AI Trading Mindsets

[Henry]

Again, agreed to a point (I'm in a good mood today )

Looking back at my 5 catagories (and I'm not saying they are the best choices) the #3 or middle class would be conservative - and this is where most teams would be - maybe 50% of all teams. The other 16 teams would be either liberal toward propects (rebuilding) or liberal toward veterans (competing). By forcing this type of breakdown, you would in effect place some teams incorrectly - thus creating errors. In addition, during the off season, you could re-evaluate the teams and change direction based on their previous season's results.

Yes - all for the sake of "personality" but I think this is what some of the greater Trade complainers are looking for - some sort of "logic" in the grand plan. It may not always be good logic, but at least it's based on something rather than simply a ratings comparison and current needs (which by the way are calcualted the same for all teams regardless of their standings.

I don't know... I feel something must be done to "simulate" logical behavior - yet knowing ull well it's only "simulated" LOL

Still think it's worth a look though.

Henry

[ghulten]

Originally posted by WLight
Regarding having AI teams screw up in self-characterization, I think most of the real-life mistakes in that regard (examples given being the Mets and Pirates this year) happen early in the season or in the offseason.

Quick recollection of a midseason self-assessment mistake: Montreal acquiring Colon last year thinking they could contend. I'm sure there are numerous others.

Originally posted by WLight
I think it'd be hard to have the AI duplicate the current folly of the Mets organization spending loads of money every offseason thinking they're going to contend when we all know their squad is crap and the additions they make are generally ill-advised. But if the game is built to make an assessment each year it shouldn't be too hard.

But I think you're missing my point:
1. The Mets don't see their own folly, even though we do.
2. In order to be realistic, the AI needs to screw up just like the Mets are now, and even though the Human player will see and scream about AI deficiencies.

Originally posted by WLight
Regarding the viability of the Pirates trading prospects for vets in July, I think that would depend on what the NL Central looks like at that time.

See, you're illustrating my point exactly. You are mistaking a fluke hot start by the Pirates for being some indication that they can contend. If the hot start continues until July 1, and they're at .500 and only 5 games out, should that change their plan (if they even have one?) I don't think it should, because they are on the track to being a winning organization in the next few years, and sacrificing those propects that will make them successul to prop up a three-month mirage will only set them back. Should the AI do this occaisionally? YES! Because real-life human GMs screw this up constantly. See the Montreal example above.

I think you're all agreeing with me without agreeing with me.

[ghulten]

Originally posted by Henry
Yes - all for the sake of "personality" but I think this is what some of the greater Trade complainers are looking for - some sort of "logic" in the grand plan.

Henry

I just don't think this is realistic because I don't think most real teams have any personality or plan at all; I think they act almost schizophrenic sometimes. :-)

And I agree with the underlying need for a sense of logic; that somehow all the moves add up and are part of a "big picture master plan" - I bet real-life Met and Ranger fans feel the same way.

[Henry]

What's funny is the game already duplicates the real world, and yet we're trying to fix it ROFL !!

Henry

PS: ok... that isn't really the problem though... the ones that complain the most are the ones that feel trading is too easy, so I suspect tightening it down somewhat will fix that - then bring on those that complained it was too hard in the frst place

It all comes down to "logic"... even if the logic is flawed, some form of logic is necessary to reduce the number of complaints.

maybe - maybe not... am I beginning to repeat myself ?!?!?? LOL

[ghulten]

I'll just agree to disagree with you, Henry.

I don't think some kind of "logic" is necessary, because it order to be realistic, it has to be deliberately flawed, leading back to threads like this one where everyone complains about the poor logic.

This thread reminds me a lot of the call-in radio sports shows that you can pick up around the country sometimes. I can imagine that fans of the Brewers, Tigers, Marlins, and most other teams call in and complain about the "poor AI," "faulty decision-making," and lack of "logic" in their teams' roster strategies.

Of course, every one of us could do better, right? <wink, wink>

[Tom Martell]

There is a flaw in the trading AI. Bottom line is John Halama for Jesse Foppert makes no sense, but the AI thought it was fine. There is no reason to trade a great prospect for a poor SP/RP under any circumstances.

Furthermore, teams that are contending shouldn't be trading away major pieces of their puzzle. A team can tell if it is contending by looking at its win % and how many games back it is from 1st place/wild card. A threshold can be set that says, "you are contending". A contending team during the season hsa nothing to do with preseason or organizational mindset - it is just a point at which teh team tries to make a run at the playoffs because it is already doing well. Teams doing well don't give up midseason.

[ghulten]

Originally posted by Tom Martell
Furthermore, teams that are contending shouldn't be trading away major pieces of their puzzle. A team can tell if it is contending by looking at its win % and how many games back it is from 1st place/wild card. A threshold can be set that says, "you are contending".

I disagree. Look at Montreal last year, trading for Colon. I bet they'd like to have those prospects back now, or even two weeks after the trade. Classic case of mistaken self-evaluation at mid-season. And it hurt the team, long-term. Omar Minaya must have a "flaw in his trading AI."

[WLight]

ghulten, I think Montreal is a special case. No one knows where their organization is right now. I think a significant part of the idea behind the Colon deal arose from a sense that if they were ever going to win they had better get it done asap. I don't think they mistakenly assumed they were truly in contention, I think there was an awareness that their days were numbered. Given what we know now about that franchise's future, they may wish they had their prospects back (give a potential buyer something for their money), but at the time mortgaging their future seemed perfectly reasonable, even if their being in contention was a long shot (and as for this year, I mean, they could contend. Imagine if they had had the money to keep Colon, you're looking at Colon, Vasquez, Ohka, Armas? The way those guys are pitching right now that is a sick rotation. And their offense isn't below average, either).

And, for ctorg:

People will complain about the trade AI no matter what. Some think it's too easy, and if it is made harder, others think it's too hard. I think the above idea would make the trade AI more intelligent and more palatable to everyone on both sides. Therefore, I think it is an important step and not something that should wait.

I disagree. I don't think this idea will make the trade AI more intelligent, I think it requires the AI to be more intelligent for it to work in the first place. If the generic trade AI as it currently exists can't distinguish between a "good" and "bad" trade (and I'm not throwing my hat in the ring on this one, just for the sake of argument let's assume that it has problems, as complaints about it are widespread), then how can we expect it to distinguish between a "good for contender" trade and a "good for rebuilding" trade?

I do agree with you on the party line that the trade AI will never be perfect and there will never be complaints. I'm just saying maybe the more basic issues should be addressed before we get into adding complex new trading features. I haven't been around for many previous releases so I don't have a sense of scale on the complaints about the 5.1 trade AI, but it seems like many people aren't happy with it.

I do think this is a great idea though, for the most part, and I've always supported it whenever it comes up.

[CareyScurry]

There have been several good posts in this thread.

I think this is a fantastic idea. Questions:

1) Should the gamer to be able to set/change the mode of a computer controlled team (to be able to create, say, Mets syndrome).
2) Should the logic have any degree of randomization? (if San Diego ranks 23rd in the league and 4th in its division in "talent"and is the second youngest team and blah blah blah - then there is a 89% chance that they will pursue a "rebuilding" strategy, a 6% chance they'll go for vets, etc...
3) How frequently are strategies assessed? Changed?
4) How diverse a range of options are people interested in? Should some GMs/scouting directors be more focused on (as well as able to evaluate) pitching or hitting or vets, etc? (ghulten addressed this nicely.)
5) How dramatically does a team's plan change how they view deals? If a team is rebuilding, would that lessen their interest in a vet by 5 or 50%?
6) Can these features be optional?

I think that the lack of, uh, a plan is a substantial (if *totally* understandable) flaw in the current AI - probably that which interests me most.

[WLight]

'Mets syndrome.' Ha! I love it.

[ghulten]

WLight: I don't Montreal's special status is what made them pull the trigger on that deal. Please go back and read the archived ESPN.com articles about the Colon deal that were written last July. That deal was made because Minaya thought Montreal could win the NL East. Simple poor judgement.

EDIT: Poorly worded: I should have said "Montreal could make the playoffs." My point got lost because of the mistake, but that's life.


CareyScurry: Your questions are great. I'd love to know what the answers are for many real-life MLB teams right now, especially the one about how frequently strategies are assessed, or whether they even have a conscious strategy. I would argue that at least 2/3 of real MLB teams don't have a conscious roster-building strategy or philosophy. Why, then, are we asking OOTP AI GMs to have one?

[cfeedback]

...with the AI not being able to pick the right strategy for itself.

at least it would be a coherent strategy...something that I don't see in the game now

if some stupid AI GM decides to make a run for it, hopefully it won't be a half-assed one anymore. and maybe he'll get lucky...this is baseball after all.

[ctorg]

I am much more interested in playing a game that challenges me than I am in playing a game that accurately reflects real life baseball. I mean, I want it to reflect major league baseball to a degree, but if there were a way to allow the computer to assess its position and make trades accordingly, that would make the game more competitive and make me happier. If the game ends up being better than a real life GM at assessing itself, fine. If it's less schizophrenic, fine. Ideally, I'd like the computer to behave not so much like a real life GM, but like a real life fellow OOTP player.

I'm sure it would still make mistakes anyway. I can just imagine a team of 36-year-old declining stars trading away its future for one final run at the pennant, then tanking in the second half as its players all decline. I can just imagine a team of rookies that manages to make a run to the top of a weak division at the end of May, then trades away a bunch of those rookies for stars, only to find out that the first two months were more of a fluke than anything.

I think that trade logic is the area of the trade AI that really needs the biggest kick. Sure, there are trades that look bad in real life, but in real life, at the time of the trade, there is almost always some logic behind making the trade, even if that logic is flawed or it doesn't work out the way it was intended.

OT: This is a great thread, by the way, a shining example of how constructive ideas should be put forth in this forum, rather than one of the shouting matches that occasionally develop here. All arguments on both sides have been intelligently presented, and it's nice to see.

[Henry]

ctorg,

Personnally I agree that a "logic tree", even if flawed, is a better solution that leaving the decisions "to the wind". ghultan's position however, is noted - that there will be those that feel it still isn't "logical" becasue it makes mistakes anyway - but I feel that is a misunderstanding that an AI can't think no matter how well you program it. As far as that goes, this thread has confirmed humans don't think that well either

I'm hoping we can get some clarification from Markus on how the AI balances it's decision now - which will help us come up with a "personality" plan that might be worth looking into.

Henry

[WLight]

Originally posted by ghulten
WLight: I don't Montreal's special status is what made them pull the trigger on that deal. Please go back and read the archived ESPN.com articles about the Colon deal that were written last July. That deal was made because Minaya thought Montreal could win the NL East. Simple poor judgement.

I disagree with you. Colon went to Montreal on June 27th, 2002, when Atlanta was 48-30 and Montreal was 41-36, 7 games out of first place and just ahead of Florida at 40-38 and NYM at 39-38.

They didn't trade away Brandon Phillips to make up a 7-game deficit at the All-Star break. Nobody thought they were winning the East; at best they'd be contending for the wild card.

In fact, speaking of archived ESPN articles, here's one that says just that.

The Cleveland Indians figured they had no chance to win this season, even with ace Bartolo Colon. For the Montreal Expos, this might be their last chance.
...
"I hope it sends a message to our fans and players that we are trying to be competitive and trying to make the playoffs,'' Expos general manager Omar Minaya said.
...
The Expos, who escaped baseball's plan to eliminate them in the offseason, have been one of the majors' best success stories this year. Despite a team thrown together in spring training, the Expos are 41-36 and in the middle of the NL wild-card race in what might be their final season in Montreal -- or anywhere.

The story clearly does not imply that the Expos acquired Colon to make a run at the division title. In fact, it indicates that the opposite is true, that they made that deal out of a sense of desperation, that they had one last-ditch chance to make their presence felt. It was an all-or-nothing move, taking into account the widespread doubts about their future, not one simply based on faulty reasoning.


Even better, here's baseball genius Rob Neyer's take on the transaction, quoted in full. I'll let you draw your own conclusions.

A year ago, who'd have thunk it?

Who'd have thunk that in the middle of the summer of 2002, it would be the Expos trading prospects for a star pitcher, and that the Indians would be dumping salary and building for a future that might be two or three seasons down the road?

It's a good move for the Expos, though ... or it would be if they really had a legitimate shot at the postseason. But while one report suggested that the Expos are "in the middle of the NL wild-card race," I'm not so sure.

At this moment, the Expos trail the Giants by three games, the Diamondbacks by five games, and the Dodgers by seven games. Don't you think it's likely that one of those teams will wind up with the wild card? (The Expos also trail the Braves by 6½ games in the NL East race.)

It's also worth noting that while the Expos are five games over .500 at 41-36, they've actually been outscored on the season, with 354 runs allowed to 346 runs scored. That's a sign they may not be as good as their record indicates.

Of course, if this is the Expos' last season they don't have anything to lose, and they should trade all of their prospects in pursuit of one last bit of desperate glory. But until somebody figures out how you can kill a franchise other than the Expos, I'm not at all sure this is their last season.

[WLight]

Sorry for the OT. Please feel free to move the post if necessary, or we could take the discussion to another forum.

[BleacherBum]

I am much more interested in playing a game that challenges me...

I couldn't agree more. So often these threads become an argument over how some trade is or isn't realistic. That's missing the point for me. I feel that if the CPU's actions are not competitive, then it hinders the playability of the game.

[Henry]

I want to thank you guys for offering your opinions and keeping this an interesting thread I'm going to be busy this weekend (stripping my front lawn!!) but I'd like to try and summarize everything here in a shorter format (maybe a bullet list) and see how much of it we can agree on... maybe give everything a point system to rank things....

If we agree overwhelmingly on certain parts - we may be able to send it to Markus to consider.

Henry

[ghulten]

Henry,

I'm with you, and I think most everybody here except for you has missed my point completely, but I'm not going to belabor it any longer.

I hereby nominate you to try and bring some coherence to the ideas here (not mine) that enjoyed something akin to universal support.

Thanks,
Gordy

[Henry]

Originally posted by ghulten
Henry,

I'm with you, and I think most everybody here except for you has missed my point completely, but I'm not going to belabor it any longer.

I hereby nominate you to try and bring some coherence to the ideas here (not mine) that enjoyed something akin to universal support.

Thanks,
Gordy

LOL ... don't give up Gordy. I think your point will surface more clearly once we see who agrees with what. Some of these threads are hard to follow, some others are just hard to READ, and others are hard to stomach !! but this one has been fun so far !!

henry