Why are so many players put on waivers when the Offseason begins?

[bigd51]

Originally Posted by Syd Thrift

Also, and this is something that’s both hard to code and which will undoubtedly make the AI less competitive, is that it really, really needs to incorporate Bayesian priors into the reasoning. At one extreme, you’ve got a 40 year old guy who scouts say has suddenly fallen off a cliff but has given you production for the past several years, you can’t just cut him or demote him to a backup role off of a single bad report. In that case it’s probably got to be at least a couple bad reports and some poor play afield as well. If you draft a guy in the first 5 rounds, you are not going to forsake them just because your scout - often the same guy who rated them highly enough to pick - now thinks they’re a bum. No, you’ll keep them and make them prove they can’t play because you’ve already sunk a lot of draft capital and a big signing bonus into them. This is how humans think and while there are exploitable flaws, these are human flaws, and I think that this in conjunction with that “this is our team and we’re sticking with it” mentality would, if not exactly make the AI smarter, would cut down on the particular issues people see.

This is exactly the issues I have that could have safeguards in place to prevent from happening. Too many times, I'll see a team draft a 1st-round player only to release him 2-3 years later b/c TCR dropped his potential rating. Dylan Crews seems to be especially susceptible to this...

In one save, Crews was drafted 2nd overall and flat-out released by the team who drafted him after just 2 years in the minor leagues (can't remember which team it was). I was able to sign him as a minor league FA.

In my current save, he was:

- Drafted 4th ovr by the Rangers in '23

- Spent the entire '24 season in A+ w/ a wRC+ of 101, OPS+ of 105, and 1.7 WAR.

(Easily worthy of a promotion to AA for '25, no? Well... no, apparently. Because reasons.)

- Demoted, started and finished the entire '25 season back in A-ball posting a wRC+ of 111 and OPS+ of 110 and 2.5 WAR.

- Traded by the Rangers in Dec. of '25 along w/ Josh Hartle (58th ovr pick in '23 by Cleveland. Acquired by Rangers via trade in '24) to the Marlins in exchange for Andrew Nardi (65 ovr closer)

- Flat-out released by Miami just 7 months later in June of '26

- Signed as a MiL FA by Boston where he currently is.


Drafted 4th ovr and he's been traded, released, and signed as a MiL FA on his 3rd team in 4 1/2 seasons in the minors and hasn't even made it past A+ despite posting these stats:




For IRL comparison, Braden Shewmake was drafted 21st ovr in '19 by the Braves and he's still toiling around AAA on their 40-man at 25 years old.

People can try to spin it however they want to justify these things, but there is no logic behind the AI's decisions whatsoever. I still remember watching the pre-release Franchise deep-dive and finding it all too telling how poorly AI roster management is after Matt and T.J.'s reaction to Ohtani being traded to the Braves in exchange for JR Ritchie and... JR Ritchie. Even acknowledged it could be a bug and yet we still get completely unrealistic decisions devoid of any and all logic.

[MathBandit]

My experience differs. To me the fact that 1st-round picks get so so so much leash IRL is one of many examples of GMs making very flawed decisions.

If I have a prospect whose potential doesn't meet MLB standards and is 23 years old, I couldn't care less if he was a 1st overall pick from two years ago or an undrafted free agent from 5 months ago. I would cut or not cut him just the same either way if I need that roster slot.

[PoopyJoe69]

I think its fair to say, between the roster logic and the still-inflated amount of trades still occurring even after patches designed to decrease them, OOTP's roster AI is the most outwardly flawed I've seen it in years.

[PSUColonel]

Originally Posted by PoopyJoe69

I think its fair to say, between the roster logic and the still-inflated amount of trades still occurring even after patches designed to decrease them, OOTP's roster AI is the most outwardly flawed I've seen it in years.

Not to start a war o the matter, but I believe this may very well be the result of trying to fuse the engine to accommodate both PT and the single. player career mode. This is my guess however.

[Paul Reuschel's Mustache]

Under "AI Settings", do the Player Evaluation AI Settings only impact AI decisions when it comes to trading? Or do they also impact how the AI generally manages rosters?

It's very unscientific, but I've experimented with setting the AI Trading Preference all the way right to "favor prospects", while giving 100% to "Ratings Weight" in the Player Evaluation AI Settings. And it appears that the AI is almost never releasing or waiving decent to good young players anymore, which was a problem before.

Two weeks before the trading deadline, I adjust the Player Evaluation AI Settings to 55-25-15-5. After the deadline, I move it back to 100% Ratings Weight.

There must be a lot of moving parts under the hood when it comes to the AI. Looking through old AI threads going back a number of OOTP versions, AI roster management has obviously been an issue for many years. I'm trying to adjust the settings so that the AI doesn't release or waive decent to good young players while holding on to over the hill scrubs.

[PSUColonel]

Originally Posted by Paul Reuschel's Mustache

Under "AI Settings", do the Player Evaluation AI Settings only impact AI decisions when it comes to trading? Or do they also impact how the AI generally manages rosters?

It's very unscientific, but I've experimented with setting the AI Trading Preference all the way right to "favor prospects", while giving 100% to "Ratings Weight" in the Player Evaluation AI Settings. And it appears that the AI is almost never releasing or waiving decent to good young players anymore, which was a problem before.

Two weeks before the trading deadline, I adjust the Player Evaluation AI Settings to 55-25-15-5. After the deadline, I move it back to 100% Ratings Weight.

There must be a lot of moving parts under the hood when it comes to the AI. Looking through old AI threads going back a number of OOTP versions, AI roster management has obviously been an issue for many years. I'm trying to adjust the settings so that the AI doesn't release or waive decent to good young players while holding on to over the hill scrubs.

This does NOT surprise me at all. The problem is that people here WANT the more stats weighted AI evaluation to work well, but it just doesn't. The higher you go with ratings as the predominant factor in AI evaluation, the better performing the AI will be.

What a lot of people seem to forget (or just ignore) IMO is that the stats are driven from the ratings, and it's just as simple as that. Yes, there will be variances over the course of time (slumps, hot streaks, etc) but if you want the AI to do the best job it can in managing everything from rosters to contracts, to trades, RATINGS and NOT stats must rule the day.

I have been using 65/20/10/5 and this is likely too low as well. The truth is, something like 70/15/10/5 is probably even better. Or even 85/10/5/0.

[Daniel_09]

Originally Posted by PSUColonel

This does NOT surprise me at all. The problem is that people here WANT the more stats weighted AI evaluation to work well, but it just doesn't. The higher you go with ratings as the predominant factor in AI evaluation, the better performing the AI will be.

What a lot of people seem to forget (or just ignore) IMO is that the stats are driven from the ratings, and it's just as simple as that. Yes, there will be variances over the course of time (slumps, hot streaks, etc) but if you want the AI to do the best job it can in managing everything from rosters to contracts, to trades, RATINGS and NOT stats must rule the day.

I have been using 65/20/10/5 and this is likely too low as well. The truth is, something like 70/15/10/5 is probably even better. Or even 85/10/5/0.

This is absolutely true for both current and historical seasons and I have found it helpful to use 70/15/10/5 to NOT see many roster changes that could be considered insane (trades, waivers, etc.).

And remember that the stats and total numbers of each player are the reflection and results of the ratings and does not work vice versa.

[PoopyJoe69]

Originally Posted by Daniel_09

This is absolutely true for both current and historical seasons and I have found it helpful to use 70/15/10/5 to NOT see many roster changes that could be considered insane (trades, waivers, etc.).

And remember that the stats and total numbers of each player are the reflection and results of the ratings and does not work vice versa.

Does reducing AI Trade Frequency to "Low" achieve an accurate number of trades throughout the regular season and offseason?

[PSUColonel]

Originally Posted by PoopyJoe69

Does reducing AI Trade Frequency to "Low" achieve an accurate number of trades throughout the regular season and offseason?

I would leave this setting alone, and play through a season or so and then see how it's performing. You have to remember, when the roster sets are "manufactured" and not game generated, the AI is going to see a lot of things it might not like, and it will maybe take some time to sort the issues out.

I am not saying lowering the trading frequency shouldn't be done, but I am instead advocating a "wait and see" approach.

[OutS|der]

Historical with recal is a whole different game. The human will always have the advantage since we know the future.
The AI is playing the game as is if the development engine and TCR is the only thing deciding the future. Even with house rules you'll always have the advantage.

[Syd Thrift]

Originally Posted by PSUColonel

Not to start a war o the matter, but I believe this may very well be the result of trying to fuse the engine to accommodate both PT and the single. player career mode. This is my guess however.

Yes, because of all the trades the AI makes in PT…

[PSUColonel]

Originally Posted by Syd Thrift

Yes, because of all the trades the AI makes in PT…

He was also talking about roster AI…and you KNOW this…as it was said directly in his post.

[Syd Thrift]

Originally Posted by PSUColonel

He was also talking about roster AI…and you KNOW this…as it was said directly in his post.

…because of all the roster AI moves made in PT…

[PoopyJoe69]

Originally Posted by PSUColonel

I would leave this setting alone, and play through a season or so and then see how it's performing. You have to remember, when the roster sets are "manufactured" and not game generated, the AI is going to see a lot of things it might not like, and it will maybe take some time to sort the issues out.

I am not saying lowering the trading frequency shouldn't be done, but I am instead advocating a "wait and see" approach.

I ran test leagues with default settings on in both standard OOTP 24 and OOTP 23 leagues, the OOTP 24 league had 129 trades occur in the regular season whereas the OOTP 23 league only had 66 trades during the regular season. This test was with both trade frequencies on "average". When setting frequency to "low" in OOTP 24, there were 62 trades that occurred, although these all happened within the month of July.

[Sweed]

Originally Posted by PoopyJoe69

I ran test leagues with default settings on in both standard OOTP 24 and OOTP 23 leagues, the OOTP 24 league had 129 trades occur in the regular season whereas the OOTP 23 league only had 66 trades during the regular season. This test was with both trade frequencies on "average". When setting frequency to "low" in OOTP 24, there were 62 trades that occurred, although these all happened within the month of July.

Taking out the covid 2021 season four of the last five MLB seasons had the month of July with over 60 trades, with 62 being the lowest.
https://www.nbcsportswashington.com/...-years/249619/ The link breaks down, over the last 20 years, July, day before deadline, and deadline day.

Haven't spent a lot of time trying but the searches I tried didn't come up with a site that gives total season numbers. Baseball Reference lists all transactions but not total numbers (at least I couldn't get to them, maybe a pay subscription?) IDK.

Take it for what it's worth.