Does pitcher stamina need to be changed?

[joefromchicago]

I just noticed this post on the OOTP Reddit forum and it piqued my interest. And since it focuses on pitchers in historical leagues, I decided to post it here. Here's the original post from RayFowler:

Pitcher Stamina, as implemented, is the worst design flaw in OOTP

I say this despite loving OOTP, but I am a stickler for realism to the point that any time I start a new OOTP game, the first thing I do is fix all of the pitcher staminas.

• What is pitcher stamina? This is a statistic that essentially controls how many pitches a player can throw before he gets "tired" and have to be pulled from the game to avoid injury or getting excessively shelled.
• What is the point of pitcher stamina? I'm not the dev, but I suspect that the stamina stat is a brute-force method to prevent players from cheesing the game running a starting rotation full of closers.
• Why is it a bad design flaw? Well, it doesn't mirror reality very well. While some pitchers in reality obviously have more stamina than others, the majority of pitchers that you normally think of as a "bullpen" specialist started games regularly before coming up to the majors.
• Examples? Well, I just looked at the top 10 all-time saves leaders and 8 of them were starters in the minors. Of the two who weren't, Trevor Hoffman started briefly in the minors, and Jonathan Papelborn was the only one who was exclusively a starter (groomed as a closer in college).
• How is this a problem? Well, the main issue is that pitcher stamina is derived from usage data (how their team chose to play them) but is treated in the game as an unchangeable physical measure (like speed or fielding ability). This means that if you are playing a historical sim and a real-world manager decided to move Dennis Eckersley to the bullpen, OOTP requires you to make the same decision as well (because Eckersley's stamina drops like a rock the next season).
• How else? If you are playing a historical game but using the OOTP talent engine, very often you will see famous workhorse starting pitchers (examples include but aren't limited to Phil Niekro, Nolan Ryan and Wilbur Wood) become permanent bullpen specialists because maybe their first couple of seasons were spent in the bullpen. To me, this is almost a fatal flaw for historical sims. In addition, whatever algorithm is used to calculate a pitcher's stamina sometimes comes up with a ridiculously low value like '1', which makes a pitcher almost completely worthless
• How can it be fixed? Actually, this is the funniest thing of all. It's almost trivial to fix, conceptually, but just really tedious to implement. What I literally do now, for historical sims, is just bump EVERY pitcher's stamina to max. This seemed like a crazy idea at first, but you know what happened? The best 5 pitchers on the team became the starters and this very closely matched the real world anyway. The game already recognizes that some good pitchers cannot be starters (not enough good pitches), so a lot of very good closers remained closers. In addition, if you have a starter go on the DL (when he didn't in real-life), then you have the same option the real manager would have had: move one of your bullpen guys into the rotation. This happens a lot in reality, but the pitcher stamina restriction for relievers makes this impossible in OOTP.

What I would ideally like to see is some sort of startup option for pitcher stamina to set a pitcher's stamina as the "maximum" stamina exhibited in any of his seasons, not a career or 3-year average. That would be the most accurate and ideal solution to this problem.

That is all I have to say on this. I had to get it off my chest.

So, is pitcher stamina broken for historical leagues? I have some ideas on this topic, but I'd like to see what everyone else thinks.

[luckymann]

Paging ALB123... ALB123, call for you in the Historicals Forum.

[joefromchicago]

I'll start things off by addressing some of RayFowler's points:

What is pitcher stamina? This is a statistic that essentially controls how many pitches a player can throw before he gets "tired" and have to be pulled from the game to avoid injury or getting excessively shelled.

I could be wrong, but I think that stamina measures two different things in OOTP: (1) how many pitches a pitcher can throw before he gets tired (in-game stamina); and (2) how fast a pitcher recovers after pitching (between-game stamina). The manual states that stamina "is a measure of how many pitches a pitcher can throw before tiring." The manual also describes the league "pitcher stamina" setting in the Stats & AI page as both "a measure of how quickly pitchers tire in your league" and "a league-wide adjustment of how quickly pitchers tend to become fatigued." Seriously, the manual lists "pitcher stamina" twice and describes it in two different ways. So there's some ambiguity here as to what "stamina" actually is, but it appears that the individual ratings and the league setting together control both in-game stamina and between-game stamina.

What is the point of pitcher stamina? I'm not the dev, but I suspect that the stamina stat is a brute-force method to prevent players from cheesing the game running a starting rotation full of closers.

That's certainly part of it, although there are two main determinants for a pitcher being a starter in OOTP: (1) stamina; and (2) number of pitches. A pitcher needs at least three pitches in order to be a starter (except for knuckle-ballers, who can get by with two). So even if you have a pitching staff composed entirely of pitchers with top-rated staminas, you still can't make them into starters unless they also have at least three pitches.

Why is it a bad design flaw? Well, it doesn't mirror reality very well. While some pitchers in reality obviously have more stamina than others, the majority of pitchers that you normally think of as a "bullpen" specialist started games regularly before coming up to the majors.

I've had a lot to say about the way that OOTP assigns stamina ratings to pitchers in the pre-reliever era who started but who didn't log a lot of innings. In short, my conclusion is that OOTP punishes those pitchers in two ways: by giving them bad pitching ratings and by giving them low stamina ratings. That, in effect, punishes them twice for being bad pitchers. But it's not that they were bad because they didn't pitch a lot, it's that they didn't pitch a lot because they were bad. In other words, their lack of skill is not necessarily correlated to any lack of stamina. So OOTP can definitely improve in this respect.

How is this a problem? Well, the main issue is that pitcher stamina is derived from usage data (how their team chose to play them) but is treated in the game as an unchangeable physical measure (like speed or fielding ability). This means that if you are playing a historical sim and a real-world manager decided to move Dennis Eckersley to the bullpen, OOTP requires you to make the same decision as well (because Eckersley's stamina drops like a rock the next season).

Despite my problems with OOTP's handling of pitcher ratings in the pre-reliever era, I will diverge somewhat from RayFowler on this point. As I see it, his focus is on the era from about 1960 to around 1990. That was a transitional period where a definite split developed between starters and relievers. OOTP rates pitchers so that the game will produce realistic results. That means rating relievers so that they relieve and starters so that they start. So if you're playing in this period, you just have to accept that some pitchers are going to be relievers and some are going to be starters. If you want to change that, then you have to accept that you're not playing historical any more, you're playing a fictionalized version of history.

I think RayFowler's gripe here is that he can't use pitchers in ways that they weren't used historically. I think that's a fair point, but it's a problem that's largely on the margins. As I see it, OOTP does a bad job of assigning stamina ratings to starters who didn't pitch a lot of innings. But for pitchers who spent most of their time in the bullpen, there has to be a mechanism in place to prevent them from being used in unrealistic ways. For OOTP, that mechanism is, in part, the stamina value that each pitcher is assigned.

Granted, that may be a "brute-force" way of achieving the desired result, and the gamer is necessarily constrained by decisions that were made by the on-field personnel at the time. So yes, Dennis Eckersley goes from being a high-stamina starter to a low-stamina reliever because of a decision made by Tony LaRussa and not because Eckersley suddenly became too frail to start. But OOTP is trying to get as close to being historically accurate as possible, and if you want Eckersley to remain a starter, then you have to acknowledge that you're no longer playing a historical simulation. That's not a problem with the way OOTP handles stamina ratings, that's a problem with the gamer wanting to play both historical and fictional at the same time.

[Garlon]

There really is no issue with how OOTP is handling pitcher stamina.

1. A pitcher is going to be given SP stamina if at least 40% of their Games Pitched were as Games Started. The formula is based upon IPouts per GS. This works in tandem with the League Totals Modifiers. Remember that you can choose to have pitcher Stamina determined based upon 1 season, 3 seasons, 5 seasons, or career.

2. If every pitcher were given SP potential stamina this would be extremely detrimental to the real starting pitchers historically as it would put relievers into the rotation and you will have guys like Mariano Rivera taking home the Cy Young Award as a starting pitcher.

3. Everything is currently in the game to create proper outcomes. Some settings need to be set to better default. For example, Use Relievers should be set to Normal for every season. Remember that the League Totals Modifiers will take care of IPouts per GS, so what remains is getting proper bullpen usage in terms of IPouts per relief appearance. If you set your pitching rotation and pitching staff limits correctly you can get great results for any season.

This guide will show you settings as well as historical statistics for various seasons. There are columns that show the expected average innings for relief pitchers for each team as well as the expected number of relief appearances. If you use the suggested strategy settings you will end up with really good results. Note, when I originally created this guide I put Use Closers as Very Often for every season, but I will update this. The reason I set it to that is for competitiveness because doing so has the computer identify the bullpen ace and by setting use closers to very often the computer will use their best reliever more appropriately in my open. The only issue with this is that you get higher save totals. I may have a revised version to provide appropriate closer settings.

[Garlon]

Another issue is how people are using the import settings.

For single season replays the default import settings are probably fine.

For 5yr I suggest

Batters 165/154
Pitchers 17/14

Remember, that for Starting Pitchers those values automatically get multiplied by 4. So with 5yr recalc that is 17*4*5 = 340 IP for starting pitchers. For relievers the sample size is 17*5 = 85 IP. This will keep the best players in the lineups and rotation and have them perform very well.

As a side note, by Setting Position Player fatigue to Very High (my suggestion) you will get more realistic result for batters. As you move this setting to Very Low you will get more dominance from the best starting pitchers. This is because on a Very Low setting the lineups tend to be stronger because the players do not need as much rest and pitchers who are below average will not perform as well because they are facing stronger lineups and the result is that good pitchers end up performing even better. With position player fatigue set to very high the best batters will actually perform more accurately in terms of both playing time and performance. As weaker batters get a bit more playing time the best batters, like Ty Cobb, will end up performing better when they do play and they will still be playing a realistic amount of games.

[joefromchicago]

Originally Posted by Garlon

There really is no issue with how OOTP is handling pitcher stamina.

1. A pitcher is going to be given SP stamina if at least 40% of their Games Pitched were as Games Started. The formula is based upon IPouts per GS. This works in tandem with the League Totals Modifiers. Remember that you can choose to have pitcher Stamina determined based upon 1 season, 3 seasons, 5 seasons, or career.

That's not exactly true. OOTP also flogs a pitcher's stamina rating if he didn't pitch a certain minimum number of innings. As I've mentioned in my posts on pitcher usage in the pre-reliever era, that's punishing a pitcher twice for being a bad pitcher.

There are also several problems with calculating stamina based on IP/GS. First, those stats are, as far as I'm aware, unavailable prior to 1916. Secondly, it's a stat that measures a bunch of other things besides a pitcher's durability that are beyond the pitcher's control. In addition to measuring how tired a pitcher might get, a pitcher's IP/GS stat is also dependent upon the manager strategy and the strength of the team's bullpen. It's sort of like basing a hitter's power rating on the number of RBIs he collected.

Finally, and most importantly from my perspective, the stamina ratings are a big factor in determining how the AI uses pitchers. In the modern era, that's usually not a problem, as there's a clear split between starters and relievers. But in the pre-reliever era, that split didn't exist. All pitchers were expected to be starters, which, in game terms, would mean that all pitchers should have "starter stamina." But that's not the case. According to the formula you laid out, any pitcher who started in less than 40% of his appearances gets "reliever stamina" as a default. And, as a result, the AI will put any pitcher who has "reliever stamina" into the bullpen, where he will be used like a modern-day reliever.

I'll give one example. Myles Thomas pitched for the Yankees in 1926. He had 13 starts out of 33 appearances, which put him at 39%, just shy of the magic 40% level needed for receiving "starter stamina." In OOTP, the result is that he gets a 40 stamina (20-80 range). And that's despite his 6.8 IP/GS ratio. Kent Greenfield of the Giants also had a 6.8 IP/GS ratio, but 71% of his appearances came as a starter, so he gets a 75 stamina rating. The AI designates Thomas as a middle reliever, even though the default is a five-man rotation and Thomas had the fifth-most starts for the Yankees that year (Sad Sam Jones, who had 11 starts to Thomas's 13, is slotted instead as the team's fifth starter*). If my past experience is any guide, that means that Thomas will rarely, if ever, start for the Yankees in any replay of the 1926 season.

You're correct to note that the ratings are all designed to produce realistic results and that they are all intended to work together to achieve that aim. But this is one situation where they work together to produce the opposite result. The stamina ratings combined with OOTP's reliever model just don't work for the pre-reliever era. I'm not optimistic that OOTP will change its reliever model any time soon. That means that, if we want some improvements in this area, it will have to come from refiguring the stamina ratings.

EDIT: *That's wrong. OOTP slots Walter Beall, who had nine starts, ahead of Thomas in the rotation. Jones had 23 starts.

[Garlon]

I suggested that the GS requirement be reduced to 33%. It used to be 50% and I told Markus that there just was not enough SP in the league. He agreed to lower it to 40% and in general this seems fine, but there are some issues for some teams whose 5th starter did not pitch many innings.

IP per GS is estimated for pitchers. For pitchers with more than 40% starts their IPouts per GS is (IPouts - (9*Relief Appearances))/GS. Relief appearances is simply G - GS. This in combination with the League Totals Modifiers will get pitchers to within a fraction of an out of their actual IPouts per GS. Note, if you want the best results you need to go into the engine file and set the weight for basing the Stamina League Totals Modifier calculation to 1 instead of 100.

[Garlon]

I tried a replay of the 1926 season using real transactions and lineups and those pitchers still make their starts.

[joefromchicago]

Originally Posted by Garlon

I suggested that the GS requirement be reduced to 33%. It used to be 50% and I told Markus that there just was not enough SP in the league. He agreed to lower it to 40% and in general this seems fine, but there are some issues for some teams whose 5th starter did not pitch many innings.

IP per GS is estimated for pitchers. For pitchers with more than 40% starts their IPouts per GS is (IPouts - (9*Relief Appearances))/GS. Relief appearances is simply G - GS. This in combination with the League Totals Modifiers will get pitchers to within a fraction of an out of their actual IPouts per GS. Note, if you want the best results you need to go into the engine file and set the weight for basing the Stamina League Totals Modifier calculation to 1 instead of 100.

Well then, that settles it: stamina is broken, at least for historical leagues.

[joefromchicago]

Originally Posted by Garlon

I tried a replay of the 1926 season using real transactions and lineups and those pitchers still make their starts.

If stamina only works for replay leagues using real transactions and lineups, then it's not working correctly.

[Garlon]

There is nothing broken with stamina. You get correct IPouts/GS for pitchers with career leagues without using real transactions and lineups. The game will always use the best available pitcher, and unlike real managers, the game can see the player ratings and make better decisions.

[Garlon]

If you want to see the real problem with historical leagues, take a look at the fielding results. Piazza and Rodriguez have about the same caught stealing percentage and there are players who perform impossibly good and impossibly bad, particularly at middle infield positions.

[joefromchicago]

Originally Posted by Garlon

I tried a replay of the 1926 season using real transactions and lineups and those pitchers still make their starts.

Well, of course pitchers are going to match their historical starts totals when you use historical lineups. That's sort of the bare minimum that we should expect from the game. But you didn't get historical results in your replay.

For instance, your starters didn't make as many relief appearances as they did in 1926. That's to be expected, given the OOTP pitching model. As a result, your top starters didn't log as many innings as they did in real life, while your relievers made way too many appearances. Garland Braxton, in your replay, had 33 saves. In real life, Braxton didn't have 33 (retrospective) saves in his entire career.

You may have gotten some realistic ratios for IP/GS (I can't tell from the results you posted), but, as I mentioned before, IP/GS measures external variables in addition to a pitcher's endurance. As such, getting realistic IP/GS results in a replay would almost qualify as a statistical anomaly, sorta' like flipping a coin ten times and having it land heads every time. To the extent that your IP/GS results matched historical results, it would be due to a combination of stamina ratings, strategy settings, and luck.

But then those sorts of ahistoric results should have been expected. I saw the same thing while examining how OOTP's pitching model falls apart during the pre-reliever era. My point regarding stamina in particular, however, is not just that the stamina ratings themselves are flawed (they are). It's that they work in combination with the OOTP pitching model to produce some wildly inaccurate results for that era, specifically with regard to relievers. When OOTP assigns a low stamina to a part-time starter, it effectively consigns that pitcher to the bullpen, and then the AI uses that pitcher like a modern-day reliever. And there's no way out of that hole, because pitchers who aren't in the rotation are automatically designated as relievers and the AI never uses them as part-time starters (even though, theoretically, it could).

I'm confident that the only reason guys like Braxton and Myles Thomas and Bob Shawkey started in your replay is because you were using historical lineups. If you hadn't, they would have never gotten a starting assignment. And that, to me, is just one more indication that the OOTP pitching model doesn't work for the pre-reliever era. OOTP doesn't do a very good job with pitchers who straddled the line between starter and reliever, and that is due, in part, to the way it handles stamina.

[Garlon]

If you have too many pitcher in the league with starting stamina you will end up with worse results. Pitchers who really were starters will get bumped in favor of those who had a small sample of innings. Things work in replay mode though. The saves may have been high because I set use closers to sometimes. The save does not mean much though, only that you were the last pitcher in a close game in which your team won. If the game identifies Braxton as the best reliever, why not use him to finish games?

[scurvycure]

Originally Posted by Garlon

... As a side note, by Setting Position Player fatigue to Very High (my suggestion) you will get more realistic result for batters...

I am not able to find where the setting for Position Player Fatigue is located.I see on the League Settings>Stats and AI page the General Strategic Tendencies, which includes a setting for Pitcher Stamina, but I am not seeing Player Fatigue there or on any of the other Settings pages. Where do I find this?

[Garlon]

You will find it under League Settings>Players & Facegen>Position Player Fatigue. Note, you can set this when create your league if you use the advanced mode, but the very high setting is not available there for some reason. You are able to change this to Very High after creating the league though.

[joefromchicago]

Originally Posted by Garlon

If you have too many pitcher in the league with starting stamina you will end up with worse results. Pitchers who really were starters will get bumped in favor of those who had a small sample of innings.

Worse results? Worse than what? The results that OOTP already produces? Because those are pretty bad to begin with.

I wouldn't have a problem with penalizing pitchers with a low number of IP by giving them low stamina ratings, if those low stamina ratings didn't, in turn, force them into the bullpen where they would be stuck for the rest of the season. The OOTP pitching model is based on a strict separation of the pitching rotation from the bullpen. In the pre-reliever era, however, there wasn't that split. There was just a pitching staff, and every pitcher was expected to be able to start and relieve. Right now, the stamina ratings work in tandem with the pitching model to reproduce that modern-day split between rotation and bullpen in years where there was no split. I don't expect OOTP to change its pitching model. That leaves stamina as an area where we can hope for some improvements.

Originally Posted by Garlon

The saves may have been high because I set use closers to sometimes. The save does not mean much though, only that you were the last pitcher in a close game in which your team won. If the game identifies Braxton as the best reliever, why not use him to finish games?

As I've mentioned many times elsewhere, in the pre-reliever era, a team's best reliever was usually a starter. Garland Braxton was actually a pretty good reliever for the 1926 Yankees. But in high-pressure situations, I'll wager that Miller Huggins called on one of his starters to get the Yanks out of a jam. Of the 22 retrospective saves that the Yankees had that year, 14 went to the top four starters. So if the game identifies Braxton as the best reliever, then the game isn't being historically accurate.

[Garlon]

If they used a starter to finish out the games then that is why they needed Braxton to make those starts historically because the starter had just pitched in relief. Kind of a wash. That was the manager's decision and not necessarily what provides the optimal strategy.

What I am saying is that if you make these other pitchers stronger then you are making the best pitchers more average. There is only so much room for performance in historical leagues because of the League Totals that need to be met. I agree with you though that there needs to be more pitchers with higher endurance. I really pushed Markus to move the criteria to 33% but he only agreed to 40% and I needed a mountain of evidence to show him that there was not even an average of 5 starting pitchers for some teams in modern seasons. He did not want to make the percentage a user option in the game so he coded it to 40%. If it was 33% then there would be more of the types of pitchers you want. For example, those pitchers on the 1926 Yankees you want to have more endurance would have better ratings. I think there needs to be an option in the game to use 40% of 33% as the criteria. I told Markus that a pitcher whose GS is at least 33% of their Games Pitched still is pitching more innings as a starter than as a reliever. It was actually a significant benefit moving the criteria from 50% to 40% but I do not think he will agree to give us an option in the game to adjust it further.

Please take a look at my defensive thread.

[ChessSafari]

Fascinating and enlightening discussion, guys. Many thanks.

[joefromchicago]

Originally Posted by Garlon

If they used a starter to finish out the games then that is why they needed Braxton to make those starts historically because the starter had just pitched in relief.

Yes! That's exactly right. But an AI-run team won't make that choice because the AI refuses to designate relievers as "emergency starters," even though that option is available. Making that one change in the AI logic would go a long way to correcting some of the problems that I've been talking about.

Originally Posted by Garlon

That was the manager's decision and not necessarily what provides the optimal strategy.

Well, you're talking about two different things here: historical strategies and optimal strategies. We can debate whether Huggins was better off using a starter or a "reliever" in save situations, but I don't think there's any room to disagree on whether Huggins actually did use starters in those situations. And if the ratings are supposed to produce historically accurate results, then the focus should be on historical strategies rather than optimal ones.

Originally Posted by Garlon

What I am saying is that if you make these other pitchers stronger then you are making the best pitchers more average. There is only so much room for performance in historical leagues because of the League Totals that need to be met. I agree with you though that there needs to be more pitchers with higher endurance. I really pushed Markus to move the criteria to 33% but he only agreed to 40% and I needed a mountain of evidence to show him that there was not even an average of 5 starting pitchers for some teams in modern seasons. He did not want to make the percentage a user option in the game so he coded it to 40%. If it was 33% then there would be more of the types of pitchers you want. For example, those pitchers on the 1926 Yankees you want to have more endurance would have better ratings. I think there needs to be an option in the game to use 40% of 33% as the criteria. I told Markus that a pitcher whose GS is at least 33% of their Games Pitched still is pitching more innings as a starter than as a reliever. It was actually a significant benefit moving the criteria from 50% to 40% but I do not think he will agree to give us an option in the game to adjust it further.

I agree. You can count me as a vote in favor of that option the next time you talk with Markus.