs. I've even seen this kind of behavior from the AI in Wild Card games. That's a part of strategy that could be "tweaked" that actually might make a measurable difference in how competitive the AI is.

This text would have been really useful in your original post.

[Dyzalot]

Originally Posted by Rain King

This text would have been really useful in your original post.

Well I kind of thought that was where the conversation would go. I didn't foresee getting attacked as a troll simply because I didn't think my OP was really all that cryptic. It just seemed really obvious to me. Obviously I was wrong.

[Rain King]

I have found that it is always better to spell things out if possible.

Also, the game does have some concepts of the playoffs being different than a regular game. It will pitch an Ace on short rest or bring them out of the bullpen in an elimination game, for example.

I think from the developer's point of view it is a very slippery slope to get aggressive with bench players. Generally, they are likely looking for a high enough leverage scenario to pinch-hit so as not to waste a player that could be better used later in a more important situation. In this situation they would need to know to override that leverage threshold because it is unlikely to get better and that is not a simple concept.

Every release this kind of stuff gets a little better, I'm sure this will get there at some point also.

[CBeisbol]

Originally Posted by Dyzalot

Puzzles are fun!

That may be
But they aren't the most constructive way to solve a problem

People, at least a few, liked trying to figure out when they would have pinch hit.

But, here's a thing with no context, figure it out, isn't a great way to solve a problem.

Originally Posted by Dyzalot

Listen, if you guys think this is a remotely realistic box score, then produce one from the past 50 years in MLB that is even close.

Originally Posted by Rain King

Here is an example from 2019.

https://www.baseball-reference.com/b...01904160.shtml

So, last year, then

[Dyzalot]

Originally Posted by CBeisbol

So, last year, then

Not sure how "close" that was since the team was down 10 early as opposed to only being down 4 in the 3rd but sure. At least that one makes sense as they used the bench to rest regulars instead of pinch hitting. I wouldn't have posted an example like that because it has some logic to it.

But, here's a thing with no context, figure it out, isn't a great way to solve a problem.

I gave plenty of context. To see a team lose game #1 of a WS where the relief pitchers got all three plate appearances should be a "red flag" to anyone looking at the box score. Again, I thought it was pretty easy to see such that my "puzzle" comment was more of a joke than anything else.

[joefromchicago]

Originally Posted by Dyzalot

I gave plenty of context.

No. No you didn't. Since your initial post, we have learned:

• The bullpen was fully rested;
• There were eight relievers available;
• There were seven bench players;
• That one of those guys was a two-way player;
• You like puzzles.

You've mentioned this information because you thought it was relevant to your point. That information, therefore, must also have been relevant when you made your initial post. Yet you didn't include that information in your initial post. So no, you didn't provide "plenty of context." You provided a box score.

And your numbers don't add up. According to you, the Giants had: (a) seven relievers; (b) six bench players; (c) one two-way player. That's 14. Considering that you have eight position players and let's say four starters (it's the postseason, you can get away with four), that's 26 players. Now, your roster limit may, indeed, be 26, but then that would be yet another piece of the context that you still haven't provided.

In addition, at least one of those bench players wouldn't be available for regular pinch-hitting duties because he's the backup catcher (unless you think that it's realistic for a manager to send in his backup catcher as a pinch hitter in the third inning). And if the two-way player is a better pitcher than hitter (we don't know - yet more context that we don't have), then you really only have five players available to pinch hit. No manager knows how many of those guys he'll need for the late innings or defensive subs or whatever, so some managers are conservative about using pinch hitters early.

Also, you keep emphasizing that the bullpen was fully rested, but this is game 1 of the WS, which means there's a game tomorrow. So the manager not only has to be aware of how rested his relievers are today, but also how rested they'll be for tomorrow's game. The Giants used four relievers in this game, so that's half of the bullpen, and it was a blow-out. I sure hope the Giants' starter in game 2 was a lot better than Anderson was in game 1, and I'm sure the Giants manager was hoping the same thing.

In addition, you point out that the AI had Burzell bat in the third inning "so he could come out in the top of the 4th to pitch to one batter before being removed." Again, what's the context here? Is Burzell the fifth starter who is in the bullpen for the post-season for just this sort of situation? If so, I'd let him in as long as possible too. And why was he lifted in the fourth? You seem to think that the AI knew beforehand that Burzell would only register one out in the fourth inning, but what human manager can know that? Did he allow a base runner? Did he reach his pitch limit? Was he injured? Was he ejected? Just more context that's missing.

I like puzzles too. The puzzle here is why you think you provided enough information in the initial post but then keep providing more information as the thread goes along. If there was enough information in the initial post, there wouldn't be any need to give us more information later. Truly, it's a puzzle.

[Matt Arnold]

Yeah, there's definitely a lot here that doesn't quite look right at a glance, but hard to say what exactly is going on. The AI is definitely more aggressive in playoff games, and I would expect that even late in a 9-2 or 10-2 game they would usually be pinch hitting for any pitchers.

The main exceptions I can see are if one of those guys who batted was the two-way pitcher, they generally speaking won't pinch hit for them. That could account for one of those ABs. The early case where they let the reliever hit I think is fair, given that's the 3rd inning, although given he was removed a batter into the next inning, that doesn't match correctly since it doesn't even seem like it was because he was getting into trouble.

So yeah, I'm sure there are some things we can tune from this, but as alluded by others, the more information available, the easier it is to dig down. Even in a case like this double-checking what situations the pitchers were coming up in could point to where the problems lie most.

[CBeisbol]

Originally Posted by Matt Arnold

The early case where they let the reliever hit I think is fair, given that's the 3rd inning

They should be more aggressive in the early innings

Teams need to use their bullpens "backward" in games like this. Using the short relievers early to keep the game close. Any time you're down (or the underdog, say a team projected to finish second over the course of a season) you are looking for higher variance. Take more risks. Increases the chance you'll come back, but also opens you up to getting blown out.

A good example, is bringing the infield in. This strategy increases the chance of cutting off the run at the plate, but also increases the chance of giving up another hit leading to further runs.

Bring in the higher leverage relievers early, decrease the chance of more runs (and allows for more aggressive pinch hitting) but opens the team up to habing worse relievers available for later inning high-leverage situations, should they arise.

Of course, only the sabermetriciest AI manager would do that

[Matt Arnold]

Originally Posted by CBeisbol

They should be more aggressive in the early innings

Teams need to use their bullpens "backward" in games like this. Using the short relievers early to keep the game close. Any time you're down (or the underdog, say a team projected to finish second over the course of a season) you are looking for higher variance. Take more risks. Increases the chance you'll come back, but also opens you up to getting blown out.

A good example, is bringing the infield in. This strategy increases the chance of cutting off the run at the plate, but also increases the chance of giving up another hit leading to further runs.

Bring in the higher leverage relievers early, decrease the chance of more runs (and allows for more aggressive pinch hitting) but opens the team up to habing worse relievers available for later inning high-leverage situations, should they arise.

Of course, only the sabermetriciest AI manager would do that

They are set in playoff games to handle that a little bit more. Especially if you're swapping out with men on base early, they'll target a "middle relief" option first, and then swap to a long reliever an inning or two later.

[CBeisbol]

Originally Posted by Matt Arnold

They are set in playoff games to handle that a little bit more. Especially if you're swapping out with men on base early, they'll target a "middle relief" option first, and then swap to a long reliever an inning or two later.

I'd, so maybe some AI managers should, be more aggressive.

Down 3-1, you don't want to go down 5-1 or 6-1 or 7-1. So if the starter is going to get pulled it needs to be for a good reliever.

Then, again, I'd use the relievers in reverse.

In this kind of game, the higher leverage situations are in the earlier innings when there's still time to catch up. So I'd be using my highest leverage pitchers against 4- 5 of their better hitters (maybe using a lower leverage pitcher against the bottom of their order) and not letting a pitcher bat unless, maybe, it was nobody on two outs where a hit might actually decrease my expected number of runs over the next two innings (slight increase in the current inning, coupled with a decrease in the next inning since my leadoff hitter (or worse, my lead off and #2 hitters both hit and failed to produce a run in the previous inning)). All the while looking for a spot where I can bring in a long reliever to pitch 2-3 innings where they are both unlikely to give up a run (say facing the middle of the order so if they get in trouble, it'll be runners on vs the bottom of the order, or getting a few easier outs before facing the top of the order) and come to bat.


So, yeah, it's complicated in that there are multiple factors to consider and it's hard to weight them all properly.

But AI managers should have different, and sometimes, vastly different styles and philosophies. Also, probably difficult

[dkgo]

The AI Buck Showalter though is still waiting for that save situation to bring Zach Britton in!