Matrix Games Forums

Posted: **Wed Dec 28, 2005 9:32 pm**

perhaps someone else has talked about this issue here before, I would like to stress the importance of a completely new, much more intelligent AI. I have played this game in all these years since its release, though the maps are big(which I like), OOB very accurate at the start of game, due to extremly poor AI, the game became too easy for human player.
I rememebr there are at least these following problems of AI
(1)they like to stack all in one tile, 9 formidable units in the single tile, easily rounded up by me and anniliated in 2 terms, otherwise they can establish a very good defense line

(2)they like frontal assault, even if the terrain and opposing forces are so infavorite to them, in one huge desert fox campaign (scenario from game depot site), I just put my British forces near Al Aghelia,organize a strong defense line, with formidable artilerry support, the desert fox leads all his elite panzer troops, they literally die like ants in frontal assult, they never think about diverse attack or use their artilerry , they just come , assult and die, no more fun to fight against such a stupid AI

(3) the don't use their weapons wisely, in another desert fox huge scenario, Italian mass forces attacked my British army new sollum after advancing 20 tiles, they were decimated very easily by me, later I turn the axis in manual to see how they organize their attack, to my great surprise, ALL their artillery remains 20 tiles behind their assaulting troops, staying immobile for 20 turns, I mean All the artilery units

without hugh enhancement to AI of COW, I don't think I will try the new version from SPWAW

the old "campaign series "has a much better AI, in a lot of instances I was nearly defeated by AI in CS, in COW, only once in Bulge 1944 scenario, I was defeated( a shame, defeated by AI..., but in computer +2 level)

Posted: **Wed Dec 28, 2005 10:26 pm**

ORIGINAL: panzerariete

perhaps someone else has talked about this issue here before, I would like to stress the importance of a completely new, much more intelligent AI. I have played this game in all these years since its release, though the maps are big(which I like), OOB very accurate at the start of game, due to extremly poor AI, the game became too easy for human player.
I rememebr there are at least these following problems of AI

(3) the don't use their weapons wisely, in another desert fox huge scenario, Italian mass forces attacked my British army new sollum after advancing 20 tiles, they were decimated very easily by me, later I turn the axis in manual to see how they organize their attack, to my great surprise, ALL their artillery remains 20 tiles behind their assaulting troops, staying immobile for 20 turns, I mean All the artilery units

#3 sounds like a problem with the scenario.

You're not going to hear a lot about what is and isn't in the new version. Matrix has a firm rule about not disclosing any information about what is or isn't in a new version, or what the release date is. It's too easy to be burned by giving out information, and having things not turn out as expected.

The PO is being looked at, but I don't know exactly what will or will not be done.

Posted: **Wed Dec 28, 2005 11:00 pm**

The problem with the PO is that it considers each formation in isolation, and formations only care about their own objectives.

The consequences of this are that the PO will ignore disaster in one sector as it is unable to switch units from one sector to another and if a single formation is outflanked, it will only begin to react if it loses control of one of its objectives. A breakthrough in one sector will never be reinforced from elsewhere and attacks can only proceed along a straight line.

One can fix these problems to a limited extent by manipulation of the objective chains in the editor, but there is only so much one can do. In order to get a PO which can play like some of the better armies of history, it would have to be able to consider formations not in isolation, but in reference to the entire situation, with the ability to redeploy troops as needed.

Posted: **Thu Dec 29, 2005 5:09 am**

ORIGINAL: golden delicious

The problem with the PO is that it considers each formation in isolation, and formations only care about their own objectives.

The consequences of this are that the PO will ignore disaster in one sector as it is unable to switch units from one sector to another and if a single formation is outflanked, it will only begin to react if it loses control of one of its objectives. A breakthrough in one sector will never be reinforced from elsewhere and attacks can only proceed along a straight line.

One can fix these problems to a limited extent by manipulation of the objective chains in the editor, but there is only so much one can do. In order to get a PO which can play like some of the better armies of history, it would have to be able to consider formations not in isolation, but in reference to the entire situation, with the ability to redeploy troops as needed.

In other words, it's missing a general, just like most other games, and many armies

I've got some good ideas, and depending on a whole lot of other factors, I may and may not be able to do much before the first release. There may be some easy things that can be tuned to make it play better. The fact that it does have the concept of formation AI, makes it a lot simpler than starting from scratch. The formation and unit AIs look like they should do a reasonable job.

One restriction is that for most RTS games (where much of the AI research seems to have been done,) the big responsibility for the AI is to play an entertaining game, and to present a challenge. For this game, there is an added constraint that it should also be able to play historically.

It will be a lot of fun to see how far I can take the AI.

Posted: **Thu Dec 29, 2005 9:17 am**

ORIGINAL: ralphtrick
One restriction is that for most RTS games (where much of the AI research seems to have been done,) the big responsibility for the AI is to play an entertaining game, and to present a challenge. For this game, there is an added constraint that it should also be able to play historically.

Hmm, are you now moving TOAW from being a wargame to a historical simulation? I´ve seen that discussion many a times in many forums and regarding many games, and nothing really useful has come out of it. It´s an either or thing I guess, and it´s ultimately up to the authors to decide if they want to have the players replay history or play historic battles under different, sometimes ahistoric or unrealistic assumptions

At the current state, TOAW can be used for little of both, but the historic simulation require a fair amount of house rules, and can´t really be played against the AI(?). Again, it´s another crucial decision wether to make the game a real solid human vs AI (See Highway to the Reaich and COTA for example, where a huge amount of time and effort was spent to make the AI fun to play against) and on the other hand TOAW and War in the Pacific for which the AI is more of a training ground before going up against less predictable and more skilled human players. Based on the discussion here, it seems unlikely that Matrix will go for both improving the game as such (and I´ll leave the discussion reg what needs to be improved to others [;)]) and improving the AI. On the other hand, I´m not involved so I may be completely off...

/Robert

Posted: **Thu Dec 29, 2005 2:24 pm**

ORIGINAL: ralphtrick
In other words, it's missing a general, just like most other games, and many armies

More or less. What's interesting is that the PO does a reasonable job of reflecting the Soviet way in warfare, though still far from perfect.

Glad to hear you're getting to grips with this.

Posted: **Thu Dec 29, 2005 2:27 pm**

ORIGINAL: Widell

Hmm, are you now moving TOAW from being a wargame to a historical simulation?

Ahem, from the manual;

"[TOAW] is a combined simulation and scenario editor..."

At the current state, TOAW can be used for little of both,

Yeah. If a lot of the new features are made optional, I don't see why both models cannot be refined simultaneously.

Posted: **Thu Dec 29, 2005 4:15 pm**

ORIGINAL: golden delicious

ORIGINAL: ralphtrick
In other words, it's missing a general, just like most other games, and many armies

More or less. What's interesting is that the PO does a reasonable job of reflecting the Soviet way in warfare, though still far from perfect.

Glad to hear you're getting to grips with this.

Well, maybe just the 'Soviet way' in late WW2 (43-45) with echeloned attack formations massing against objectives. The 'Soviet way' changed considerably after the death of Stalin...

I too am happy Ralph is looking at possible ways of improving the AI! (Btw, I also agree with Ben's major point above that the AI operates almost exclusively at the formation-objective level and fails to react appropriately to the overall [strategic] situation.)

Posted: **Thu Dec 29, 2005 4:49 pm**

I personally hope it's a game mainly for human-computer playing, not as human-human game, I love those huge scenarios of 200 turns(a lot in game depot site to download), I can't imagine a human-human playing by email for 200 turns.

I prefer historical OOB and historical distribution of units in the beginning of scenario, then in all the coming turns the AI can have their own hidden multiple "tactic decisions" to choose , sometimes they make frontl atatck, sometimes they attack from behind, so we have to guess each time how AI will react even if we replay the game. We gamers know histroy details well, if we know AI will follow strictly history, we can guess easily how they move, that's not fun anymore.

even the campaign date can be variated, the AI can choose to launch "battleaxe"or Alamein offensive in any random date, to give human player full surprise.

For me it should be more a "war simulation" instead of history simulation , but with initial setting all historic. We can do all the "what if " with historical units at hands of human player and AI hand.

In one word, only first turn is historical, all rest turns can be what if...

Posted: **Thu Dec 29, 2005 5:16 pm**

ORIGINAL: sstevens06

Well, maybe just the 'Soviet way' in late WW2 (43-45) with echeloned attack formations massing against objectives. The 'Soviet way' changed considerably after the death of Stalin...

Fiddlesticks. History ended in 1945....

I too am happy Ralph is looking at possible ways of improving the AI! (Btw, I also agree with Ben's major point above that the AI operates almost exclusively at the formation-objective level and fails to react appropriately to the overall [strategic] situation.)

I wonder if it's even "almost" exclusively. In theory, there's the "Independent" formation orders setting which is supposed to make a formation support other units- but does that even work?

Posted: **Thu Dec 29, 2005 5:18 pm**

ORIGINAL: panzerariete

In one word, only first turn is historical, all rest turns can be what if...

Pretty much. A lot of scenarios straightjacket the players into historical patterns through various means.

That being said, one does have to bear in mind that matters took their historical course for good reasons.

Posted: **Thu Dec 29, 2005 5:20 pm**

ORIGINAL: golden delicious
Ahem, from the manual;

"[TOAW] is a combined simulation and scenario editor..."

He, he, I knew I shouldn´t have gotten into this [;)]. It´s all a game of words. Simulation yes, it simulates operational warfare as a game:

ORIGINAL: TOAW:CoW Manual page 5
Our goal is to model these campaigns as games, with a serious effort to represent some of the real problems faced by commanders in the field

Historical Simulation, not really, specially not when talking about the AI as was the case in this discussion, unless very well "programmed" by the scenario designer. Not sure how many scenario designers that design to have the AI play the AI and repeat the historical result of a campaign vs how many try to create scenarios for, let's call it, even games, whatever that means (another can of worms there [:D])

But then again, these my own very humble and personal opinions and interpretations, so they shouldn´t of course be taken too seriously I guess

Posted: **Thu Dec 29, 2005 5:56 pm**

ORIGINAL: golden delicious

...

I wonder if it's even "almost" exclusively. In theory, there's the "Independent" formation orders setting which is supposed to make a formation support other units- but does that even work?

Beats the hell out of me - I've tried using the 'Independent' formation order for Front- (Army Group-) and Army-level assets and haven't experienced any satisfactory results.

Posted: **Sun Jan 01, 2006 11:53 am**

The AI is very poor. I doubt it will ever be excellent, because that would probably require a complete redesign of the underlying data models. But improvements should be possible without too much effort.

Disclaimer: I have a PhD in AI, and my major interest in TOAW is designing situations for AI vs AI play. I doubt that I represent a huge part of the customer base

Anyway, I think it's useful to think in terms of two major AI categories:

- "Tactical" level: how the AI controls individual units within a formation.
- "Strategic" level: how the AI coordinates the activities of different formations.

Conceptually, it's not hard to describe what an excellent AI would look like. At the "tactical" level, it would be reasonable at handling common military tasks such as: forming a defensive line; avoiding being cut-off; withdrawing to a new defensive line; counter-attacking to restore a line; etc etc etc.

TOAW is poor at all of these, IMO partly because it follows the common Jim-Dunnigan-era approach of basing the tactical level on a completely inadequate follow-the-numbers objective model. But even with the poor modelling, I think it should be possible to fix some of the worst problems: eg formations on the defensive having no concept of being cut-off, and taking no action to avoid it.

To get a really good AI, IMO you need to go beyond the objective hex model. I want to be able to paint a line of hexes and tell a formation that it is to defend that line, and paint another as a secondary defense line, to which it should withdraw if the first becomes untenable - etc. For this to work, the AI has to have some understanding of a "line", a "withdrawal", "untenable" etc etc. It requires a lot more smarts than the objective model, but on the other hand without those kind of smarts it can never be a really good AI. It's certainly doable, but it would be a lot of work of a fairly specialized (ie, expensive) nature.

TOAW pretty much completely ignores the "strategic" aspects. Obviously, the ideal is an AI "general" which can set tasks for formations and coordinate them. Now you need an AI which has some understanding of a further layer of abstractions, at the formation level; again, a lot of expensive work to do it well.

But I think there might be a couple of "strategic" things which could be added to TOAW without too much work. In particular, I'd like to see a real concept of "strategic" level reserves, along these lines:

- Ability to designate Formation A as a "reserve" for Formations X, Y, Z ...
- Under certain circumstances, the AI assigns Formation A as a reinforcement for one of X, Y, Z ... by giving it the same orders and objectives. It might be that Formation X on the attack is doing very well, and the AI acts to reinforce success. It might be that Formation Y on the defense is under great pressure and the AI decides to buttress it.
- Once the trigger which led to the assignment of Formation A as a reinforcement no longer applies, the AI returns it to "reserve" status, and it disengages & moves to the rear.
- Ideally, there would also be an event which would allow a formation to be swapped in/out of reserve status depending on its loss/supply/readiness levels, as compared to othr formations.

Obviously, this sketch would need to be fleshed out before it could be implemented, but I think it could result in something which significantly improves both the quality and the realism of the AI.

In particular, the triggers would require the AI to be able to recognize "success", "crumbling defence" etc etc. This would be non-trivial, but it has to be remembered that the goal here isn't the creation of some perfect military genius - just an AI which does stuff which is plausible.

Longer term, I hope some thought can be given to opening up the AI somehow or other to 3rd party hobbyists. It would be great to see an engine which let 3rd parties code up a "counterattack to restore line" or "perform fighting withdrawal to new line" tactical behavior, or a "general-in-chief". I can imagine competitions between different AI's, if enough people got interested.

Posted: **Sun Jan 01, 2006 6:26 pm**

ORIGINAL: Szilard

The AI is very poor. I doubt it will ever be excellent, because that would probably require a complete redesign of the underlying data models. But improvements should be possible without too much effort.

Disclaimer: I have a PhD in AI, and my major interest in TOAW is designing situations for AI vs AI play. I doubt that I represent a huge part of the customer base

Anyway, I think it's useful to think in terms of two major AI categories:

- "Tactical" level: how the AI controls individual units within a formation.
- "Strategic" level: how the AI coordinates the activities of different formations.

Conceptually, it's not hard to describe what an excellent AI would look like. At the "tactical" level, it would be reasonable at handling common military tasks such as: forming a defensive line; avoiding being cut-off; withdrawing to a new defensive line; counter-attacking to restore a line; etc etc etc.

TOAW is poor at all of these, IMO partly because it follows the common Jim-Dunnigan-era approach of basing the tactical level on a completely inadequate follow-the-numbers objective model. But even with the poor modelling, I think it should be possible to fix some of the worst problems: eg formations on the defensive having no concept of being cut-off, and taking no action to avoid it.

To get a really good AI, IMO you need to go beyond the objective hex model. I want to be able to paint a line of hexes and tell a formation that it is to defend that line, and paint another as a secondary defense line, to which it should withdraw if the first becomes untenable - etc. For this to work, the AI has to have some understanding of a "line", a "withdrawal", "untenable" etc etc. It requires a lot more smarts than the objective model, but on the other hand without those kind of smarts it can never be a really good AI. It's certainly doable, but it would be a lot of work of a fairly specialized (ie, expensive) nature.

TOAW pretty much completely ignores the "strategic" aspects. Obviously, the ideal is an AI "general" which can set tasks for formations and coordinate them. Now you need an AI which has some understanding of a further layer of abstractions, at the formation level; again, a lot of expensive work to do it well.

But I think there might be a couple of "strategic" things which could be added to TOAW without too much work. In particular, I'd like to see a real concept of "strategic" level reserves, along these lines:

- Ability to designate Formation A as a "reserve" for Formations X, Y, Z ...
- Under certain circumstances, the AI assigns Formation A as a reinforcement for one of X, Y, Z ... by giving it the same orders and objectives. It might be that Formation X on the attack is doing very well, and the AI acts to reinforce success. It might be that Formation Y on the defense is under great pressure and the AI decides to buttress it.
- Once the trigger which led to the assignment of Formation A as a reinforcement no longer applies, the AI returns it to "reserve" status, and it disengages & moves to the rear.
- Ideally, there would also be an event which would allow a formation to be swapped in/out of reserve status depending on its loss/supply/readiness levels, as compared to othr formations.

Obviously, this sketch would need to be fleshed out before it could be implemented, but I think it could result in something which significantly improves both the quality and the realism of the AI.

In particular, the triggers would require the AI to be able to recognize "success", "crumbling defence" etc etc. This would be non-trivial, but it has to be remembered that the goal here isn't the creation of some perfect military genius - just an AI which does stuff which is plausible.

Longer term, I hope some thought can be given to opening up the AI somehow or other to 3rd party hobbyists. It would be great to see an engine which let 3rd parties code up a "counterattack to restore line" or "perform fighting withdrawal to new line" tactical behavior, or a "general-in-chief". I can imagine competitions between different AI's, if enough people got interested.

I agree with most of what you say (except for the second sentence.) I believe that the AI can be made to play an excellent game for many scenarios. Since TOAW doesn't do random maps, that means that the designer can help a lot with the rough spots.

One reason that I took the role of programmer for TOAW is because I have a deep interest in Wargames and Wargaming AI (as opposed to 'real' AI. I've almost got an MS in CS, 20+ years programming experience and a stack of AI books, and I'm working my through 'elmer'(the PO code.)

I believe that the 'tactical' AI actually should play better than it does. I'm looking at why it plays so poorly, and what I can do short term to make it play better. I cant, and won't promise anything, but I've got some ideas, and I'll see how well they pan out in the time before the initial release.

The data structures aren't set up that badly for this work. I've got some redesign to do, but that's more for personal preference than anything else.

I've actually thought about allowing the designer to draw lines, or at least annotate the objectives with a width. That's way down the road, and may never happen.

I could see the potential for a game with this engine which allows you to set formations, draw lines, etc. instead of at the unit level. The AI would have to be pretty good for that to work, though. Something like Combat Mission, but at this level. I don't have the patience to play some of the huge scenarios.

The Open AI idea is also one that I like, but I don't know if it will happen. I would like to see what the Civ4 AI's toolkit looks like before I even look at doing anything along those lines.

I need to add a disclaimer that all this is pure speculation. It's possible that the new version of TOAW may not sell enough that I will want to add new features, there may be other things that can be done which are more marketable, expansion packs may not sell, I may lose interest, etc.

Ralph

P.S. You'd probably appreciate that I'm playing the AI against itself to test out the changes that I'm hoping to make.[:D]
P.P.S. If you have any AI vs. AI scenarions, I'd like to use them for testing.[&o]

Posted: **Mon Jan 02, 2006 11:32 pm**

ORIGINAL: ralphtrick

P.S. You'd probably appreciate that I'm playing the AI against itself to test out the changes that I'm hoping to make.[:D]
P.P.S. If you have any AI vs. AI scenarions, I'd like to use them for testing.[&o]

Tks, Ralph - I'll dig something out & clean it up a bit.

Another thing which would be really helpful for testing, especially larger scenarios, is the ability to let the AI play itself with screen updates turned off & news/debugging info sent to a log file. I assume that this would let an AI vs AI game play much faster. A breakpoint thing to go along with this (updating the screen & pausing) would be handy also.

Posted: **Tue Jan 03, 2006 12:50 am**

ORIGINAL: Szilard

Another thing which would be really helpful for testing, especially larger scenarios, is the ability to let the AI play itself with screen updates turned off & news/debugging info sent to a log file. I assume that this would let an AI vs AI game play much faster. A breakpoint thing to go along with this (updating the screen & pausing) would be handy also.

You should read this page;
http://www.tdg.nu/download/Unsuported%20features.htm

The noupdate switch is basically what you're describing.

Posted: **Tue Jan 03, 2006 8:12 am**

ORIGINAL: golden delicious

You should read this page;
http://www.tdg.nu/download/Unsuported%20features.htm

The noupdate switch is basically what you're describing.

Not really. The main effect of noupdate seems to be just to skip the unit move animation; you still get a screen update for the unit's final destination, combat result (as far as I recall) etc etc. What I'm talking about is something where you don't get any screen updates at all, until you hit a breakpoint (presumably, implemented by new event type).

I expect that this would speed things up quite a bit; there's still a *lot* of screen updating going on, even with noupdate.

You'd also want some way of capturing (at least) event news strings in a file, for later analysis. I can't remember whether any of the logging switches will do this for you at the moment. Probably, but you also get reams of other junk which often you won't be interested in & which would be a pain to wade through.

Posted: **Tue Jan 03, 2006 12:02 pm**

ORIGINAL: ralphtrick

It's possible that the new version of TOAW may not sell enough

Hi!

That's the key.

While I know some people have "huge" expectations for dramatic improvements, the first priority is sales. Once you get the basic game selling, then the world's your oyster and all kinds of options present themselves.

I have no idea what the actual size of the TOAW community is, hopefully that community will fully support your efforts. From there, this is an excellent game for just about any experience level player. With luck it could be a great entry level game for people into wargaming. Getting those people "in" will ultimately determine if TOAW is a success or not.

So I am hopefully that the TOAW community will take the long view. Shortsideness will only lead to ugly confrontations and mixed signals. Note, I am not saying people shouldn't voice thier opinion, but that bashing each other over what's best is a totally destructive and is in the end, an unproductive adventure. If we want a future for this game, we will have to work hard together as a community.

Ray (alias Lava)

Posted: **Tue Jan 03, 2006 4:26 pm**

ORIGINAL: Szilard

Not really. The main effect of noupdate seems to be just to skip the unit move animation; you still get a screen update for the unit's final destination, combat result (as far as I recall) etc etc. What I'm talking about is something where you don't get any screen updates at all, until you hit a breakpoint (presumably, implemented by new event type).

Just checked. You're thinking of the "nodelay" switch. I started a PO vs PO game of Balkans 12 using noupdate. No screen updates whatsoever right through the end of the first turn.

To break the flow, just click on the screen. This will give you the option to switch off the PO (as always in a PO turn).

You'd also want some way of capturing (at least) event news strings in a file, for later analysis.

That's what the 'uberdude' switch is for.

Really. Norm was very helpful when he designed this stuff.

Matrix Games Forums

will AI be improved greatly?

will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?

RE: will AI be improved greatly?