Posted by barracuda3 on 11/29/2020 4:18:00 AM (view original):
Thanks for assembling (and publishing) this data! Great stuff.

You explain that "in-game" fatigue represents fatigue due to Marshall exceeding his allocated pitch count in a given game based on his RL IP/G. Is "pitch-based" fatigue simply based on his % as displayed at the beginning of each game in which he pitches, or is there something more to it than that?

Correct, on the pitch-based fatigue being his % as displayed at the beginning of each game. Pitch-based fatigue is simply the % of his allocated total pitches he’s projected to accumulate and elbirdos formula which you can access through the FAQ pinned in this forum can tell you that allocation.

In short, 90% fatigue for pitch-based fatigue is a projected total pitch count of that seasonal allocation plus 10%. So, in Marshall’s case, he’s allocated 3,474 pitches for a season. 90% fatigue would equal 3,821pitches over a season. That displayed % equals a pitchers current % as protected to finish the season with using the formula 1-(allocatedpitches-(accumulatedpitches/team games*162))/allocatedpitches). For example, if through 4 team games Marshall had thrown 92 pitches, he’d be displayed at 92/93% (not sure how the display handles rounding), as he’d be projecting for 3,726 total pitches, or 252 (7.25% more than allocated).

I mention briefly that there is also appearance fatigue, but I have a post on that elsewhere (though I still owe some data in it), but we weren’t really dealing with appearance fatigue in this league, and in cases we would have, it’s effect would’ve been similar to the pitch-based. We also now know that appearance fatigue has a pitch-based weighting, so I don’t really need to break it down separately.

I'm amazed by the effort you put into this, and I'm starting to understand more and more of it as I try to dive into it.

In this MM league, I was trying to see how many owners could push MM to his real-life innings and beyond while remaining competitive (if that was their goal, at least). I ran into appearance fatigue early because he was getting into too many games for me. I had to play with his PC settings a lot over the first quarter of the season and organize the bullpen differently when he was 100% or resting. Eventually I settled on 25/35 for him, and that got him enough innings while maintaining effectiveness. This data really helps show why going way over his max allotted PC in a game is asking for trouble.

Posted by redcped on 11/29/2020 12:50:00 PM (view original):
I'm amazed by the effort you put into this, and I'm starting to understand more and more of it as I try to dive into it.

In this MM league, I was trying to see how many owners could push MM to his real-life innings and beyond while remaining competitive (if that was their goal, at least). I ran into appearance fatigue early because he was getting into too many games for me. I had to play with his PC settings a lot over the first quarter of the season and organize the bullpen differently when he was 100% or resting. Eventually I settled on 25/35 for him, and that got him enough innings while maintaining effectiveness. This data really helps show why going way over his max allotted PC in a game is asking for trouble.

I tried really hard to push my MM to 250 IP. I rested him in game 162 to get ready for playoffs, but could've gotten another 2-4 IP out of him if I'd let him go one last game. He finished with a league-leading 235 IP, and quite middle-of-the-pack 4.29/.277/1.51 slash (to go with a 12-11-6 record). I had him set to a 40/40 PC and a 70% autorest. He pitched in 91 games, 1 at 69%, 12 at 70%, 1 at 74%, 1 at 88%, and all 76 of the rest were 94-100%. When you look at the charts from the previous page, with those PCs pushing 40 or going slightly over (54 games had him go over 40, with a high of 47 reached 4 times), he often was pitching with a combined fatigue dropping into 40-60% effective range, and was mostly in the 70-85% combined range.

Knowing what I do now, I'd handle him very differently...

just4me,

This thread has prompted me to dive into the whole pitch count issue a bit more thoroughly, especially since I'm using a bunch of high-IP 19th century guys in the about-to-start Pai Gow league, and I don't have much experience with them. I see that you wrote:

Now, from an in-game fatigue perspective, Marshall has a RL 1.97 IP/G, which, with his sim-assigned PC, would give him a max PC per game of 32.78 to stay at 100% during a game.

I see how you get the 32.78, using the pitch count formula from the first post of the classic elbirdo thread. However, reading through that thread I see that on the second page, in his 4/26/2012 post, elbirdo wrote:

I was looking over this and I would like to update my formula with what I'm using now

I multiply BFP by 3.392 add 4.444 for each BB and add 1.916 for each strikeout.

For Gagne

BFP = 306
BB = 20
SO = 137

3.392(306) + 4.444(20) + 1.916(137) = 1389.24 pitches

Looking at the numbers for my pitchers in the Pai Gow leagues, the difference between the two formulas is negligible, and doesn't ever seem to be enough to alter the way I would set pitch counts. In Marshall's instance, according to my calculations, using the "updated" elbirdo formula would yield 3,430 pitches, for an IP/G of 32.36. Again, negligible, but I'm wondering if you might know why people tend to use the "original" formula; whether it's simply a case of the initial post never having been updated or if the "updated" formula was discredited in some way.

Wow, herculean effort here. Really impressive. I love this topic because it is so misunderstood in our community and this does a great job trying to get to the bottom of it.

I'll admit I haven't had time to go through it all in detail but can share some context that may be helpful for discussion and analysis from my time playing LIVE in the early 00's. For context, LIVE used the same engine simulated games of SLB use, but a difference was that you got the actually see the Fatigue % of a pitcher change in-game.

• Pitchers are not at 100% until they hit their IP/G allocated PC. Instead, pitchers start dropping below 100% immediately. Pitchers with high IP/G in real life may stay in the upper nineties for quite some time, but definitely not 100. Pitchers with low IP/G fall below 100 after a single pitch.
• In-game fatigue is not linear. As implied by the above point, a pitcher with good IP/G may stay in the upper nineties for a while, but the closer they get to their real life IP/G equivalent pitches in the game, the faster they will fatigue. I don't have great intuition around how/when the slope increases but it was always gradual. Each 10% goes away faster than the previous one.
• In-game fatigue is based not only on IP/G but also season usage. The closer a pitcher is to their sim-allocated pitch total for the season, the quicker they will fatigue in-game. This may or may not also be related to simply how far along in a season it is (# of games played by the team) — I think this last part is unlikely though.

Posted by barracuda3 on 11/30/2020 3:31:00 AM (view original):
just4me,

This thread has prompted me to dive into the whole pitch count issue a bit more thoroughly, especially since I'm using a bunch of high-IP 19th century guys in the about-to-start Pai Gow league, and I don't have much experience with them. I see that you wrote:

Now, from an in-game fatigue perspective, Marshall has a RL 1.97 IP/G, which, with his sim-assigned PC, would give him a max PC per game of 32.78 to stay at 100% during a game.

I see how you get the 32.78, using the pitch count formula from the first post of the classic elbirdo thread. However, reading through that thread I see that on the second page, in his 4/26/2012 post, elbirdo wrote:

I was looking over this and I would like to update my formula with what I'm using now

I multiply BFP by 3.392 add 4.444 for each BB and add 1.916 for each strikeout.

For Gagne

BFP = 306
BB = 20
SO = 137

3.392(306) + 4.444(20) + 1.916(137) = 1389.24 pitches

Looking at the numbers for my pitchers in the Pai Gow leagues, the difference between the two formulas is negligible, and doesn't ever seem to be enough to alter the way I would set pitch counts. In Marshall's instance, according to my calculations, using the "updated" elbirdo formula would yield 3,430 pitches, for an IP/G of 32.36. Again, negligible, but I'm wondering if you might know why people tend to use the "original" formula; whether it's simply a case of the initial post never having been updated or if the "updated" formula was discredited in some way.

For me, it was more, I built my spreadsheet for PCs a long time ago and never even noticed the update to adjust it, so was still going off the original post.

Posted by ozomatli on 11/30/2020 4:42:00 AM (view original):
Wow, herculean effort here. Really impressive. I love this topic because it is so misunderstood in our community and this does a great job trying to get to the bottom of it.

I'll admit I haven't had time to go through it all in detail but can share some context that may be helpful for discussion and analysis from my time playing LIVE in the early 00's. For context, LIVE used the same engine simulated games of SLB use, but a difference was that you got the actually see the Fatigue % of a pitcher change in-game.

• Pitchers are not at 100% until they hit their IP/G allocated PC. Instead, pitchers start dropping below 100% immediately. Pitchers with high IP/G in real life may stay in the upper nineties for quite some time, but definitely not 100. Pitchers with low IP/G fall below 100 after a single pitch.
• In-game fatigue is not linear. As implied by the above point, a pitcher with good IP/G may stay in the upper nineties for a while, but the closer they get to their real life IP/G equivalent pitches in the game, the faster they will fatigue. I don't have great intuition around how/when the slope increases but it was always gradual. Each 10% goes away faster than the previous one.
• In-game fatigue is based not only on IP/G but also season usage. The closer a pitcher is to their sim-allocated pitch total for the season, the quicker they will fatigue in-game. This may or may not also be related to simply how far along in a season it is (# of games played by the team) — I think this last part is unlikely though.

This is excellent, and I did play a ton of live back in the day, and remember seeing the fatigue % drop, but couldn't remember at all the whens and hows of it. I do think in-game fatigue may still be linear in its displayed value just because we thought for years that pitch-based and appearance fatigue were not linear, but as we've broken those down, they've turned out to be very straightforward, though appearance fatigue does use modifier. Even the effects of fatigue, which admin says is the same across all fatigue types was assumed to be not linear, but the more I play with it, the more I'm convinced it is incredibly linear and simply applied as a multiplier on the base stats used in the decision tree (i.e., 70% fatigue = OAV*1.3). In-game fatigue is probably similarly linear, but may have a modifier or a baseline change to accelerate the effect. To your last point, appearance fatigue has a modifier based on season usage in regards to total pitches related to allocated pitches, so I wouldn't be surprised if in-game fatigue had a similar modifier used as the accelerator, but that it was still linear once that multiplier is added.

When I update the post with a link to the google sheet of data, I'll change the wording on in-game fatigue and the combined fatigue to better reflect that this is a percentage of the allocated pitches for the pitchers IP/G and not the true in-game fatigue %; which, without live-play, we may not be able to nail down exactly.

these charts still shows the relationship of in-game fatigue and performance, and to the point above, I believe shows a fairly linear effect.

Thanks for the input, Mike, great points to help better communicate this topic.

I've got the google sheet linked and added a note to the first post on in-game fatigue definitions.

The system is clearly broken. I just won a division by 2 games over an opponent who started '16 Kershaw for the last eight games of the regular season in a $255M league. He started every game at 100% and threw 45, 42, 42, 42, 46, 101, 78, and 95 pitches in those games. There is no way that should happen. In 29.1 IP, he had a 1.43 WHIP, gave up 23 runs (1 unearned), and 5 HR. He gave up 15 of those runs in the last three games which were against me, losing two out of three. So, he likely faded down the stretch. But, he should never have been able to have started that many consecutive games at 100%.

Even in the league in which a starter's pitch count has to be set at 40-40 for every game, the starter isn't at 100% the next game. How can this happen?

You can definitely get guys far enough below their IP/162 that the consecutive games penalty threshold goes away at the end of the season. Ripping off the last 8-10 games of the season at 100% is not that hard if they are super rested.

Posted by jfranco77 on 12/7/2020 2:51:00 PM (view original):
You can definitely get guys far enough below their IP/162 that the consecutive games penalty threshold goes away at the end of the season. Ripping off the last 8-10 games of the season at 100% is not that hard if they are super rested.

That should be easy to fix.

I listened to this man Mr. Franco and it paid dividends. Together in a league, I posted Why is Ed Morris $30,203,800 - IP/162 = 848.00 show fatigue when he has only used a fraction of the % allowed at any point during a season. Consecutive games penalty. Now I pitch him every other game and No issues.I don't begin to understand the mathematics of it.
Morris has 848 IP's - ( 63 GS - GP - CG ) meaning, each game he pitched had an average of 13.47 Innings, yet when you look at his IP/G as a whatif stat it is 9.22. I'm curious, but not curious enough to investigate why?
848 by 63 is 13.47 IP/G - Whatif gives Morris 9.22 IP/G - To reach 848 and average 9.22 he would have to start and complete 91.47 games
Baseball-Almanac has him at 63 GS - GP - CG and 541 Total IP's
541 by 63 = 8.59 IP/G - Makes a little more sense.
Thanks Mr. Franco, your generosity of knowledge sharing is keeping me in or close fto the playoffs every season since.

Posted by barabajagal on 12/7/2020 3:52:00 PM (view original):
I listened to this man Mr. Franco and it paid dividends. Together in a league, I posted Why is Ed Morris $30,203,800 - IP/162 = 848.00 show fatigue when he has only used a fraction of the % allowed at any point during a season. Consecutive games penalty. Now I pitch him every other game and No issues.I don't begin to understand the mathematics of it.
Morris has 848 IP's - ( 63 GS - GP - CG ) meaning, each game he pitched had an average of 13.47 Innings, yet when you look at his IP/G as a whatif stat it is 9.22. I'm curious, but not curious enough to investigate why?
848 by 63 is 13.47 IP/G - Whatif gives Morris 9.22 IP/G - To reach 848 and average 9.22 he would have to start and complete 91.47 games
Baseball-Almanac has him at 63 GS - GP - CG and 541 Total IP's
541 by 63 = 8.59 IP/G - Makes a little more sense.
Thanks Mr. Franco, your generosity of knowledge sharing is keeping me in or close fto the playoffs every season since.

I’ve got a full break down on How Appearance Fatigue (consecutive game fatigue) really works if you want to see the nitty gritty details, and to also understand how Kershaw above could go 8 straight games at 100% (my best is 21 end of season games at 100% with ‘85 Clarkson, and 162 games at 100% with ‘88 Silver King and ‘86 Dan Casey throwing them for less than 10 pitches per game).

As for Morris’ IP/G calculation, his IP are being prorated to 162 games by converting his IP by the number of games his team played. Not all teams play the same number of games every year and it creates interesting prorations. So his IP are prorated, but his GS are not in those displayed fields (there is one location where they display the prorated amount), but because of rounding the prorated GS and IP will give a different number of IP/G than his raw 541/63.

Isn't it true that when a player (pitcher or hitter) is picked up on the WW during the season, you receive the % of their IP/PA based on the % of the schedule played by your team? I thought this had been implemented several years ago. If so, it's a significant flaw that someone who is drafted on a roster at the beginning of a season and not used will have 100% of their IP/PA available regardless of when they finally start to play.

This is a really basic issue that seble's replacement could/should address quickly. It's true that it would be hard to do with '16 Kershaw in any league other than a high-cap league with lots of AAA players. But, it could certainly be done with some of the low IP studs in lower cap leagues if an owner wanted to do so.

They did implement a fix for this for the playoffs. Now when playoffs start your pitchers are prorated to having used 95% of their pitches, similar to like you called out, how acquiring a player from trade or WW works. There’s not much benefit to it during the regular season. Getting those 8 starts at the end of the season means you’ve been sitting on at least $3m, if not more, that’s a solid disadvantage at that point to not use that player all season for an 8 game stretch at the end?

I understand not liking the lack of realism, but from a gaming the sim perspective, it’s not really troubling. There’s no advantage to it.