[Syd Thrift]

I see people saying that the AI needs to get better, and I don't disagree; unfortunately, many times it's just "make trade more better" on the one extreme or "here is a situation where I feel like the game should not have done this one thing even though it sort of makes sense in context". Neither, I don't think, give the devs much to work with.

Instead, I'm going to lay down a few rules that I'd like to see the AI abide by that I don't think it takes into account, or at least doesn't do often enough. Please feel free to add more of your own, tell me where my rules are dumb, and so on.

- When constructing a roster, the primary backup for any given position player should not be a person who's starting somewhere else. That also means that if you're carrying, say, 4 OFers, that 4th OF should at least be able to play passable CF. IME having your starting 2B also be the 2nd string SS leads to situations where, frankly, not very good shortstops wind up playing waaaay too much.

- After like 1950 or so, rotations are pretty well set and few if any teams are using starters in relief. I'm doing a game that I just "took over" after simming out from 1946-68 and basically what seems to have happened in the 60s is that some pitchers started to get used like it was 1925 all over again, which led to some... weirdness.

- This might hamstring the AI further but I feel like it ought to make a determination in which, if it feels a given player is good enough to start, that player needs to start. If you have 2 3rd basemen who should both start, train one of them at a new position, maybe trade one, but I don't think it's historically or currently accurate to have a situation where you're sitting one of them.

- Speaking of, especially with fictional leagues I think the AI should be a bit more willing to retrain older players to play left field and first base. It's actually not that bad at that but this could definitely improve IMO.

- Conversely it feels like sometimes OF range in particular doesn't get correlated well enough with speed. Like, sure, there are times when a guy who doesn't steal much and isn't really a terror on the basepaths can become a league average CF (Dave Henderson from the 1980s comes to mind) but conversely I think it's very, very rare that a very fast player doesn't have the *range* to be a good CF. There may be other reasons why you don't want to put a Rickey Henderson or a Brian Hunter in CF - the lack of a quality arm (which I think is what kept Henderson in left) or bad, bad hands (that was Hunter's issue IIRC) - but not range.

- 2 way players were rarely if ever used from around the early 20s until... Shohei Ohtani, really, and the game should reflect that. If a player can do both, the AI ought to make a decision on whether said player with pitch or play in the field and stick to that. Yes, there were *very* occasional exceptions but even the biggest outlier of a 2-way guy - Brooks Kieschnick - pitched a grand total of 72 games in his career, all of those at the tail end of his career when he was essentially not used in the field anymore. Rick Ankiel made the transition from pitcher to outfielder but was never used as a 2 way player either. Another example I can think of, John Olerud, was never used as a pitcher in the big leagues in spite of being a lights-out starter in college. Hell, even Babe Ruth really only had like 2 seasons in which he was used as a 2-way guy beyond the occasional spot appearance. I appreciate that the game produces guys like this that you can make choices with, but, like, Ohtani is just about the only guy who's ever done both on a regular basis in the past hundred years.

- This may also make the AI less competitive but sunk cost fallacy exists and it should fall victim to it. If a guy gets signed to a 10 year, $250M contract, he should start until it is 100% crystal clear that he can't play anymore. Say what you will but there was literally a guy in the league this very season who was finally released after 5 *years* of replacement lever play as a starter, and the contract was a big part of it.

- This is a repeated gripe of mine and I'm never sure if I get the point across, but the game is waaaaay too good, like it's perfect at this, at noticing when veterans begin to lose it. Scouting can be properly inaccurate and scattershot when players are new but what I never see is the AI mis-evaluating vets. If you see a 10 year veteran have suddenly lower ratings, even on the lowest scouting settings you can be 100% sure that that player actually took a talent hit. You will never, ever see a talent drop to a vet that didn't actually happen, and you will also rarely if ever see a veteran player's talent dip without a scouting report that tells you so. What ought to happen instead, in order to reflect reality, is that it should take multiple years for scouts to accept that a vet has lost it. In the aforementioned case of Pujols I'm sure that he kept getting playing time because they had the real-life equivalent of an OOTP scouting report that said he was still a 75/80 player. You should have to decide with these older players whether your scout is right and the guy's just in a slump or if he's lost it.