There was a recent discussion around this: https://www.whatifsports.com/forums/Posts.aspx?topicID=531518&TopicsTimeframe=180&TopicsPage=0

I put in two test teams with identical rosters, pitching staffs, etc... the only difference between the two teams was the roles I assigned the pitchers. On team one, I assigned the pitchers being tested as RPs and on team two, I assigned them as SPs. In both cases, they and the other pitchers on the roster had the same advanced settings.

The pitchers chosen for this tests were chosen specifically for their high IP/G. I wanted to use high IP/G pitchers that could throw a significant amount of pitches, as if there was a cap on SPs being used as RPs, that should show clearer. I chose

1966 Juan Marichal who had an IP/G of 8.31
1972 Gaylord Perry who had an IP/G of 8.36
1975 Gaylord Perry who had an IP/G of 8.36
1975 Catfish Hunter who had an IP/G of 8.41

Each of them are good on in-game fatigue as starters for ~130 pitches. I put their PC to allow up to 140 pitches to capture as much of their IP/G allotment as possible.

Given their IP/G, we should expect them to hit 0% on the in-game fatigue around 130 pitches. If there was a PC limit for SPs used as RPs, then we'd see that 0% effect starting earlier.

Team One Results (Part 1 - The Simple Stats):

Pitcher	IP	ERA	OAV	WHIP	HR/9	BB/9	K/9	Pitches	Pitches/G
72 Perry	392.33	8.53	0.346	2.06	1.45	4.75	2.48	6670	125.85
Hunter	372.67	7.56	0.320	1.87	1.26	4.54	2.49	6155	123.10
Marichal	338.33	8.43	0.349	2.04	1.81	4.47	2.71	5681	113.62
75 Perry	220.67	7.67	0.328	1.83	0.77	3.67	3.96	3557	104.62

Team One Record: 30-132 - 2nd worst team in entire league.

Team Two Results (Part 1 - The Simple Stats):

Pitcher	IP	ERA	OAV	WHIP	HR/9	BB/9	K/9	Pitches	Pitches/G
72 Perry	428.00	2.59	0.202	1.02	0.53	2.46	2.00	5842	121.71
Hunter	398.33	2.96	0.201	1.01	0.95	2.28	4.61	5400	117.39
Marichal	381.33	2.78	0.206	0.90	1.06	1.11	5.55	4946	115.02
75 Perry	221.00	3.87	0.260	1.27	0.65	2.04	6.64	3218	119.19

Team Two Record: 97-65 - NL Wildcard, 3rd best record in NL.


Added here for reference:
Team One Pitches Per PA:

72 Perry	3.35
Hunter	3.36
Marichal	3.33
75 Perry	3.33

Team Two Pitches Per PA:

72 Perry	3.35
Hunter	3.35
Marichal	3.30
75 Perry	3.37

Team One Results (Part 2 - The Detailed Stats):

Final Thoughts:

This was clear from just the first few games... the thread linked above details some stats from the first few starts of each league, and that pattern continued for all 162 games in both leagues... the final simple stats posted above should make it pretty clear alone, and over the next few days I'll have the breakdown of each team by pitches 1-100 and 100+. The difference after pitch 100 is remarkable for the team used in relief, and unnoticeable in the team used as starters.

See comment below in reply to ArelenWilliam, there may be a need to run again with a slightly different set of starters for one level of detail, but even without a clear line of sight to exact pitch counts (I estimate their PC at certain points by looking at their final PC for the game and subtracting their average pitches per PA counting PA backwards until I get to 100. Even with that slight approximation of pitch 100, This test was one of the most clear and easiest to gather data on I've ever done here at WIS. The results screamed out with every single game.

I'm a day early, thought both leagues finished.. team two finishes tomorrow. I'll finish this up then...

Thank you for doing this j4m. I had heard a long time ago from @TrentonJoe that he was told by WIS staff that when a starter us used in relief, his in-game maximum pitch count is 75% (reduced by 25%). Sitemailing TJ...

Posted by ArlenWilliam on 4/15/2022 10:48:00 PM (view original):
Thank you for doing this j4m. I had heard a long time ago from @TrentonJoe that he was told by WIS staff that when a starter us used in relief, his in-game maximum pitch count is 75% (reduced by 25%). Sitemailing TJ...

These four all had the same IP/G and similar BB and K rates, so their pitch counts per game would be similar in terms of fatigue, but even with the small differences they had, from the observations here, it's either a flat 100 PC assigned to SPs used in relief, or somewhere between a 20-25% reduction of their actual PC. To know for sure whether it's a flat 100 or the % reduction, I'd need to run again with some lower IP/G pitchers who should be good for around 100-105 pitches. If there's no noticeable decrease in performance, it's a flat 100 pitches, but if there's a drop off after pitch 75-80, then it would be clear that it's the % and not the flat 100.

Using 130/140 as the pitch count seems extreme. Not sure about others, But I rarely set pitchers over 100/110. Instead of choosing pitches with lower IP/G, why not re-run using the same pitchers and simply lower the pitch count. I guess I'm not as interested in seeing how many pitches they can throw in a single game, as I am in seeing what kind of performance drop I can expect for using SP out of the pen.

The matter I addressed stands and thank you very much, J4M. Questions abound.

Posted by just4me on 4/16/2022 7:46:00 PM (view original):

Posted by ArlenWilliam on 4/15/2022 10:48:00 PM (view original):
Thank you for doing this j4m. I had heard a long time ago from @TrentonJoe that he was told by WIS staff that when a starter us used in relief, his in-game maximum pitch count is 75% (reduced by 25%). Sitemailing TJ...

These four all had the same IP/G and similar BB and K rates, so their pitch counts per game would be similar in terms of fatigue, but even with the small differences they had, from the observations here, it's either a flat 100 PC assigned to SPs used in relief, or somewhere between a 20-25% reduction of their actual PC. To know for sure whether it's a flat 100 or the % reduction, I'd need to run again with some lower IP/G pitchers who should be good for around 100-105 pitches. If there's no noticeable decrease in performance, it's a flat 100 pitches, but if there's a drop off after pitch 75-80, then it would be clear that it's the % and not the flat 100.

While this is not definitive, I have to assume the coding is a % reduction, not a flat PC. For the simple reason that many many SP have IP/G that equate to less than 100 pitches.

If they use a flat PC, that would imply the following:
-- Some SP (those with low IP/G) will fatigue SLOWER as relievers than they will as starters
-- All starters are assumed to be able to throw exactly the same # of pitches as relievers before fatigue sets in; this would be a completely different way of applying fatigue than they use for starters - why would they code two models?
-- Or they use a flat 100 PC for high IP/G starters and a % reduction for low IP/G starters - again, why would they code two models?

Can easily test it by putting Mike Marshall on 95-100pc I think. He’d likely fatigue well before then