AI DFAs people randomly more

[Cactusguy21]

Never seen this before in playing OOTP for many years.

The AI for seattle in 2023 just put Diego Castillo on waivers in late July. He has 55/55 ratings, a low salary, and an ERA of 0.21 (!!!) in 43.2 IP (!!!), good for 2.5 rWAR and 1.2 fWAR. And the wildest thing is, nobody claimed him! I didn't claim him because it's unrealistic, but none of the other teams did either.

The Dodgers AI also DFAed Trayce Thompson when he had 1.1 WAR. You can kind of squint at their roster and see why they would get rid of a 55/55 player, but shouldn't they try and trade him first?

48/33/14/5 AI eval ratings, with hard mode and neutral favor set.

[Sweed]

Maybe the Dodgers did try to trade him? Not sure how one would know if they had or had not?

I don't play with the current roster but knowing Castillo's skill ratings and, probably just as important, salary would help put it in context. If he's overpaid or high salary for his position the AI may stay away.

Back in the day IIRC the Red Sox put Manny on waivers every year and nobody took his contract. Not saying this is the case for you here, but it can be the case in OOTP.

[Cactusguy21]

Originally Posted by Sweed

Maybe the Dodgers did try to trade him? Not sure how one would know if they had or had not?

I don't play with the current roster but knowing Castillo's skill ratings and, probably just as important, salary would help put it in context. If he's overpaid or high salary for his position the AI may stay away.

Back in the day IIRC the Red Sox put Manny on waivers every year and nobody took his contract. Not saying this is the case for you here, but it can be the case in OOTP.

1) They could have at the very least gotten some minor leaguers for him. I did commissioner mode and tried to trade him and got a few offers of 40/45 type pitching prospects.

2) Castillo's ratings are 60/65/55 and pitches are 60, 70, 60. His salary is 2.95 million, and he is arb-eligible next year at 4m estimation. He has 2 options remaining as well. If Seattle were to try and trade him at the deadline, he'd be one of the top prizes, probably would net a top 100 prospect.

[kidd_05_u2]

Originally Posted by Cactusguy21

Never seen this before in playing OOTP for many years.

The AI for seattle in 2023 just put Diego Castillo on waivers in late July. He has 55/55 ratings, a low salary, and an ERA of 0.21 (!!!) in 43.2 IP (!!!), good for 2.5 rWAR and 1.2 fWAR. And the wildest thing is, nobody claimed him! I didn't claim him because it's unrealistic, but none of the other teams did either.

The Dodgers AI also DFAed Trayce Thompson when he had 1.1 WAR. You can kind of squint at their roster and see why they would get rid of a 55/55 player, but shouldn't they try and trade him first?

48/33/14/5 AI eval ratings, with hard mode and neutral favor set.

You have to remember that the roster management AI is handling something that is extremely complex, so 'brain farts' like these are bound to happen from time to time unfortunately. I have to say that I don't see crazy DFAs like these in my saves though.

If you run a league with standard settings and are seeing great players DFAd more than occasionally, it may be worth it to make a bug report.

[Cactusguy21]

I get that, but even then the AI has a valuation model and should check that before it does a DFA

[Sweed]

Originally Posted by Cactusguy21

1) They could have at the very least gotten some minor leaguers for him. I did commissioner mode and tried to trade him and got a few offers of 40/45 type pitching prospects.

2) Castillo's ratings are 60/65/55 and pitches are 60, 70, 60. His salary is 2.95 million, and he is arb-eligible next year at 4m estimation. He has 2 options remaining as well. If Seattle were to try and trade him at the deadline, he'd be one of the top prizes, probably would net a top 100 prospect.

1. I can only wonder how the trade engine works along with user settings. Each setting "allows" x number of trades? Or the likely hood a trade will be attempted? Or the likely hood it will be accepted? IDK how the "low to high" setting changes the likely hood of trades, but they have to do something. Maybe it's something that failed a check in the trade engine?

I'm not saying you have it set on low, but that wouldn't be the point anyway. It could fail a check on normal, I think?

In any case just read your OP and, like, you trying to understand. Which means I can only speculate.

The funny(?) thing is there's a discussion going on now about how there are too many trades, and bad ones at that. So, who knows?



2.Thanks for the stats.

[Charlie Hough]

Originally Posted by kidd_05_u2

You have to remember that the roster management AI is handling something that is extremely complex, so 'brain farts' like these are bound to happen from time to time unfortunately

I haven't done enough testing in OOTP 24 to see if there has been sufficient improvement, but bad waiver transactions occurred with very high frequency in OOTP 23. At times, I had to lock players to their MLB teams to stop the AI from making repeated moves that made no sense relative to the ratings and salaries involved. Players who should have been starters were being waived in an effort to send them to the minors and call up inferior players, and many players were constantly being called up and sent down without a good reason in terms of injuries or performance.

An occasional anomaly is one thing, but this was happening throughout every season across a sizable number of transactions. I've reported it more than once, and I'm hoping this has already been addressed or will be significantly improved in the future.

[Bunktown Ballers]

Originally Posted by Sweed

Maybe the Dodgers did try to trade him? Not sure how one would know if they had or had not?

I don't play with the current roster but knowing Castillo's skill ratings and, probably just as important, salary would help put it in context. If he's overpaid or high salary for his position the AI may stay away.

Back in the day IIRC the Red Sox put Manny on waivers every year and nobody took his contract. Not saying this is the case for you here, but it can be the case in OOTP.

Pretty sure the AI is not that smart to do that when the AI can't make a competive lineup on a daily basis

[PSUColonel]

This is a big reason I have been advocating for higher ratings evaluation for the AI. I am wondering if the new default "incorporate stats into scouting reports" could be playing a role in some of the odd AI behavior we have been seeing in 24. As Charlie mentioned he had seen odd waiver moves in 23, but I can honestly say I really did not...so long as my evaluation was 65/20/10/5.

This setting is definitely new this year in that it was never checked as default since its creation a few years back. I would like to see this feature working well, but maybe it isn't? I don't know...just a thought though.

[Sweed]

Originally Posted by Bunktown Ballers

Pretty sure the AI is not that smart to do that when the AI can't make a competive lineup on a daily basis

I'm not saying OOTP puts a guy on waivers knowing his contract wouldn't be picked up. I'm saying a guy, even though a star or even very good, may not be picked up because of his contract. OOTP is good at signing players to large realistic contracts, but as soon as it does, it tends to look at them as a negative in terms of a team's budget.

Which kind of ends up being a catch 22 (in a good way). Team gets it's star player at a realistic price and though it now thinks he should maybe be traded, he can't be because he has his star contract.

[Charlie Hough]

Originally Posted by PSUColonel

As Charlie mentioned he had seen odd waiver moves in 23, but I can honestly say I really did not...so long as my evaluation was 65/20/10/5.

Thanks for mentioning this. I was using 50/30/15/10 or 55/30/15/5 in OOTP 23. But I was planning to experiment with 60/30/10/0 in OOTP 24, and maybe I will try your settings instead.

However, some of the transactions I was seeing in OOTP 23 didn't make sense in terms of the players' statistics either. But hopefully things will improve with an even greater emphasis on ratings. I might even try 80/20/0/0 because that is probably an accurate reflection of how I evaluate players as a human user. It would be interesting to see what happens in that case.

[billyray1984]

Originally Posted by Charlie Hough

Thanks for mentioning this. I was using 50/30/15/10 or 55/30/15/5 in OOTP 23. But I was planning to experiment with 60/30/10/0 in OOTP 24, and maybe I will try your settings instead.

However, some of the transactions I was seeing in OOTP 23 didn't make sense in terms of the players' statistics either. But hopefully things will improve with an even greater emphasis on ratings. I might even try 80/20/0/0 because that is probably an accurate reflection of how I evaluate players as a human user. It would be interesting to see what happens in that case.

In theory, more weight given to ratings is a good thing and in a vaccum, it is, but be ready for a lot of roster moves which might be more optimal, but are very unrealistic when compared to IRL baseball.

Players will start getting waved/sent to the minors/cut more if their ratings drop, even if their stats are good, older players will get signed in FA and then will be put on the bench in favour of younger players if their ratings drop and overall the league will be getting much younger since usually a player's rating is never as high as when he enters the league and the next few seasons. Also, players with one bad year will lose A LOT of value.

At the end of the day, all of this behaviour is good when trying to get a roster as optimal as possible when it comes to tools (and I'd argue that as a human GM playing a game, I will usually prefer to optimize tools instead of stats at all costs since stats are not predictive), but IRL that very rarely happens.

Preference is usually given to high paid vets instead of rookies, especially when said vets are popular, teams will keep younger players in the minors to manipulate their service time, teams will give players a shot even if they're been bad the last two or more years, etc.

[PSUColonel]

Originally Posted by billyray1984

In theory, more weight given to ratings is a good thing and in a vaccum, it is, but be ready for a lot of roster moves which might be more optimal, but are very unrealistic when compared to IRL baseball.

Players will start getting waved/sent to the minors/cut more if their ratings drop, even if their stats are good, older players will get signed in FA and then will be put on the bench in favour of younger players if their ratings drop and overall the league will be getting much younger since usually a player's rating is never as high as when he enters the league and the next few seasons. Also, players with one bad year will lose A LOT of value.

At the end of the day, all of this behaviour is good when trying to get a roster as optimal as possible when it comes to tools (and I'd argue that as a human GM playing a game, I will usually prefer to optimize tools instead of stats at all costs since stats are not predictive), but IRL that very rarely happens.

Preference is usually given to high paid vets instead of rookies, especially when said vets are popular, teams will keep younger players in the minors to manipulate their service time, teams will give players a shot even if they're been bad the last two or more years, etc.

Yes, it is very difficult to balance. It seems you have the choice between a better AI opponent (challenge) or realism (a lesser AI opponent)

To me, anything below 55 in the ratings department is going to yield some very odd moves. That is if you want a challenging AI experience.. It has its pitfalls too though in terms of realism. When a player is still performing well, but his ratings take a dip, the AI is going to see it (but so can the human I'd argue) and maybe waive a player or even release him even though he is being paid lots of money and hasn't really dipped in performance. (yet). So it's like you are giving the AI the old magic eye (The Natural reference) in that it can see all.

If you go with a more stats heavy equation, the AI is going to overreact to statistical performances even though players' may be rated well. It just seems there is no way to have your cake and eat it too. You sort of have to choose which type of experience you want for your saves.

So in saying that, the only two AI evaluations I consider anymore are 65/20/10/5 and 55/25/15/5. The latter is an attempt at having both worlds, but it definitely sways more toward a challenging game.

Now, 50/30/15/5 seems completely neutral, and theoretically it should be, but in all honestly doesn't take ratings into account enough to be as challenging.

[Cactusguy21]

The problem is, that if anything, having a higher input of stats should have kept Castillo on the roster, as he was the best reliever in baseball in terms of pure statistics.

[PSUColonel]

Originally Posted by Cactusguy21

The problem is, that if anything, having a higher input of stats should have kept Castillo on the roster, as he was the best reliever in baseball in terms of pure statistics.

good point.