AI to AI trade logic is awful

[HRBaker]

I think I will, however, hold up on transferring my v23 leagues to v24 until I see if the "first year" AI adjustment is watered down by an update - or if we solidify on a pattern of settings that does the same.

The idea of suddenly moving players around, even if it is only for one year, unsettles me about my leagues.

[Sweed]

Originally Posted by HRBaker

I think I will, however, hold up on transferring my v23 leagues to v24 until I see if the "first year" AI adjustment is watered down by an update - or if we solidify on a pattern of settings that does the same.

The idea of suddenly moving players around, even if it is only for one year, unsettles me about my leagues.

Fair response. I'd add that dependent on how you play it may be unnecessary?
I play out all games, just finished ST, and have progressed to getting three games into my regular season. With trades on "very low" I've seen less trades than I had in v23 on "normal". Add in they'll surely have some sort of tweak out shortly, and well before my season advances anywhere near the trade deadline, I think I'll be fine.

Not trying to convince you, or anyone else, but FWIW the trades that did happen in my game were not "bad" trades. Major stars did not switch teams "willy-nilly", the same players were not traded multiple times. As a matter of fact I looked through all trades and NO player was traded more than once. I'd argue if I had known of the issue before hand and had set my trading to "very low", just after import, there would not have been enough trades in my game. If I were to import now I'd probably put it on low and check the results day to day (which is my normal MO anyway) and tweak up or down dependent on that. Anyway, just some observations from my experiences.

If you sim weeks at a time? Maybe more of a question on what to do.

[Pdubya64]

Great stuff Bears, kudos for laying it out so nicely.

I think you are spot-on, particularly with the Darvish exercise. There is no real metric to explain the value of experience in a Baseball player. Being a soft skill it's a real struggle to quantify and implement it in a game like OOTP.
But I think experience/savvy, choose your terms, is the answer to how players manage to negotiate and persevere in the face of tough times and a brutal GM "gauntlet" (for lack of a better term).

We are likely just stuck with the approximations we have- but hope springs eternal, eh?

[Bears5122]

Originally Posted by Pdubya64

Great stuff Bears, kudos for laying it out so nicely.

I think you are spot-on, particularly with the Darvish exercise. There is no real metric to explain the value of experience in a Baseball player. Being a soft skill it's a real struggle to quantify and implement it in a game like OOTP.
But I think experience/savvy, choose your terms, is the answer to how players manage to negotiate and persevere in the face of tough times and a brutal GM "gauntlet" (for lack of a better term).

We are likely just stuck with the approximations we have- but hope springs eternal, eh?

Totally agree. Experience is tough to quantify in a game. I think one areas to increase value is in team chemistry and development. Young players should rarely be seen as leaders while veterans should have a higher percent that are. Maybe something where younger players develop slightly better when playing alongside veterans (especially those who are leaders). Certain veterans may have a feature listed where they are seen as "teachers".

In regards to ratings, I think you could regress more dynamically. An aging hitter might lose speed and power, but see an increase in their eye as they age. A pitcher loses stuff but may add control or additional pitches to their arsenal.

It's minor tweaking, but I do feel like veterans are a bit undervalued in the game.

[Bears5122]

Originally Posted by Sweed

Very constructive. One could always quibble over the details but big picture it's on point. I would add that I'd like to see the AI take more note of a player's injury history before awarding that long term contract. I look at that "fragile" or "wrecked" player and think, "no way". I think the AI looks at those guy's ratings and stats and thinks "yeah" without noting he's usually missing quite a bit of time.

Same.

Byron Buxton is an example of a guy I think mirrors how injury concerns can impact contracts. Remove the injuries and that's a $200 to $300 million player. But with his injury proneness, he signed a $100 million contract instead.

[billyray1984]

Originally Posted by Bears5122

In regards to ratings, I think you could regress more dynamically. An aging hitter might lose speed and power, but see an increase in their eye as they age. A pitcher loses stuff but may add control or additional pitches to their arsenal.

It's minor tweaking, but I do feel like veterans are a bit undervalued in the game.

Using a higher tcr (120 +) will produce vets who sometimes have a resurgence in later years or develop finer tools, but when guys start to regress, they do regress fast. This is especially true for pitchers when they lose velo.

I do agree that the current system is not equipped to handle complete outliers : you'll never see Barry Bonds get walked as much as he did in 2004, Rickey Henderson never steals as much as he did IRL and you'll never get an RA Dickey or a Rick Ankiel type of story, but at the same time, these players are so rare that it's kind of a given that a simulation that is based on tryin to give an accurate representation of a baseball season as a whole will no have such extreme outliers.

[Sweed]

Lukas commented in the patch thread about the status of the "too many trades issue" in the new patch thread.
https://forums.ootpdevelopments.com/...52#post4997052

His post and my response

Originally Posted by Lukas Berger

We don't typically publish a roadmap, not least because things can just be too much in flux, especially for the first little bit after a release. Also sometimes our focus may change for one reason or another and we don't like to set potentially disappointing expectations for folks if we do change focus.

We've looked at the trading and we'll keep looking at it, but we really don't want to do anything knee-jerk here and possibly make things worse. If you've noticed some of the recent comments as folks get through multiple seasons indicate that while trading might be a bit overheated in the first season, this settles down to more of the expected level in the following seasons (which is similar to what we've found).

So it's likely that there are other factors involved here, including the AI's typical tendency to like to remake the rosters a bit in the first year.

Given that this is quite a complicated issue, with a lot of moving parts, we'll likely end up looking at it even more and running more sims over the next few weeks and come to a conclusion if there are changes needed for a future patch.

Originally Posted by Sweed

Thanks for the detailed answer on the frequency of trades issue. I have miy frequency now set to "very low" thinking there would be something coming soon in a patch. I know this is very complex, and probably easily made worse, so I do understand you guys wanting to take your time and be sure before changing any code.

With your answer here noting your tests are showing a slowing down (what I am seeing in my game**) after initial the initial surplus (for lack of a better term) of trades over time I'll go back to my original setting of "normal" and see what happens. I do play my game day to day and manage every inning of my team's games. If things go sideways I'll see it in time to not break anything. I play slow enough I could always "task manager" my way out, if need be

** keep in mind my game is played day to day so my observations are not based on a long period of time.

I understand the caution of the development team wanting to be sure before changing code. With that in mind it seems, that most of us that have advocated and changed our trade settings to very low or low, are now not in a position to see what the long term effects are. Maybe a switch back to "normal" is fine after the initial AI run through the team's roster makeup?

As I said, with this in mind, I'll be going back to "normal" and see what happens. I know on "very low" I only had two trades IIRC in ST and now a week into my regular season one, and that was three non-descript players. One, a corner OF, that could be a MLB bench player at his best and help his new team now, and the other team got two average AA pitchers in return.

I'll report back after I give "normal" a retry.

[matbar]

This trade just happened in a new league with the newest patch.

Padres trade a 45/55 prospect to the Giants for a mid reliever and $35.2 million!

[sutnopia]

Simulated three seasons with two new saves and the latest patch (for the record).

With low trade frequency...
(number of trades-what season (1-2-3))
January: X-12-18
February: X-1-1
March: 2-0-0
April: 0-1-3
May: 1-1-0
June: 5-0-0
July: 44-28-31
November: 37-15-X
December: 22-45-X

With average trade frequency...
January: X-16-8
February: X-0-0
March: 6-1-1
April: 1-2-0
May: 0-0-0
June: 5-0-1
July: 90-52-49
November: 41-9-X
December: 40-44-X

For what it's worth

Low trade frequency the first season in a new save and then back to average feels like others have said resonable.

[Pdubya64]

One oddball thing I have noticed with this year's NY Yankees; they don't value 24yr old Oswaldo Cabrera very much.

In two different starts with different teams I was able to get him in a Shop Player trade for not a whole lot in return. He is a 50ish potential, but perhaps they just have enough utility guys?

When I got him for Schoop with the Tigers Cabrera was tearing up the league starting as 2B for the first month or two before the new patch dropped. It's not completely out-of-whack, but I thought it seemed a loss of a good young talent.

[Sweed]

Originally Posted by sutnopia

Simulated three seasons with two new saves and the latest patch (for the record).

With low trade frequency...
(number of trades-what season (1-2-3))
January: X-12-18
February: X-1-1
March: 2-0-0
April: 0-1-3
May: 1-1-0
June: 5-0-0
July: 44-28-31
November: 37-15-X
December: 22-45-X

With average trade frequency...
January: X-16-8
February: X-0-0
March: 6-1-1
April: 1-2-0
May: 0-0-0
June: 5-0-1
July: 90-52-49
November: 41-9-X
December: 40-44-X

For what it's worth

Low trade frequency the first season in a new save and then back to average feels like others have said resonable.

Thank you for testing and posting your results. Seems to confirm what Lukas said they have seen with in house testing, IE

"overheated in the first season, this settles down to more of the expected level in the following seasons (which is similar to what we've found)."

.

Unlike others I am sticking with my 25-25-25-25 and after the initial rush of trades in offseason all is ok. I honestly haven't seen a difference in the quality of the players in trades compared to what I saw in v23 and before, IE only the volume has approx. doubled.

I play day to day and am back on average with the 4-25's. I will be keeping a close eye on things as late June through the trade deadline approaches. I'd have no problem changing to very low for that month in this first season if it looks to be necessary.

[Rain King]

Originally Posted by Pdubya64

One oddball thing I have noticed with this year's NY Yankees; they don't value 24yr old Oswaldo Cabrera very much.

In two different starts with different teams I was able to get him in a Shop Player trade for not a whole lot in return. He is a 50ish potential, but perhaps they just have enough utility guys?

When I got him for Schoop with the Tigers Cabrera was tearing up the league starting as 2B for the first month or two before the new patch dropped. It's not completely out-of-whack, but I thought it seemed a loss of a good young talent.

This is the problem with evaluating the Trade AI based on the real-life counter-parts.

While they try to get the players as close to real-life as they can, it isn't really possible to emulate the front-office evaluation of those players per team.

[PSUColonel]

Originally Posted by Rain King

This is the problem with evaluating the Trade AI based on the real-life counter-parts.

While they try to get the players as close to real-life as they can, it isn't really possible to emulate the front-office evaluation of those players per team.

I think the point he might be trying to make is that 50 rated guys should be worth more in OOTP. It seems the ratings have been a bit inflated however, and so that may be why the AI doesn’t seem to be valuing a 50 overall player as much as it may have in the past. I personally am not a fan of this “sliding scale” of evaluation.

[Galeg]

Originally Posted by PSUColonel

I think the point he might be trying to make is that 50 rated guys should be worth more in OOTP. It seems the ratings have been a bit inflated however, and so that may be why the AI doesn’t seem to be valuing a 50 overall player as much as it may have in the past. I personally am not a fan of this “sliding scale” of evaluation.

Personally, I think ZIPS projections have been getting worse the past few years and since OOTP uses ZIPS....

[Rain King]

Originally Posted by Galeg

Personally, I think ZIPS projections have been getting worse the past few years and since OOTP uses ZIPS....

I think the nature of the 2020 season has hurt the reliability of all projection systems for the last couple of years.

[andyhdz]

Originally Posted by Sweed

Thank you for testing and posting your results. Seems to confirm what Lukas said they have seen with in house testing, IE .

Unlike others I am sticking with my 25-25-25-25 and after the initial rush of trades in offseason all is ok. I honestly haven't seen a difference in the quality of the players in trades compared to what I saw in v23 and before, IE only the volume has approx. doubled.

I play day to day and am back on average with the 4-25's. I will be keeping a close eye on things as late June through the trade deadline approaches. I'd have no problem changing to very low for that month in this first season if it looks to be necessary.

I am a big fan of the straight 25s so let us know if you continue to see wonkiness with that going forward.

[Guthrien]

I personally think that valuation doesn't work right, and if that's true, frequency matters less (though it's obviously way overboard on normal). Were it as simple as changing ratios, they'd have much easier time adjusting it. It reminds me of how people try to fix a lot of games by just finding the perfect general settings, it sounds hopeful, but then a patch comes. IDK Sgt Mushroom thinks they probably can't patch it, at least anytime soon and he's a lot more experienced at tweaking this game than most. There are a lot of people simming seasons on video. But again, I have no desire to be negative either, I do have faith it will be fixed. I simply would like it equitable with 23 which wasn't perfect, perfection is way too high a bar for a game this complex.

[MathBandit]

Originally Posted by Guthrien

IDK Sgt Mushroom thinks they probably can't patch it, at least anytime soon and he's a lot more experienced at tweaking this game than most.

This gives me more faith that it will be fixed quickly than anything else I've seen.

[Pelican]

One thing I’ve noticed with high trade frequency is a whole lot of AI to AI trades of minor league players. The trades seem to be even, and, other than just “churning”, perhaps are filling needs or deficiencies. But you don’t see very many trades like this IRL - not at that level. The major league trades have been a bit wonky, but nothing too far fetched. I’m going to dial back the trade frequency to low, just to calm things down.

[mjfootball]

Honestly no one has any budget space or cash available to do any big trades in my first year 2023 sim so hasn't been too much in this one.