So, uh, how's the AI?

[84CubsFan]

I realize that beta testers are under (necessary) confidentiality agreements, but can anyone talk a bit about their impressions of the new version's AI?

[battists]

Originally Posted by 84CubsFan

I realize that beta testers are under (necessary) confidentiality agreements, but can anyone talk a bit about their impressions of the new version's AI?

Some discussion here:

How the AI handling waivers?

Was there anything in particular you were interested in? "AI" is a pretty broad topic...

[Matter2003]

It's better----there were a couple of pretty big things that got fixed, still refining and improving at this point however...

[CHCfan]

My biggest AI interest right now is, How well is the returning "Team Focus" feature working? Do teams clearly act as a rebuilding/win-now team would in real life?

[Markus Heinsohn]

Originally Posted by CHCfan

My biggest AI interest right now is, How well is the returning "Team Focus" feature working? Do teams clearly act as a rebuilding/win-now team would in real life?

We're working on it

[84CubsFan]

Originally Posted by battists

Was there anything in particular you were interested in? "AI" is a pretty broad topic...

I basically want to know about two main areas where, in my opinion, the OOTP2006 AI was lacking:

1) whether the AI "knows" how to use all of the features of the game (roster management, rule 5 draft, trade negotiation, etc.)

2) whether the AI "understands" how to manage a good game of baseball (lineup/rotation management, proper substitutions, situational play calling, etc.)

I can only imagine how hard it is to code "smart" AI, so I hope my query comes across as only a genuine question about the new version and not simply criticism of the old. Thanks, Battists and Markus, for the responses. You guys are doing a great job.

[84CubsFan]

No one else wants to weigh in on this?

AI is arguably the most important deciding element in the "fun factor" of OOTP and there was an eery silence about it before v.2006 came out. I see in the new version specs that there have been "improvements," but any tester's anecdotes to tell? Is the computer a competitive rival GM and opposing manager?

[ctorg]

Originally Posted by 84CubsFan

I basically want to know about two main areas where, in my opinion, the OOTP2006 AI was lacking:

1) whether the AI "knows" how to use all of the features of the game (roster management, rule 5 draft, trade negotiation, etc.)

2) whether the AI "understands" how to manage a good game of baseball (lineup/rotation management, proper substitutions, situational play calling, etc.)

I can only imagine how hard it is to code "smart" AI, so I hope my query comes across as only a genuine question about the new version and not simply criticism of the old. Thanks, Battists and Markus, for the responses. You guys are doing a great job.

I would think you'd actually have to be more specific than that. Try to leave as little room as possible for subjectivity. What, to you, would signify that the AI "knows" or "understands" these things. Be very specific. Make lists if necessary. Otherwise you'll just get vague answers.

[Raidergoo]

AI- better in spots, a tool in spots. Lineups are often very good.

The one I liked when testing the shop player function.

I shopped a very disposable 3 star MR.

I got 12 offers from teams offering similar junk.

I got an offer of a 4 star RF, which was intriguing.

The RF was Tony Oliva, a 34 year old Tony Oliva, a .220 hitting Oliva who had not hit in 1 1/2 years, and was still under two more years of a long term contract.

His ratings were good, but the AI was set up to heavily weigh stats (current and old) over ratings.

The AI was saying, in effect, "Will no one rid me of this troublesome priest?"

[84CubsFan]

Originally Posted by ctorg

I would think you'd actually have to be more specific than that. Try to leave as little room as possible for subjectivity. What, to you, would signify that the AI "knows" or "understands" these things. Be very specific. Make lists if necessary. Otherwise you'll just get vague answers.

Fair enough. AI queries:

-- does the computer sign rookies and then release and the re-sign them again? (this has partly been answered in the other thread, but only partly)

-- does the computer give insane multi-year contracts to 38 year old veterans?

-- does the computer use the rule 5 draft and take advantage of it?

-- does the computer leave big talent "unprotected" and thus subject to the rule 5 draft?

-- does the computer recognize rebuilding/win now strategies (already asked, but a great question)?

Those are a few specific questions off the top of my head and they only apply to the general management AI. In-game AI is another matter.

Basically, though, I actually am looking for subjective impressions about how the AI feels (Matter2003's answer was exactly the sort of thing I'm interested in, for example). If it feels good to the majority, I'd love to know that; if bad, I want to know that, too. I still have a few days to pre-order, and will not buy if it looks like AI is one of the last priorities for this version as, frankly, it seems to have been for v.2006.

[battists]

Originally Posted by 84CubsFan

No one else wants to weigh in on this?

AI is arguably the most important deciding element in the "fun factor" of OOTP and there was an eery silence about it before v.2006 came out. I see in the new version specs that there have been "improvements," but any tester's anecdotes to tell? Is the computer a competitive rival GM and opposing manager?

I think there have been improvements, but this is one of the least quantifiable areas. Additionally, tester blogs and such will be released some time soon, and, with profuse apologies, I don't have enough time to go into too much detail right now.

FWIW, I have no doubt that you'd get arguments against AI being the "most important deciding element in the fun factor of OOTP".

[84CubsFan]

Originally Posted by Raidergoo

AI- better in spots, a tool in spots. Lineups are often very good.

The one I liked when testing the shop player function.

I shopped a very disposable 3 star MR.

I got 12 offers from teams offering similar junk.

I got an offer of a 4 star RF, which was intriguing.

The RF was Tony Oliva, a 34 year old Tony Oliva, a .220 hitting Oliva who had not hit in 1 1/2 years, and was still under two more years of a long term contract.

His ratings were good, but the AI was set up to heavily weigh stats (current and old) over ratings.

The AI was saying, in effect, "Will no one rid me of this troublesome priest?"

That sounds encouraging!

[84CubsFan]

Originally Posted by battists

I think there have been improvements, but this is one of the least quantifiable areas. Additionally, tester blogs and such will be released some time soon, and, with profuse apologies, I don't have enough time to go into too much detail right now.

FWIW, I have no doubt that you'd get arguments against AI being the "most important deciding element in the fun factor of OOTP".

Agreed that it's one of the least quantifiable areas (so is the concept of "fun factor").

Completely understand not having the time! No worries.

Agreed that I would get arguments against AI being the most important deciding element in the fun factor. But for solo play, I stand by my statement.

[Raidergoo]

Originally Posted by 84CubsFan

I still have a few days to pre-order, and will not buy if it looks like AI is one of the last priorities for this version as, frankly, it seems to have been for v.2006.

Please define AI. I have a good idea from your question, but want your explanation.

There is a transaction team banging away on waiver issues, a trading team bang away on trade issues, and so forth. I recently thought the AI gave away draft picks too freely. It's better. Another person thought the AI added cash too freely to trades, and I think they are happier. Is the AI more flexible keeping a free spot ( or several) on the 40 man roster? Better. As good as a human? No, not unless you compare it to any KC Royals GM over the last decade.

[BigCity]

I'm going to throw in a related item - or my 2 cents worth on the subject anyway...

I think we'll all agree that in the real world, managers and GMs are all over the ballpark (pun intended). Some make choices some of us think are genious, some think are "ok". Other feel certain GMs/managers are idiots, others think they are good.

This is the "human emotion" we are expressing here concerning the AI. Some of us feel one way, some another.

As such, we ALL will NEVER agree if the AI is "smart" or "dumb".

I still think that AI GMs and managers should have a personality rating that is hidden from us. The coding that makes these decisions should have a range from good to bad. When these AI personalities are placed, you and I would see a range of good and bad decisions - just like in real life.

You would find it almost impossible to rip off the New York GM (for example) in a trade - but regularly take advantage of Cleveland's GM (just like in real life!).

I think this environment would provide the most "fun".

[84CubsFan]

Originally Posted by Raidergoo

Please define AI. I have a good idea from your question, but want your explanation.

There is a transaction team banging away on waiver issues, a trading team bang away on trade issues, and so forth. I recently thought the AI gave away draft picks too freely. It's better. Another person thought the AI added cash too freely to trades, and I think they are happier. Is the AI more flexible keeping a free spot ( or several) on the 40 man roster? Better. As good as a human? No, not unless you compare it to any KC Royals GM over the last decade.

AI: how the computer "plays" as a non-human controller of teams. Very difficult to code an AI that can rival a human player, agreed, but not impossible -- some strategy games do this very, very well.

My assessment of v.2006 was that, probably for very logical reasons, the majority of the design energy was spent on features -- what the game could do: how many stats it kept, how many MLB rules it followed, how much customizability it had when creating leagues, the design of the interface, what the player cards looked like, what the file structure would be, etc. Definitely necessary stuff, all. But it was also apparent (to me at least) that decidedly less energy was devoted to the actual play of the game itself. The game shipped with computer managers either not using the new management features of the game correctly, or not using them at all. Only some of these problems were solved in the patches. This was and continues to be a big problem for me in terms of fun factor.

If the design team is still spending the bulk of its energy on design features and not any, or hardly any, on how the computer "plays" against human General Managers and Field Managers, then I want to know that.

Henry makes the right point that no one is going to agree on what, exactly, makes a "good" GM vs. "bad." GMs should ideally be coded with certain strategies/tendencies/idiocies, etc. My only criterion for this is that I not be forced to have to stretch the laws of believability in order to justify why the computer AI is simulating "bad" management when the real fact of the matter is that there was not enough code attention paid to it to give us (the consumers) a good game experience in the first place.

[ctorg]

The ultimate AI is one that can learn. When it makes moves that cause it to have problems, it can assess that and adjust what it does to account for the problem. Basically, if it makes a move and can determine that the net effect of the move was negative, it will stop making that move and do something else instead. It takes a while for a computer to learn in such a way, but once it does, it can be really good.

Some years from now, hopefully OOTP will employ such an AI and we'll be complaining that it's too difficult to beat and never seems to make any mistakes.

[YankeePride]

Originally Posted by BigCity

I'm going to throw in a related item - or my 2 cents worth on the subject anyway...

I think we'll all agree that in the real world, managers and GMs are all over the ballpark (pun intended). Some make choices some of us think are genious, some think are "ok". Other feel certain GMs/managers are idiots, others think they are good.

This is the "human emotion" we are expressing here concerning the AI. Some of us feel one way, some another.

As such, we ALL will NEVER agree if the AI is "smart" or "dumb".

I still think that AI GMs and managers should have a personality rating that is hidden from us. The coding that makes these decisions should have a range from good to bad. When these AI personalities are placed, you and I would see a range of good and bad decisions - just like in real life.

You would find it almost impossible to rip off the New York GM (for example) in a trade - but regularly take advantage of Cleveland's GM (just like in real life!).

I think this environment would provide the most "fun".

Agreed. But, instead of making it a hidden rating, it should correlate to their preferences (the "Prefers OBP or AVG screen" Hopefully, that makes sense).

As long as GMs are different enough to notice a pattern then I'd be satisfied. I should know what GM wants to build a veteran team just by looking at his roster. I should also know what GM prefers speedy baserunners over power hitters just by looking at his roster/draft page/minor league system/etc.

Perhaps to one person the GM that favors guys that hit .300, but doesn't care about OBP would be the idiot of the league. But, to another player, that GM could be a genius - perhaps using his limited resources to build a team of average hitters instead of spending loads of cash on three 30-40 HR guys.

That's the difference I want to see because it exists in the real world. Obviously, it isn't as drastic, but that's why there are so many potential GMs in the real word, as there are in the OOTP world.

If a certain league created in OOTP favors power hitters because of small ballparks and terrible pitching then the GM that realizes this factor will likely be successful. Whereas the GM that doesn't take advantage will field an uncompetitive team and lose his job quickly. That can all be based on their preferences, which already exist.

(on another note, I think there are way too many managers/GMs in 2006 will great ratings. It makes signing coaches uninteresting when there isn't a clear difference between one manager and another. Hopefully, this has been cleaned up a bit).

[Raidergoo]

Originally Posted by 84CubsFan

If the design team is still spending the bulk of its energy on design features and not any, or hardly any, on how the computer "plays" against human General Managers and Field Managers, then I want to know that.

I see Markus spend a lot of effort on 1) Aging and Development 2) Historicals 3) various transactation related items. 1 and 2 is now down to a minor bit of tuning. 3 is under attack.

New feature design is pretty much over if it is not coming from Andreas.

[RANGER11JP]

Originally Posted by YankeePride

Agreed. But, instead of making it a hidden rating, it should correlate to their preferences (the "Prefers OBP or AVG screen" Hopefully, that makes sense).

As long as GMs are different enough to notice a pattern then I'd be satisfied. I should know what GM wants to build a veteran team just by looking at his roster. I should also know what GM prefers speedy baserunners over power hitters just by looking at his roster/draft page/minor league system/etc.

Perhaps to one person the GM that favors guys that hit .300, but doesn't care about OBP would be the idiot of the league. But, to another player, that GM could be a genius - perhaps using his limited resources to build a team of average hitters instead of spending loads of cash on three 30-40 HR guys.

That's the difference I want to see because it exists in the real world. Obviously, it isn't as drastic, but that's why there are so many potential GMs in the real word, as there are in the OOTP world.

If a certain league created in OOTP favors power hitters because of small ballparks and terrible pitching then the GM that realizes this factor will likely be successful. Whereas the GM that doesn't take advantage will field an uncompetitive team and lose his job quickly. That can all be based on their preferences, which already exist.

(on another note, I think there are way too many managers/GMs in 2006 will great ratings. It makes signing coaches uninteresting when there isn't a clear difference between one manager and another. Hopefully, this has been cleaned up a bit).

YankeePride, I LOVE your avatar.


What we might need in all this is some kind of history factor. What I mean by this is, for example, a player who spent a lot of years with a team might NOT be traded near the end of his carreer, or maybe be prefered if he HAD be traded to fill in with his former team. Or a GM and a manager have some years together and the GM leaves or gets fired and goes elsewhere and has a greater chance of signing some former players or manager (Like George and Billy, sort of). Not to sound like a chick flick, but it is some of these relationships that make baseball interesting. You can also have negitives like managers or players not wanting to go to a team with a certain GM/manager.
Just a thought. How to do it? I'm not sure. It would require some kind of database I would think. It doesn't need to be perfect, just maybe a factor.

JP