AI to AI trade logic is awful

[Thundercrack]

Originally Posted by Lukas Berger

Right, because the Red Sox never traded Babe Ruth in his prime for a bunch of washed quad A players.

Real teams make bad trades when trading with each other. The AI should also make bad trades when trading with itself.

The idea that the AI should somehow be far, far better than real, live human GM's, with all the countless terrible trades they've made, and never do anything questionable is just kind of hard for me to get a grip on.

I don't say anyone's wrong to think that, but just speaking personally, it bothers me to see this happen in real-life far more than it does in a game, which is just mimicking real-life, for both better and worse.

This is so silly. Are you saying the AI is coded to occasionally make ridiculous trades that make no sense talentwise or even at a financial level? You have to realize that what you said makes no sense

[Sweed]

Originally Posted by Thundercrack

This is so silly. Are you saying the AI is coded to occasionally make ridiculous trades that make no sense talentwise or even at a financial level? You have to realize that what you said makes no sense

Not what I took from his post.

More like each AI GM and scout see players differently based on their skills as evaluators and what their preferences are. OOTP should not be coded to "make bad deals" for realism, and I don't believe it is. It can however make bad deals because it tries to simulate different AI staff have differing opinions that result in players being evaluated differently by each team.

[Dr Naysay]

Until the trading frequency is resolved what setting should we put AI trading at?

[Lukas Berger]

Originally Posted by Sweed

Not what I took from his post.

More like each AI GM and scout see players differently based on their skills as evaluators and what their preferences are. OOTP should not be coded to "make bad deals" for realism, and I don't believe it is. It can however make bad deals because it tries to simulate different AI staff have differing opinions that result in players being evaluated differently by each team.

Yes, this.

It's not so much the AI is coded to be bad, but that if even humans, who are much smarter than the AI at its best, still make bad trades, the AI is never not going to do this at times, in spite of our best efforts. Unless we one day get big enough to hire like 200 top level Google programmers to work on nothing but the AI for a year. And even then, there are no guarantees.

[Thundercrack]

Originally Posted by Lukas Berger

Yes, this.

It's not so much the AI is coded to be bad, but that if even humans, who are much smarter than the AI at its best, still make bad trades, the AI is never not going to do this at times, in spite of our best efforts. Unless we one day get big enough to hire like 200 top level Google programmers to work on nothing but the AI for a year. And even then, there are no guarantees.

Or, just maybe, there’s something off with the trading algorithm in 24?

[jcard]

Originally Posted by Lukas Berger

Right, because the Red Sox never traded Babe Ruth in his prime for a bunch of washed quad A players.

Real teams make bad trades when trading with each other. The AI should also make bad trades when trading with itself.

The idea that the AI should somehow be far, far better than real, live human GM's, with all the countless terrible trades they've made, and never do anything questionable is just kind of hard for me to get a grip on.

I don't say anyone's wrong to think that, but just speaking personally, it bothers me to see this happen in real-life far more than it does in a game, which is just mimicking real-life, for both better and worse.

While the underlying argument here does not even merit a rebuttal, even the Categorical Bad Trade of baseball history it cites is misrepresented. Everyone knows that there were extra-baseball considerations behind that trade. The consideration received for Ruth was not a bunch of Quad-A players, but $100-125k and a 300,000 owner-to-owner loan.

[wallewalls]

I think the issue is more so that trade frequency is too high by default, and even so on low, that there are more bad trades happening as a result.

And yes, the Cobb/Ruth comparisons are ridiculous. Those happened 100+ years ago for one, and should not be the justification for these types of deals happening in game. Even with the Archer to the Pirates deal that happened a few years ago, at the time it was probably considered an overpay, but it wasn't as egregious as it looked just a few months later. Archer completely broke down, but he wasn't broken down at the time of the deal.

[BBGiovanni]

I don't think it's ever going to be perfect, and I've always turned the frequency down to "very rare" or whatever the lowest setting is for AI trades. If I saw 71 trades in a month in a 30 team league I'd be pretty angry. And I've never seen something like that.

[Sweed]

Originally Posted by Thundercrack

Or, just maybe, there’s something off with the trading algorithm in 24?

Didn't he say words to that affect?

Originally Posted by Lukas Berger

Teams making too many trades and teams making bad trades are kind of different issues, though maybe tangentially related.

I completely agree that on the default settings teams are making too many trades. I know we've already said in a couple spots that we're looking into the amount of trades made and will rebalance things there.

I suppose part of the thing about bad trades is that if there are more trades overall, then there will also be more bad trades. Just stands to reason.

So balancing that out a bit more will likely help bring down the amount of bad trades as well, and allow a more accurate assessment of whether there are too many of those, or just too many trades in general.

[Thundercrack]

Originally Posted by Sweed

Didn't he say words to that affect?

Well, he referenced the volume, but not the egregiousness of some of the bad deals I’ve seen. Maybe turning down the volume will ultimately help the quality of the trades as well, but I don’t know that, and he didn’t address it.

[Cactusguy21]

So I was going through my trade log and noticed the Red Sox traded Jordan Walker for Daulton Varsho. I was thinking, that's a really interesting trade! Then I realized, wait, why was Walker a red sock in the first place?


turns out the Cardinals had, the day before, traded Jordan Walker for ADALBERTO MONDESI. Mondesi played 7 games all year for Boston due to injuries before being flipped for the #2 prospect in the game and Chris Stratton (a mediocre reliever who only costs 2.8m). That's outlandishly bad, never seen that before.

There's some other bad ones, which are the usual bad AI trades (like Josh Jung for two 45 POT prospects) but those aren't necessarily immersion breaking. People make bad trades all the time. But there is not a universe where Mondesi fetches the #2 prospect in the game. That is far worse on paper than the Archer trade was at the timem (obviously Archer trade didn't age well).

[Thundercrack]

Originally Posted by Cactusguy21

So I was going through my trade log and noticed the Red Sox traded Jordan Walker for Daulton Varsho. I was thinking, that's a really interesting trade! Then I realized, wait, why was Walker a red sock in the first place?


turns out the Cardinals had, the day before, traded Jordan Walker for ADALBERTO MONDESI. Mondesi played 7 games all year for Boston due to injuries before being flipped for the #2 prospect in the game and Chris Stratton (a mediocre reliever who only costs 2.8m). That's outlandishly bad, never seen that before.

There's some other bad ones, which are the usual bad AI trades (like Josh Jung for two 45 POT prospects) but those aren't necessarily immersion breaking. People make bad trades all the time. But there is not a universe where Mondesi fetches the #2 prospect in the game. That is far worse on paper than the Archer trade was at the timem (obviously Archer trade didn't age well).

Didn’t you hear, the Red Sox traded Babe Ruth for a bag of candy 104 years ago. These things happen

[Lukas Berger]

Originally Posted by Cactusguy21

There's some other bad ones, which are the usual bad AI trades (like Josh Jung for two 45 POT prospects) but those aren't necessarily immersion breaking. People make bad trades all the time. But there is not a universe where Mondesi fetches the #2 prospect in the game. That is far worse on paper than the Archer trade was at the timem (obviously Archer trade didn't age well).

I agree, but I think a big part of the problem here is that Mondesi is quite overrated in the roster set, injury proneness aside. ZIPS kind of irrationally still loves him it seems.

We'll tone that down a lot for the next patch.

[Cactusguy21]

Originally Posted by Lukas Berger

I agree, but I think a big part of the problem here is that Mondesi is quite overrated in the roster set, injury proneness aside. ZIPS kind of irrationally still loves him it seems.

We'll tone that down a lot for the next patch.

Even still, I'm not sure any team would give up the #2 prospect for a rental piece unless that player was elite. The game likes Mondesi but he's hardly an 80/80 and was injured all year.

[Lukas Berger]

Originally Posted by Cactusguy21

Even still, I'm not sure any team would give up the #2 prospect for a rental piece unless that player was elite. The game likes Mondesi but he's hardly an 80/80 and was injured all year.

Yes, that's a good point.

[Lukas Berger]

I do suspect the AI may not take injury proneness and injury history into account enough when calculating trades.

EDIT: Also, with scouting on they may in fact view Walker as a lower rated prospect than he really is. This can vary by quite a bit from team to team. No that that likely makes this a good trade in any case, but it can be another contributing factor.

[Dr Naysay]

Originally Posted by Dr Naysay

Not having a pop at anyone here per se and there are definite issues but since moving to a 90% stats only setup in last years iteration I have found out just how biased we can all be looking at ratings/potential ratings. One of my best players in last years game was a guy that I would never have stuck in the Majors had I not switched to stats. I have since seen a ton of whinging about trades on Reddit and YouTube about the AIs thinking but then looking at the trade I'm seeing a player with poor ratings whose overperfoeming those ratings at a 800+ ops throughout his Milb years which is likely why some AI trades happen. Again there are definitely issues but I think some people let their bias towards ratings seriously impact their attitude towards trading.

Just to punctuate my point I have just seen a video criticizing the AI trade engine because Seattle traded Matt Brash "whose one of the best starting pitchers in OOTP right now."

Matt Brash:


The first of two prospects deemed "not great":



The other prospect who is also an Elite Outfielder:




The rest of the deal was Ketel Marte to Seattle and Andres Munoz and Bryce Miller to Arizona. When I look at that deal Seattle traded two relievers(Munoz is elite) and a pitching prospect for a large improvement on middle infield and some prospects that could already be better than Packard and AJ Pollock both in fielding and hitting.

[sutnopia]

Originally Posted by Lukas Berger

I agree, but I think a big part of the problem here is that Mondesi is quite overrated in the roster set, injury proneness aside. ZIPS kind of irrationally still loves him it seems.

We'll tone that down a lot for the next patch.

Tone that down for the next patch means the patch available as public beta? (24.2.44)?

Thanks in advance!

[PSUColonel]

Originally Posted by Thundercrack

Well, he referenced the volume, but not the egregiousness of some of the bad deals I’ve seen. Maybe turning down the volume will ultimately help the quality of the trades as well, but I don’t know that, and he didn’t address it.

He said he wants to start by lowering the trading frequency, and then re-examine the quality of the trades. If it’s deemed things are still not working well, I’m sure there will be adjustments made.

[Sweed]

With the developers now taking a look at this I quit looking so closely at my imported league and just turned down trading to low. I play out all games so my style means I can easily keep an eye out for the time being.

One thing I did do was to compare total trades from my 22 to 23 import (first off season through ST start up) and my 23 to24. Both games were imported into the newer version right after the World Series. v22 to v23 had 56 trades in that window. v23 to v24 had 97. My thought at looking to the first off season in each was my rosters were being looked at by an AI that had been improved/changed, and there might be an adjustment time as the AI reevaluated everyone from a v23 perspective? Kind of like real life MLB rosters have a "shakeout" period in a new version of OOTP. The trades almost double in v24 compared to v23 with the same settings, take it FWIW.

Then in my second season in v23 the offseason to ST trades dropped to 29. Nothing changed in the settings. A little over a half of the previous season.

I'm not saying this is the cause in v24 and everything is "ok", but could be a contributing factor if one is only looking at an original season?

In any case I'm glad they're looking into it and have to think there will be a tweak coming in a patch.