Playing time according to real life.
 

One thing that continues to baffle me when it comes to OOTP and historical play is how the AI decides lineups and pitching rotations. For example, I started a 1901 historical replay. I selected single season replay(1 year recalc) with development off. I also chose to use real transactions. Let's look at the Tigers. The real life Tigers had 4 pitchers make the majority of the starts in 1901. They were Roscoe Miller(36 starts), Ed Siever(33), Jack Cronin(28) and Joe Yeager(25). Emil Frisk made 7 starts and Frank Owen made 5. Frisk and Owen both had identical 4.34 era's,whips above 1.60 much worse than the above listed starters. So I take a look at the OOTP Tigers and the rotation setup by the AI. It has Roscoe Miller, Ed Siever, Emil Frisk and Frank Owen in that order, with Cronin and Yeager in the bullpen. This make little to no sense at all. I can somewhat accept Frisk, as he leaves the team in July, so either Yeager or Cronin would take his place(I assume this, but Ed High with 1 whole start under his belt might get selected).

Stuff like this wouldn't bug me in a draft league, but in leagues using real transactions, it bugs the daylights out of me. I know OOTP doesn't know names, but I simply see no explanation how Frisk or Owen could be selected for the rotation based on their poor showings in real life.

I wonder if the time has come for OOTP to add a real usage meter to the game. Again, I think something like this would be hard to use in draft leagues, but when the rosters are set to real life, a device like this could come in handy.

I always figured the AI was using the best rated players. In one league I set up with various favorite historical teams that had winning records, I included the 2003 Tigers, who were 43-119, to even out the overall wins & losses. The Tigers finished over .500.

In 1901 the real Tigers finished 74-61 in 3rd place. So far they are 8 games under .500 in my replay(2 months) and both Frisk and Owen are pitching just as they did in real life, horrible.

This seems strange. What are your settings for the number of relievers vs. position players on rosters?

Is your league set up to use real stats as the basis for roles and positions? Do you have the lineups set to be determined by traditional AI rather than sabermetrics?

Also, what are the ratings for these two pitchers compared to the others, and what are your settings for AI evaluation? How much emphasis is placed on ratings vs. current year stats vs. last year's stats vs. stats from two years ago?

If you've imported your full MLB history before you created your game, then hopefully OOTP could use the previous years' stats as part of its evaluation. Otherwise, it may only use ratings if you don't have any simulated statistical history.

I did import the entire history. Settings were default for the most part. I did lower hit & run frequency and set defensive replacements to very rarely. I think the default for ratings is 30 50 15 & 5. I changed that 0 80, 15, 5 just to see if it was a ratings deal, but after running computer manager on all team, nothing changed.

I'd check the other settings I mentioned as well. And you might want to try 40/40/15/5 for the evaluation settings. Does that make any difference?

Just set up the same league again. This time Roscoe Miller is a relief pitcher. Emil Frisk and Frank Owen have moved up in the rotation. Joe Yeager is in the rotation. Cronin is once again a relief pitcher. Tried the 40/40/15/5 thing and nothing changed at all.

After trying everything I could think of, I changed things back to 30/50/15/5 and changed "base pitchers stamina on" to entire career. This gave me much better results, but Frank Owen remained in the rotation. Then I changed it to based on 3 year period and bingo, Owen was now in the pen along with Frisk and the best pitchers were in the rotation.

That shouldn't be necessary, though. How would those two scrubs have significantly better stamina than the main starters with one-year recalc? Something doesn't seem right.

I agree. I'm playing 1901. This was Owen's first year. He didn't play at all in 1902 and then had a alright 20 start season in 1903. From 1904-1906 Owen would go 64 and 41. Very good numbers all around. Frisk wouldn't pitch past July of 01.

When base stamina on is set to imported season, both Owen and Frisk have a 20 for stamina, the same as Cronin, Siever, Miller and Yeager. When I change it to based on 3 years, Owen and Frisk both drop from 20 to 3. So it's clearly stamina that is driving the bus on this issue.

I know that most people don't want to go to this much trouble but I use real transactions, real lineups, and I keep a separate spreadsheet with real pitching rotations so that I can keep the rotations and bullpens accurate. For me it is way more than worth it given the excellent outcomes that are produced. I get historical bullpen usage in some cases accurate right down the line, in terms of both games and IP. I also get accurate reproduction of seasons where pitchers shuffled back and forth between the rotation and the pen. The bottom line is this: if you keep the rotation accurate, the AI will manage the bullpen correctly. It takes extra effort but it pays off handsomely.