Why would the computer waive him?

[plinko83]

I'm going through my last-minute options at the trade deadline, and noticed the computer put a player on waivers with a 3.0 WAR, 4.9 projected for the whole season, who's making a reasonable (6.4 mil) amount of money. Just curious to why the AI would waive this player in any circumstance? Any thoughts?

The Tigers are in first place in their division at 52-48, fully in the mix. I can see why the AI might trade the player, but to just give them up for nothing seems weird, right?

[kidd_05_u2]

Ages in a modern MLB setting tend to trend upward and older players run out of options, so it gets increasingly hard for the AI to handle the active roster. Odd waivers like this one will always happen from time to time.

My only suggestion is to turn off the setting that limits demotions to 5 per season. I haven't tested extensively with the option turned on, but given how the AI usually works it wouldn't surprise me if computer teams chose to put a good player through waivers instead of releasing a worse player who can't be demoted.

[Paul Reuschel's Mustache]

OOTP is a fantastic game. AI roster management has a few inherent weaknesses that sometimes result in head-scratchers such as your example.

I have a house rule when it comes to the waiver wire: if I see a decent to good player on waivers that I want, then I have to trade for him. I cannot simply put in a claim and (potentially) get him for nothing, because I know that player would (probably) never be waived in the real world. So, if he's on waivers and I want him, then I have to work out a realistic trade to acquire him. It can be frustrating when the team waiving the player is unwilling to trade him or makes unrealistic trade demands. But a house rule is a house rule.

[LansdowneSt]

Let us know if he got claimed.

[plinko83]

Originally Posted by LansdowneSt

Let us know if he got claimed.

He got claimed by me, and I had the best or second best record in the league at the time, so made it through without being claimed by anyone.

I like the idea of the house rule of not being able to claim a "mistake" player on waivers, but i had already claimed this guy, so I guess he's mine for now....

[jdw31158]

I have never, ever, seen a player put on waivers by the AI get claimed by another AI team. If a player goes on waivers and I claim him, I get him 100 % of the time. This is obviously something that needs work.

[LansdowneSt]

Originally Posted by jdw31158

I have never, ever, seen a player put on waivers by the AI get claimed by another AI team. If a player goes on waivers and I claim him, I get him 100 % of the time. This is obviously something that needs work.

I've seen it. Usually when the roster needs to be trimmed on Opening Day.

Looking at the MLB transaction wire, in real life here were the number of players claimed via waivers by month:

April - 10
May - 12
June - 7
July (thru today) - 4

It'd be great if the game could produce a report that indicated how many waiver claims there were by month so the OOTP team could see if the tempo of that particular typical roster action is modeling real life. Just for use as part of the testing process if nothing else.

[md40022]

Originally Posted by jdw31158

I have never, ever, seen a player put on waivers by the AI get claimed by another AI team. If a player goes on waivers and I claim him, I get him 100 % of the time. This is obviously something that needs work.

I do see it from time to time. It's definitely rare and could probably use some tweaking but it DOES happen from on occasion.

[damientheomen3]

Even though this player is pitching very well this season, keep in mind that both his changeup and his stamina are right on the cusp of acceptable for a starting pitcher. The AI definitely shouldn't have waived this player regardless, but I'm just saying before you go offering him any long-term extensions, keep in mind that based on the ratings we can see that he's one small talent bump down (in multiple areas) away from having to be bullpen only.

[Sweed]

Originally Posted by damientheomen3

Even though this player is pitching very well this season, keep in mind that both his changeup and his stamina are right on the cusp of acceptable for a starting pitcher. The AI definitely shouldn't have waived this player regardless, but I'm just saying before you go offering him any long-term extensions, keep in mind that based on the ratings we can see that he's one small talent bump down (in multiple areas) away from having to be bullpen only.

First two things I noticed too. He's not averaging 5 innings per start, though in today's game that may not be an issue.

As far as trading, how do we know Detroit didn't try? Obviously wasn't offered to the OP, but maybe to other teams? I'm not sure what the mechanics are for AI to AI trades and, AFAIK there wouldn't be any way to know if Detroit tried to trade him or not.

Lastly we don't know what pitching Detroit still has and how those pitchers compare to the one they waived. If they brought up a young guy that can eat
innings, they're probably better off.

And as others have noted it is AI, and from time to time there are going to be questionable moves made.

[md40022]

and if we are just dumping on Dylan Lee right now, let's also add that he's 35 years old and his ERA over his last 5 starts is 5.18. Given his age, maybe it's not just a rough patch, maybe Detroit figured his best days are behind him and it's time to pull the plug.

[jcard]

Per FanGraphs, in 2022 there were nine pitchers who finished with at least 4.9 WAR.

Nine.

Nola, Rodon, Verlander, Alcantara, Gausman, Ohtani, Fried, Strider, Bieber.

The AI move was ridiculous. The rationalizations equally so.

[Sweed]

Originally Posted by jcard

Per FanGraphs, in 2022 there were nine pitchers who finished with at least 4.9 WAR.

Nine.

Nola, Rodon, Verlander, Alcantara, Gausman, Ohtani, Fried, Strider, Bieber.

The AI move was ridiculous. The rationalizations equally so.

To be fair non of those guys averaged less than 5 innings per start and the vast majority averaged 6+. That also includes the other 21 SP on that list with a WAR of 2.8+. One could argue OOTP's AI using Lee as a SP is the unrealistic decision. That 35 stamina is just plain ugly.

The OP asked for thoughts, thread title, "Why would the computer waive him?". The only way to answer the question is to guess/suggest reasons that the AI did it. Nobody defended the decision as being good or correct. If you want to make that out as rationalizing, in the context of the AI was making a good/well reasoned decision, when nobody actually did, that's on you. He got what he asked for. Nothing more.

[kidd_05_u2]

Yeah, pretty much everyone said it is a bad decision and we are trying to come up with an explanation as to what would cause the AI to make it.

I imagine some people prefer the threads where every reply is "the developers are dumb! OOTP needs competition! Markus sold his soul to COM2US!"

Thankfully, the number of those threads declines as we get further away from release date.

[rudel.dietrich]

Because the AI is dumb. Has been dumb for 20 years now and not enough players hold OOTPDs feet to the fire.

We get rationalizations and house rules.

[kidd_05_u2]

Originally Posted by rudel.dietrich

Because the AI is dumb. Has been dumb for 20 years now and not enough players hold OOTPDs feet to the fire.

We get rationalizations and house rules.

Much dumber are the people who don't understand how difficult it is to have an AI that is almost perfect at handling baseball GM and manager decisions in an infinite number of different scenarios.

The good thing is nobody is holding your feet to the fire. Some 20 years playing and thousand hours commenting a dumb game? Talk about dumb.

[dodgerblue88]

Curious, what are your AI evaluation settings? There a subset of players in every game that consistently outperform their ratings, so I would imagine if your ai eval settings are skewed too heavily towards ratings this could explain why

[Le Grande Orange]

Originally Posted by kidd_05_u2

Much dumber are the people who don't understand how difficult it is to have an AI that is almost perfect at handling baseball GM and manager decisions in an infinite number of different scenarios

No one expects perfection. What some expect is for it to not make boneheaded mistakes (or at least, to do so only rarely). It seems to me one way to improve the AI is to gather good data on how human users make their decisions on roster moves, and use that as the basis for what the AI should do.

[Pelican]

Looks to me like the AI may be too reliant on “what have you done for me lately”. Three of the last five starts shown were awful. The other two were strong; but he lost one of them anyway. And as has been said by others, he’s forcing you to use your bullpen for four innings each start.

I can imagine there would be few takers for a fading SP on the downside of his career with a bloated contract. Those are always hard deals to make. The AI needs to get creative, and cover some of that salary, to entice a team in contention that is light on starting pitching. Maybe a veteran SP will come alive in a pennant race?

If the AI wants to simply waive this guy, there had better be a young stud SP waiting in the wings. Even if that’s true, it still does make sense to simply waive the guy.

I wonder what a scout would say. Here is where IRL a good scout or pitching coach could say “the guy has lost his fastball” or “his arm is shot” or “no movement on his pitches”. Unfortunately, even a newly-updated OOTP scouting report is going to be generic and vague. You might get “has four pitches and all are below average” or something like that.

He would be far from the first SP to fall apart in his mid-thirties. Some guys fade away, and other guys lose it all at once. Verlander and Scherzer are the exceptions to this. We marvel at them because it just does not happen often.

[kidd_05_u2]

Originally Posted by Le Grande Orange

No one expects perfection. What some expect is for it to not make boneheaded mistakes (or at least, to do so only rarely). It seems to me one way to improve the AI is to gather good data on how human users make their decisions on roster moves, and use that as the basis for what the AI should do.

You can have whatever expectation you want. What is annoying is that people who are clearly absolutely clueless about what is possible and how hard it is to get improvements in AI, keep acting as if every mistake made by the AI is the result of lazy and dumb developers.

Let's take your second sentence. How would you go about gathering the data on how human users make roster decisions? How would you translate those results into code?