My oppinion of what the AI is good for...

[Warren Peace]

Steve:

It seems to me that you are essentially saying "The AI is as good as its going to get and that is that."

I'm kind of surprised by this as on just about all other issues you guys have constantly improved the game. I must admit it sounds kind of defeatest to me.

Based on your posts It sounds like in the early development of CMBO you spent a lot of time developing the AI, found it was a time sink with diminishing returns, and decided your time would be best spent elsewhere.

From a business perspective this sounds like a reasonable move. However, this was several years ago, and I think it would not a terrible idea to visit the AI again just to see if there might be some relatively simple ideas that might be implemented. As a scientist, I have found that sometimes revisiting issues after several years can pay off big, especially if new thinking is brought to bare. I urge you to at least spend a little time considering some of the ideas brought up here.

Finally, perhaps you could consider the idea of optional AI cheats to improve AI balance. SPecifically, how about an option where the AI plays with a different FOW level then the human player? This might help make up for some of the deficiency in the operational or statagic AI. I would think this could be easily implemented.

Warren

[Sergei]

Originally posted by russellmz:

you can try to make the human dumb.

Uh... is it just me or does that sound like "you can try to make the squirrel have a hairy tail"?

Aha, you mean make the human dumbER! Hey presto, no problemo. Just take a six-pack or two. You might want to ask a sober friend to help you in clicking the go-button after that, though.

[Haohmaru]

Originally posted by Sergei:

</font><blockquote>quote:</font><hr />Originally posted by russellmz:

you can try to make the human dumb.

Uh... is it just me or does that sound like "you can try to make the squirrel have a hairy tail"?

Aha, you mean make the human dumbER! Hey presto, no problemo. Just take a six-pack or two. You might want to ask a sober friend to help you in clicking the go-button after that, though. </font>

[Battlefront.com]

As has been the case in almost all of my debates with JasonC in the past... I am done butting heads with him. I know there is absolutely no way he will ever consider that he is wrong, or at least being unreasonable, even when he clearly can't back up his claims (i.e. he can't show me that this stuff is "easy" as he claims since he hasn't even made a wargame without AI, not to mention a good one). He is also condescending and can not advance his ideas without at the same time making it known that he feels we are lazy, interested only in "spin", incompetent, or at the very least less imaginative than he is. I don't know why I ever bother having discussions with him, but for some odd reason I try thinking that perhaps he might not be a complete arse this time round. Oh well, I'll try and learn my lesson from this latest rant of his.

Steiner14

Steve, i don't understand why you can't honor that JasonC is doing nothing else, than trying to help to make the AI better.

Because at the same time Jason is dismissing our vast experience as being nothing more than "spin". He doesn't have a clue, not even the smallest hint of a clue, what a massive and unrealistic undertaking he is asking for. Worse, he insists that it is damned simple to do. How the heck can such unrealistic expectations be productive? It can't be.

We all want the AI to be better. This is a simple and easy to grasp concept. But having someone who doesn't have a clue lecture us like we are children or mental midgets is rather difficult to swallow. No, in order for a discussion of AI to be worth anything people have to come into the discussion humble. This is not something Jason knows how to be.

What I mean by humble is to accept that AI programming, in actual fact and within the reality of commercial requirements, involves a lot of time and energy for a rapidly diminishing return on investment. Any AI can be improved with more effort, but that effort undermines other improvements that could be far more valuable to the end user. I know for sure that if we asked people if they would rather 50% the game in exchange for 125% AI vs. what they have now... the overwhelming majority would vote for the better game instead of the marginally improved AI. And those who chose the AI option would quickly change their tune once they played with it.

You guys have the luxury of being able to live in a dream world when discussing topics like this. You don't have to let reality bother you, just like Jason has demonstrated, because reality doesn't affect you. You don't have to spend months programming the stuff, dealing with customer complaints about the delays, trying to figure out how to pay bills during the extensions, or (and here is the irony) having to deal with the same level of disapointment with the AI voiced by the customers as the last game. That is why Jason can sit there and hold his ground because he has nothing to lose and has nothing to risk. He will never put his money where his mouth is and make the kind of "easy" AI he is describing, so why not just be a royal pain in our arse?

Warren,

I'm kind of surprised by this as on just about all other issues you guys have constantly improved the game. I must admit it sounds kind of defeatest to me.

No... REALISTIC. We know what can and can not be done in the real world (as opposed to inside Jason's head). We have a plan to make CMx2's AI far better than CMBO/BB/AK's. However, the improvements will be based on sound, rational, realistic goals that through experience we know have a decent chance of bearing fruit. Pursuing a Grand Experiement, as Jason suggests, is what I am saying we will never do. It is suicide. And suicide for us means that you won't have any Forum to discuss how easy it would be to make a better AI because we will be out of business.

Based on your posts It sounds like in the early development of CMBO you spent a lot of time developing the AI, found it was a time sink with diminishing returns, and decided your time would be best spent elsewhere.

From a business perspective this sounds like a reasonable move. However, this was several years ago, and I think it would not a terrible idea to visit the AI again just to see if there might be some relatively simple ideas that might be implemented. As a scientist, I have found that sometimes revisiting issues after several years can pay off big, especially if new thinking is brought to bare. I urge you to at least spend a little time considering some of the ideas brought up here.

You are of course quite correct. But there is one flaw here... and that is the current CM engine's code dicates what can and can not be done in a reasonable amount of time. Ideas don't mean anything if they can't be practically implemented. Two examples in this thread are single turn PBEM and processing turn info on individual computers instead of a single "master". Simple, easy concepts... impossible for us to implement. AI is far worse.

We have pushed the current CM's AI as far as we can. We know that, we accept that, and we can live with it. Especially since CM's AI is far superior to others out there. For the next engine, however, we can learn from our previous experience and make significant improvements that are simply not possible with the current code.

Finally, perhaps you could consider the idea of optional AI cheats to improve AI balance. SPecifically, how about an option where the AI plays with a different FOW level then the human player? This might help make up for some of the deficiency in the operational or statagic AI. I would think this could be easily implemented.

Now... this is the type of discussion I am talking about being good. It is the exact opposite of the tact Jason is taking. You have made a very practical, truely easy, suggestion. It probably could even be worked into the existing code without too much difficulty (but perhaps not). It certainly can be worked into the new engine. In fact, we have already thought of this and some other "cheats".

These "cheats" will be clearly identified to the end user and totally optional. Kinda like giving the AI 150% units. This is, in effect, a "cheat". But it is a clear and optional way to play. Another one we thought of was to allow the AI to see a Human player's orders. This would, in an offhanded way, allow the AI a sorta "intuitive" ability. For example...

Let's say you are playing a Human and you see a platoon of tanks coming through a field. What is it that you do? You try and figure out where they are going and roughly what they intend to do. Well, what if the AI got to peek at the orders? This would allow the AI to get some information for it to predict what is going on in the Human's head. Coding for something like this might (and I stress might) not too difficult, but the results might (and I stress might) make for a fundamentally more challenging AI. ANd of course it would be optional.

Our philosophy is to make a decent AI for all our games (we won't abandon AI). There are, however, things we can do that can make the AI significantlly better in some situations. The best way to do this is to allow the scenario designers certain tools to help better define what the AI should attempt to do and how it should judge its progress while doing it. This will make player designed scenarios potentially better than Quick Battles, but it is better to make improvements where possible instead of holding back in one place because it isn't possible to push forward in another.

Steve

[Steiner14]

Ok, now i begin to understand.

Improvements for the AI with predictable amount of work for Charles.

Dunno if this one is already in the list, if not you should consider about it:

Improoved AT-gun capabilities for the AI: trigger levels for unhiding, choosable by the scenario designer.

Trigger could be unit 'price', vehicle class or hit-probability, destruction probability and which flags the enemy is holding or a combination of them.

Improved tank capabilities for the AI:

Give the scenario designers the possibility for trigger-cover-arcs, like for AT-guns.

But additionally the tank doesn't need to have LOS into the cover-arc. The engine calculates the hit chance from a virtual position in LOS the tank could use, while it is hidden.

Maybe even the scenario designer could determine the path, the tank should use for his initial attack.

An additional option for the designer would be handy, that determines how long a vehicle has to be within the trigger-cover-arc or if it needs to stop, until the tank begins with his attack to prevent fast passing tanks triggering the attack but until the AI-tank has reached his LOS-position, the target is already out of LOS.

[ October 31, 2003, 05:59 PM: Message edited by: Steiner14 ]

[Captain Wacky]

My opinion on what the AI is good for...

For me to poop on!

Couldn't resist

[JasonC]

Steve on page 1 - "require 500MB of more RAM and take 15 minutes to compile turns"

Me on page 1 - "it is just time spent programming it up, debugging, and refining routines. It is the input time that takes work, not the number crunching"

Steve on page 4 - "AI programming... involves a lot of time and energy for a rapidly diminishing return on investment. Any AI can be improved with more effort"

Not a word anymore about 15 minute turns. Just as I said in my first reply to him, the bottleneck is programming time not CPU time. He has admitted as much. Since I said so from the begining of our interaction, it is not reasonable that he pretends I said the opposite.

In my last I responded to a claim that I had been insulting, by asking Steve to show me anything I had said in this thread that was insulting, and asked "insulting to whom?" He did not respond with any citation but instead repeated the unsubstantiated claim.

Here are the worst things I've said, by the least charitable possible construction, in this thread. On page 1 I said "horsefeathers" to the claim it was CPU time that was the AI hold up, rather than programming time. Even Steve is now saying the hold up is programming time. I do not see anything insulting in this, but a particularly delicate sensibility might.

On page 2 I said "CM's AI would be better if any significant improvements were made to it, ever". This was meant to be a purely factual statement. Steve as much as admitted that after spending time on the AI before the release of CMBO, they basically decided to leave it as is, since. So again it was factual. I don't see anything insulting in it. I deliberately phrased it in positive terms, of what might be better.

I said an argument that CPU power equal to that needed to beat a world chess champion would be needed to improve the AI was a strawman. A strawman argument consists in arguing against a position not actually taken by the person one is talking to, but instead a less reasonable position. In allowing that the AI could be improved with greater programming effort - and without a CPU the size of deep blue's or 15 minute turns - Steve has admitted as much. I see nothing insulting in pointing this out. It clarified the position I was taking, distinguishing it from a position like "the AI must be able to beat Kasparov".

I suggested tapping users to improve the AI, mentioning the example of Mods (BTS loves to show how far the graphics have come, rightly - with such help). In saying so I said "Instead of coding one way for the AI to do something and spinning the complaints that result, program 10 and put the knobs in the hands of players". This was a positive constructive proposal. They don't have to like it or to do it, but it does no harm to them and was clearly meant to help.

Steve has as much as admitted they did code one way for the AI to do things and have not seriously revisited it. He has referred repeatedly to complaints about it. He might consider it insulting that I called what he does here "spin", but it clearly is not responding to complaints about the AI by revisiting the AI. And pointing to non-issues like CPU time rather than the real issue, programming time, to me at least qualifies as "spin". It suggests that they have the programming time and would use it, but are prevented from doing so by technical computing power limits. Now this is not the case, as Steve has by now as much as admitted. Advancing a reason you can't do something that is not the real reason is "spin".

In his last Steve pretends that I have insisted all along on changes to the current code, as easy. What I actually said was "some time, between some version and another, it is worth doing. What you want is an AI on an improving track without further input from you all. Well, that requires some one-off work to make a flexible AI engine, but is perfectly feasible." I did not say put it in a build coming out in 2 months and basically already done. Steve says that AI improvements are planned for CMx2, which qualifies as "between some version and another". I would not be the least surprised if suggests received here, from me and from others, will make their way into such improvements. There is certainly nothing insulting in this.

The least charitable possible reading might see something insulting in my criticisms of the behavior of the AI. This is quite a stretch, however. Saying the AI is dumb is not saying BTS is dumb. When asked directly why I thought the AI wound up dumb, I said they probably wrote a path finding core routine and much of the rest was ad hoc, because of time priorities. Steve has as much as admitted something similar. There is nothing insulting in this, therefore.

The AI is dumb because they haven't had the time to make it smarter. Insults to the AI are therefore not insults to them. I certainly did say that sometimes the behavior of the AI is "imbecile", that tactics that succeed against is can be "ridiculous". A boat on a hillside, or 3 strands of wire plus a TRP at the exit of a covered route, are examples of what I was thinking of. I leave it to any CM player to judge whether those tactics are profound and whether falling for them is a mark of brilliance.

I said I wasn't trying to be fair but to improve the product, and that I didn't care about spitballs thrown at me. I will detail some of those spitballs below, for the hard of hearing. I am not aware that purposefully not taking offense at deliberate insult is itself insulting. Where I come from, it is usually considered polite, even forgiving, to overlook obviously intended offense. Someone might uncharitably read is as insulting lack of humility, I suppose. It is not, however, a dictate of either politeness or of reason that one be sensitive to boorish behavior in others.

On the issue of programming time, far from insulting them over it, when someone else noted that "CM's AI was written by a single guy just a few years ago", I said "so let's change that and get lots of people working on it". I clearly am not here denying a problem with programming time - which my first reply to Steve insisted on rather than denied - but instead suggesting a way to address it. One I have repeatedly said involves work and effort to get going, but that can tap tweaking time from outside BTS, for free, once that is done. If they don't want to do this that is their choice, obviously. But there is nothing insulting in suggesting a possible solution to the acknowledged hold-up.

On the subject of noticing easy improvements, which seems to have Steve so blue in the face, what I actually said was "many probably have. But they can't do anything about them. Only a few people inside BTS can. And their time is limited". Far from denying the hold up, I was insisting upon it. As is obviously necessary when one is suggesting a solution to that problem.

I later called "comical" the claim that I was asking for perfection or human thinking, as though only those could improve the present performance of the AI. In his last Steve says that the AI can be improved if programming time is spent doing so, just that they haven't wanted to spend that time because of its return to them, other priorities, etc. To me this amounts to an admission that is is possible to improve the present performance of the AI without making it perfect, or as smart as a human. The point of my "comical" comments was that there is plenty of room in between present AI behavior and perfection. I called the comments comical because I laughed when I read them. The notion that only a perfect AI would know not to shoot a full FO at an LMG with FOW off still strikes me so.

Probably the harshest paragraph I've written in this thread dealt with Steve's string of explanations of AI weakness, simply pointing out that they weren't sitting still. Deep blue effort can't, no it will be improved, wish we had 6 months, wish we had 2 months, it'd need 10 years and millions, others can't help because it is tweaked perfectly as is. The reality has become more apparent - it takes time and that is absolutely scarce, it does not require a deep blue effort or CPU power not yet available, others could help but it is hard even to get things to the point where they could.

Most of all, this simply shows he is sensitive on the point simply because it was a deliberate decision not to improve something that could be improved. That decision might be correct. But saying so is not the same thing as saying it cannot be done, that it'd take deep blue, yada yada. Those were hand waving arguments. The honest argument was "we'd love to, but we haven't found the time". If Steve had said that from the get-go, I would not have called his comments spin. Indeed, my first reply to him was an insistence that this was the case.

That covers everything I can find in this thread that the most sensitive person might find in any way insulting. Here is what I have received in response - without providing anything further to the above list, I might add. "an orafice in your rear area" "perfectly horrible line of irrational and unreasonable bunk" "should exit this and any future discussion of AI" "Those who whine the loadest (sic) are the ones most ignored because to pay attention to them would mean disaster" "please go somewhere else to be an ignorant pain in the arse" "since I don't know what the Hell I am talking about, they don't mean a damned" "directly proportional to your ego's presentation. Condescending posts like you are well know for dictate the "spitballs"" "your naive, uninformed, uneducated, insulting rantings are not in the least bit helpful" "there is a thing called reality. If you have never experienced it...there is no possible way you can understand how clueless you are" "you don't have the slightest clue what you are talking about" "all sorts of baseless, uninformed reasons" "Great in theory, but fails in practice like all your other ideas" "So a few egomaniacs can argue about how much better a "1.5554" is here instead of another guy's "1.673"?" "the insulting and childish tone as you have displayed here (and in many other threads I might add) is completely counter productive." "Hehe... I am going to make this my sig line" "beyond Jason's ability to understand, apparently" "I'm sure Jason can give you an answer for the age old angel question if you ask him real nice". That was on page 2 alone.

Later I got "dealing with people, like JasonC, who have can't put their money where their mouth is, yet have zero problem mouthing off. His condescending, egotistical presentation (which is very, very well known here on this Forum) makes it even worse." "The fact is that he hasn't a clue what he is talking about. Not even a hint of one. Yet he drones on and on as if he does. Has Jason ever written a computer program?" "Jason howls..." "Unfortunately Jason thinks that dreaming about something and actually doing it put us on an even footing in terms of credibility. I expect that from him" "Tell that to Jason" "we have so many more enlightened posts from Jason yet to come" "Jason obviously didn't like this coming from me, so I guess he will ignore your comments as well. Afterall, informed comments are easily discarded by him." "Jason would call Charles lazy or unimaginative because he couldn't do the Theoretical Solution. I'd LOVE the chance to sit down and examine Jason's work and evaluate it in the same way. I'm sure I could tear him to pieces no matter what his job is"

"If you are nearly blind it is easiy to understand how you can be so off the mark. And that is why I understand your point of view; because I understand that you don't." "Everything you said above is 100% bunk" "What I am saying is not "spin", but fact. What you are saying is uninformed, naive, and (as is usual with you) condescending horsecrap." (Should I make that my sig line?) "Typical Jason type rants about how he knows everything with no experience and we know nothing with tons experience" "you clearly don't have a clue what you are talking about" "All the daydreaming you care to sputter out on this Forum (laced with your trademark attitude)" "hold Charles in high regard and pee on his efforts without seeing a contradiction. A man of contradictions." (Confusing himself with Charles?) "a total divorce from the reality of programming" "being a pompous, deaf, condescending git in order to get attention and make yourself feel important" "when you drop the attitude your ideas are discussed without all the extra baggage. I pride myself in answering posts in the spirt of they were posted in." "if the person/s making the suggestions are clueless AND insulting AND unreasonable... that pretty much destroys any chance of having a productive conversation "am done butting heads with him. I know there is absolutely no way he will ever consider that he is wrong, or at least being unreasonable" "He is also condescending and can not advance his ideas without at the same time making it known that he feels we are lazy, interested only in "spin", incompetent, or at the very least less imaginative than he is. I don't know why I ever bother having discussions with him, but for some odd reason I try thinking that perhaps he might not be a complete arse this time round. Oh well, I'll try and learn my lesson from this latest rant" "having someone who doesn't have a clue lecture us like we are children or mental midgets is rather difficult to swallow." "so why not just be a royal pain in our arse?"

Is there something in your eye? A beam, perhaps?

[Kellysheroes]

I'd really like to see a poll on how many people that play the CM series find the AI unchallenging with all the settings they are given to make it challenging.

How about inputing a "card" in your next version asking the customer after say 30 days of play to evaluate the AI. Hopefully everyone will send the card in and then you can tell us what the moral majority feel about the AI.

I for one find the AI a pretty good challenge, but, I don't put hundreds of hours into playing this game either. It's one of those every once in awhile type of games, since it is just a game and not a lifestyle.

I've been happy with the series so far and look forward to the Africa Korp version soon.

Many people talk about "multiplayer" and all this, but, I've been trying for almost a week to get an online multiplayer game going and nothing yet. I don't think multiplayer is as popular as many people make it out to be. Games like Everquest and Ultima and the rest do not clearly represent the total market of multiplayer "types", while there might be a large market for fantasy multiplayer, the wargaming market has never been what one would call "massive". I go back to the mid/late 60's in wargaming and even then finding four people to play EtoE was an accomplishment.

Keep up the good work, at least you have "one" happy customer! heh

[massimorocca]

Defintive solution.

1) BFC recruit JasonC with no pay.

2) Give him the old CM code ( it's now obsolete with the new developing)

3) BFC and JasonC submitted to the CM community a question. What is the most important improvement you will see in the AI behavior and how many months you will wait to have it.

4) JasonC commited himself to code ( If he cannot work at full time, as I think we can calculate the number of job hours to match exactly with the deadline fixed by the CM community)

5) If in the given time JasonC get the goal, BFC start to pay him as programmer, if not JasonC retreat himself from any technical discussion about the BFC products, after a public and verbose acknowledgement of defeat .

[General Brock]

The gist of this thread seems to be "put some A.I. into the players control" via so called dials or options etc and the response seems to keep being "IMPOSSIBLE IMPOSSIBLE IMPOSSIBLE..." I keep mentioning good ol Battles of Napoleon because it comes to mind but of course there were others. It gives lots of a.i. tweeks in to the hands of players. In very simple but effective ways. My request for all wargame makers is to put more of that control into players hands. Further more let the computer a.i. play against itself too. it's a great way to see how those tweeks affect play.

[JPS]

FWIW, the Space Empires IV (and probably earlier ones as well) AI was more or less continuously updated based on "AI modifications" made by players. Different genre, sure, but similar principle.

IMO it is good from business perspective to first prioritize good looks (i.e. mods), big advertizable numbers (i.e. extensive list of vehicles and weapons), as well as multiplayer options. However, for me personally the improvements made in these three areas have already reached a point of (strongly) diminishing returns.

For the longevity of the series I hope the primary attention will start to shift towards improving game mechanics (including e.g. how arty is handled) as well as the AI.

Best regards,

Jukka-Pekka

[Sergei]

It would be interesting to see if the non-tacAI could be arranged in a way that allows for players to mod it. Not tacAI because that could be used for cheating, but I suppose it's everyone's own business if they want to "dumb down" the AI.

But I presume that while such a thing would at best allow for passionate CM'ers to improve the game with no extra costs, to work it would require extensive documentation from BFC and opening of a part of their code in a way that it could be used by a rival. Yet doing that wouldn't necessarily improve anything, because AI programming is out of the reach of most of us. It's easier and more fun to make scenarios or skin mods, and those who could do a better job than Charles are probably already having so much work that they couldn't devote the time required by it.

[c3k]

I must admit that I'd have never thought of a player modded AI, or different versions of the AI being usable, but the idea intrigues me.

Imagine the bragging rights to "I played 'To the Volga' vs the 'Zukov' AI and beat it!" Or, "The 'Rokossovsky' build crushed me when I gave it a 50% force bonus." "'Paulus' surrendered on turn 99!"

JasonC: be happy that you've gained a special place in Steve's heart. It may not be a GOOD place, but it is special.

Steve: a lot of players are pulling hard to help you. These ideas - misguided, naive, immature, whatever - are all meant sincerely to improve the game and help BFC.

Ken

[Battlefront.com]

All I have to say about Jason's above post is... you don't get it. Obviously you really don't understand how you come off to others (and trust me, I am certainly not alone here). Difference is I know I am being harsh back at you. You're smart, but think you know more than you do. And that makes you annoying when you could instead be helpful. Same was true with people like Username, but thankfully not in the same vein.

The problem is NOT just about Charles' time. It is about hardware issues as well. I have been entirely consistant on that point. There are things that he could do which would dramatically improve the AI's abilities, but would involve gobs of RAM and processing time. Jason doesn't belive this and apparently nothing I can say will sway him.

Jason, you can make lots of unkind insinuations about my level of understanding of these issues, but they are untrue. I know about what I say because... shocker here... Charles and I have been working together directly for 5 years and indirectly for 7. If you think that during this time Charles and I haven't covered this ground in great depth... well... what more can I say.

Now, a user modable AI. First of all, we have nothing against this. But making any sort of API set for the user to tweak is a very laborous task all on its own. To have anything beyond tweakable is even worse. We do not intend on spending any time on this because we feel it will squander valuable resources which could be put to better use in other directions.

In a perfect world the AI would be truely open to the player. A few games have done this, but the ones I know of were much simplier and therefore lended themselves better to such features. We don't think that anything short of a major investment of time would produce anything useful for the end user to play around with. In short, we expect CM's guts to always be a black box without user access.

What we do plan on doing is opening up scenarios to having a lot more variables and parameter settings. This will allow the scenario designer to not only make a game more open to better AI performance, but also to enhance the Human player experience in general. For example:

The scenario designer can designate the "mood" of each side's forces, the general context of the battle, and specific mission goals and their importance. In plain English you might wind up saying that a particular scenario should work like:

Blue Force:

Tired, high morale, on the point of acheiving a long awaited breakthrough. The stakes are high because if this mission fails then the general offensive it is a part of will also fail. Some intel on the Red Force is available, with 3 tanks ID'd and at least 2 companies of troops seen within the area. There is a key hill in the middle of the battlefield that must be taken at all costs and two other minor objectives that can be taken if possible. However, if after the initial attack fails, forces should be reconsolodated and await a follow up attack.

OK, that all sound familar, like a scenario briefing? Yup. But in this case pretty much all the stuff will actually be variables set by the player. In other words, abstract text in CMBO/BB/AK will instead be definable variables in CMx2. These variables will not only give the Human player a more defined role within the game, but obviously the AI player as well.

On top of this we plan on allowing the scenario designer the ability to set up hidden "hints". For example, laying out advance routes and/or defensive phase lines. Weights can be assigned to these features in order to allow the AI the ability to choose which one it wishes to use, under some given circumstances (even if totally random), thereby keeping a scenario playable more than once and different for different players.

On top of that we plan on making some optional "cheats" available, like giving the enemy player more forces in the current CM. "Cheats" like allowing the AI to see the Human's initial deployment prior to deploying its own forces, or seeing the Human's orders at all times. This can also be something variable so as to make the AI "cheat" consistantly or somewhat randomly. Like MAYBE seeing SOME of the Human's initial deployment, or perhaps seeing ALL. Stuff like that.

The fundamental difference between these types of user controled elements and something like what Jason is advocating is how it affects the AI itself. What I described above won't make the AI's deployment logic or path plotting change one bit. However, it can customize how that logic will behave in a given scenario. It is a subtle difference perhaps, but from our point of view it is a fundamental one.

I'll end this post by saying that the more specific or obvious a suggestion about how to improve the AI, the less valuable it is. Charles is a very bright programmer who knows far more about this sort of stuff than anybody else on this Forum. Yes, even more than Jason A good general suggestion can be fleshed out by him far better than any of you can flesh it out yourself. That is not a boast but a simple fact.

That being said, this is an area of game development that we don't think we need help with. The needs are VERY obvious, the solutions (in theory) also plain to see. Unfortunately the realities of making those into code are difficult and not helped by users at all. Therefore feedback from users about improving the AI is, in effect, not likely to produce improvements in the same way discussions about Interface, unit controls, etc. can be.

Steve

[Kellysheroes]

To BFC: Many older games used objective "hexes"/"Sqares" as a means of increasing the AI's chances of winning a scenario vs a human counterpart by giving them victory points "every turn" and then a grand victory point awarded for whoever controlled the victory hex/square at the end of the scenario. Is this what is done in CM series of games?

I may be wrong but it appears to me that the victory points for flag points on the map have a set amount of victory points for controlling them and nothing per turn, I think that if the objective flags gave victory points every turn (more to the computer AI than the human of course) this would lead to some very interesting outcomes and less exploitive maneuvers by the human player.

To me then some of us could play more normal/normal setting games and still find a very challenging game without doubling the amount of units the AI gets. Since the victory locations and TIME taking them would become more of a factor. Blitzkreiging more important than "exploiting" the AI with slow calculated moves. If the AI were getting 100vp's per turn per flag, it would definitely cause some changes in the way many play/exploit the AI. 100vp's is just and example. And I have no idea how easy or hard this would be to program into the game.

It's just a thought to give another type of advantage to the AI and eliminate some exploits.

[Battlefront.com]

Kellysheroes,

To BFC: Many older games used objective "hexes"/"Sqares" as a means of increasing the AI's chances of winning a scenario vs a human counterpart by giving them victory points "every turn" and then a grand victory point awarded for whoever controlled the victory hex/square at the end of the scenario. Is this what is done in CM series of games?

No. There is no concept of "territory" in CM. At the tactical level territory is largely an unimportant issue since posession means nothing without occupation. It also doesn't mean much because one position with a strong unit can effectively neutralize the value of large sections of the map even if an enemy unit occupies it.

If the AI were getting 100vp's per turn per flag, it would definitely cause some changes in the way many play/exploit the AI. 100vp's is just and example. And I have no idea how easy or hard this would be to program into the game.

Actually, it is possible to play premade scenarios with just about the same thing. Some of the best scenarios I have played are REALLY tight on time for the attacker. As the attacker I know I don't have the luxury of waiting things out or figuring out exactly what the AI has and where. Therefore I must move forward with less caution and more haste, which can lead to nail biting game. One of the sceanrios on CMBB required me to move a beat up, exahusted assortment of German units through an unknown gauntlet of Soviet forces at night. My mission was to exit them off the other side of a long and wooded map. I didn't make it Although I did pretty well and kept my force largely intact and pretty much unharmed, the AI managed to hit me hard in my flanks when I least expected it and that slowed me down too much.

With the new stuff we are planning for CMx2 the above type of scenario can be refined even better. For example, the scenario designer can lay out defensive lines for the AI to try and hold, set primary roles for those forces (like counter attacking), etc. For the attacker task forces can be defined and advance routes suggested. For both sides a pace of operations can set in order to tell the AI what sort of tempo it should pursue. Etc, etc.

Now... tie the above comments in with earlier ones I made about the three levels of AI and their priorities. The TacAI decides how units behave to direct stimuli, like tanks being shot at by other tanks, shaken troops getting mortared, selecting optimal targets, etc. This is where ROUTINES and not VARIABLES matter most, so for the most part the TacAI will be "hands off" for the user. However, some things like Arcs and standing orders (SOPs) will allow the player some degree of on the fly control over the general behavior of a unit.

The StratAI is what decides which units do what, where, and when SPECIFIC to a given scenario as the scenario unfolds. This is the bit that paths, task force groups, standing orders (SOPs), etc. The scenario designer will be able to set various things to help guide the general flow of the battle.

The OperationalAI is what decides how the overall SPECIFIC battle will be fought before it even begins. Things like tempo, objectives, acceptable loss threshold, etc. can be set by the scenario designer not only to better adapt the AI's focus for the given scenario but to also set up more diverse victory conditions for the both sides.

The specifics of what we actually implement will not be known for some time, even to us. But hopefully you guys can see that we are not going to rest on our past acheivements and instead push forward wargaming in the same way we did when we released CMBO and CMBB. We have a very solid track record of innovation, quality, and attention to detail. After the last 4 years you guys should be comfortable with the fact that we know what we are doing and will wow you when we do it. In short... don't worry about it

Steve

[Michael Dorosh]

Originally posted by Battlefront.com:

Kellysheroes,

</font><blockquote>quote:</font><hr /> If the AI were getting 100vp's per turn per flag, it would definitely cause some changes in the way many play/exploit the AI. 100vp's is just and example. And I have no idea how easy or hard this would be to program into the game.

Actually, it is possible to play premade scenarios with just about the same thing. Some of the best scenarios I have played are REALLY tight on time for the attacker. As the attacker I know I don't have the luxury of waiting things out or figuring out exactly what the AI has and where. Therefore I must move forward with less caution and more haste, which can lead to nail biting game. One of the sceanrios on CMBB required me to move a beat up, exahusted assortment of German units through an unknown gauntlet of Soviet forces at night. My mission was to exit them off the other side of a long and wooded map. I didn't make it Although I did pretty well and kept my force largely intact and pretty much unharmed, the AI managed to hit me hard in my flanks when I least expected it and that slowed me down too much. </font>

[Kellysheroes]

To Battlefront:

Thanks for the reply. Now another question. During the purchase phase when setting up a Quick Battle, and I set the computer to automatically purchase it's units, does it take into consideration the units I purchase? If not, wouldn't this make it more of a challenge as well.

Once again I have no idea what type of coding this would require, so I'm not going to say this would be "easy" to implement! heh

I must say though I got thwacked yesterday by the computer AI, gave it +25% units and +3 bonus and I was the attacker with a 2000pt buy. I let it make up a completely random map and man did it screw me, my side of the playing field was practically all open and FLAT terrain with a small patch of trees and brush here and there and it's side of the map was covered with cover from buildings to heavy pines, ever try charging a line with over a mile of open terrain? LOL I felt like I was playing the battle of Gettysburg in WW2!

Something I didn't understand though, I set the computer to get HIGH level units and it had green and conscript and I set my units to be MEDIUM and I also got green and conscript...is this suppose to happen this way? Was a quick battle made for the Southern portion of Russia, June 1941 mid-day battle. I lost, damn AI suks! lol And here I am asking to make the AI more challenging! But, mainly I hope to get better with time so I'm just asking for the "future" actually. I don't know if it's the units or there actually was an improvement over the AI from CMBO, because I usually smoke the CMBO AI even with 25% more units and +3 with HIGH level units and it does get veteran and crack units.

Also one more question, any chance of getting the CMBO updated with the options of CMBB? I like the covered arcs and the advance and contact and shoot and scoot a lot. Would really enjoy an upgrade to CMBO with these features!!

TO MICHEAL: Yes that was what I was saying, but, to futher that example the computer AI would get DOUBLE the victory points per turn or some range higher than 1 to 1 to give it a higher chance of winning the scenario vs a driving human player that might push it out very early. Also whoever controls the flag on the last turn gets more like 250 to 500pts for having possession at end of the scenario. This was done in many of the Battles of Napoleans and SSI Civil War games, thus even a strong human opponent to the AI had a tough time of it due to the victory point squares/flags being held by the computer AI for so long.

[ November 02, 2003, 11:48 AM: Message edited by: Kellysheroes ]

[Battlefront.com]

Michael,

I think you missed the point on this one, Steve. Consider a 30 turn game.

Ah... got it. Well, I did actually manage to address his point And that is that holding general terrain is a very difficult thing to assess fairly and in the end is irrelevant from a tactical standpoint. As stated CM makes no attempts to track this stuff now and will not in the future.

In our minds the only things that matter are special areas, Victory Locations, and for these the only thing that matters is posession at the end of the game. It doesn't matter if you held the thing for 25 out of 30 turns if in the end you don't have it. Variable length games, added to CMBB, allow for a possible extension of the game if the Victory Location is still contested. But if the guy pushed off it is too weak, then an extension is unlikely. We think this is a realistic way of assessing victory and therefore have no plans to fundamentally change the philosophy.

Now... as to using "meaningless" terrain posession (as opposed to specific Victory Locations) as a factor for Victory calculations... this won't help the AI at all. At best it will hand the AI some sort of favorable victory result, but it won't make it fight any smarter. The AI already has a sense of the ebb and flow of the battle, adjusting aggression up or down depending on various factors. So in effect counting up who has what terrain isn't useful towards improving the AI.

Kellysheroes

During the purchase phase when setting up a Quick Battle, and I set the computer to automatically purchase it's units, does it take into consideration the units I purchase?

No, because this would be cheating. The AI isn't allowed to know anything more than a Human could.

If not, wouldn't this make it more of a challenge as well.

Yes, perhaps it could. At least it inherently has a chance of giving the AI a leg up for Quick Battles. Certainly something to consider. However, it won't help premade scenarios at all.

Once again I have no idea what type of coding this would require, so I'm not going to say this would be "easy" to implement! heh

Seriously, the fastest way to rub us the wrong way (ESPECIALLY Charles) is to insist that something is "easy". Best to just make a suggestion without a qualifier

Coding the AI to see the Human's unit purchases and deployment is actually quite easy. The problem comes in coding up logic that can transform that information into an advantage. For example, if you purchased King Tigers and a company of PzGrenadiers on the attack it would have to be able to figure out what mix of units would best counter it. And as can be expected, if the same units were on the defensive then the AI would likely have to be smart enough to come up with a different mix of units. This is where AI coding starts sucking up lots of time.

Something I didn't understand though, I set the computer to get HIGH level units and it had green and conscript and I set my units to be MEDIUM and I also got green and conscript...is this suppose to happen this way? Was a quick battle made for the Southern portion of Russia, June 1941 mid-day battle.

IIRC having Rarity on will affect the unit Quality as well as what units are picked. By default the Quick Battle picker tries to come up with a realistic force for that time period and Nationality.

I lost, damn AI suks! lol And here I am asking to make the AI more challenging!

In the right circumstances if the Human player screws up, the AI will have a decent chance of success. Even when a Human player does well the AI often can make him work for victory and sweat a little in the process. But as with any AI, if someone plays against it long enough it will become less challenging over time. The simple reason is because the AI will never inherently be any better than the first time you played it, nor will it offer a dramatically different approach to the next battle. On the other hand the Human player will (should!!) improve over time and adapt his own tactics to the AIs, while the AI can't adapt or learn. With multiplayer you can play 10 games vs. 10 entirely different skill and experience sets which can range from inferior to vastly superior to your own. This is why there will never, ever be a decent AI challenge when compared to multiplayer capabilities.

I don't know if it's the units or there actually was an improvement over the AI from CMBO, because I usually smoke the CMBO AI even with 25% more units and +3 with HIGH level units and it does get veteran and crack units.

There are indeed some subtle improvements in CMBB's AI (and even CMAK's), but the AI certainly becomes a lot "better" with more units at its disposal. More units means it has an easier time organizing itself (i.e. less tough choices to make) and the ability to suffer significant losses without losing the ability to continue fighting in a coherent manner.

Also one more question, any chance of getting the CMBO updated with the options of CMBB?

None. Although some have claimed it would be "easy" for us to do (there's that word we HATE seeing ), the two code bases are so different that it would take many months to make the conversion. There is no incentive for us to do that because it would mean CMx2 would be on hold until it was done. On top of that CMAK does use the CMBB improvements and that game offers much of what CMBO did plus a whole lot more.

Thanks,

Steve

[Kellysheroes]

Thanks Steve, you've really been helpful.

I don't get to play as often as I'd like, but, CMBO and CMBB and soon CMAK are at the top of my top 10 gaming library. The marriage of turn based strategy to real time combat has been one of the best things to happen to computer wargaming and I look forward to what you have instore for CMx2.

Now instead of asking repetitive questions about CMx2, where can I go to read about what's up in the coming version of the Combat Mission series?

I'm also thinking about purchasing STRATEGIC COMMAND, is it complicated to play like "Hearts of Iron"? That game takes too long, how many hours would you say on average it takes to play out a campaign of STRATEGIC COMMAND? Not counting being a new player to it, but, someone that understands the interface and the rules.

Also you may laugh at this, but, do you guys ever intend to make any other era of wargames like the CM series? Like the Civil War and Napoleanic and especially the Ancient Period times like in the BC time period?

Kellysheroes

[Haohmaru]

I've just downloaded and started getting into Biltong's Campaign Rules for the first time ever. I truly hope that the potentially unfavourable conditions and user-made maps make the ai a challenging opponent. If it manages to kick my butt on more than one occasion I will happily retract my remark of it being "embarassing". Again, I have yet to see the ai play particularly well but apparently with the right map, it can. I hope that in the future CM releases BFC will make some improvements that will allow the ai to inherently play well, (or better than currently) including qb generated maps, rather than designers having to "coax" intelligent behaviour from it. I am all in favour of having the option for the ai to cheat as in the methods Steve has mentioned. Who cares if it's unfair for the ai to know what you are going to do, if it makes it more of a challenge it sounds like a great idea.

Haoh

[Sergei]

Originally posted by Kellysheroes:

I'm also thinking about purchasing STRATEGIC COMMAND, is it complicated to play like "Hearts of Iron"? That game takes too long, how many hours would you say on average it takes to play out a campaign of STRATEGIC COMMAND? Not counting being a new player to it, but, someone that understands the interface and the rules.

Have you tried the SC Demo? Probably should give you some indication of how complicated it is.

I haven't played HOI, but considering it's like EU, I think SC (of which I've only played the demo!) is much simpler in playability. Reminds me of Panzer General. Being turn based makes stuff easier in general, of course.

[Sergei]

Oh Steve,

will you guys consider having an optional continuation-of-play at the end of a battle? So that if both sides (or human in single play) agree, at the end of the last turn the final score is calculated but the game itself is allowed to carry on until one side or the other hits "ceasefire"? This has been in some games such as Age of Rifles.

[Panzer76]

The AI can be used for:

Learning the game.

After you have learnt it, you will hopefully find human opponents and never look back. Thats where the real fun is. (not to mention the challenge)

BTW: Did I mention this is the best WWII tactical game ever?

[MikeyD]

About 'continuation of play', have you tried reopening a finished game in a new QB as a QB map and importing the troops? It's not quite what you're asking for, but it'll keep you fighting with many of the same guys on the same terrain.

Can an old QB auto-map be imported into a new QB game, or does it have to be a proper scenario map?

About the AI NOT knowing what units you've selected before selecting its own -- man, you must've programed an ESP ability into the thing then because the AI almost always matches up with my types of units really well!

[ November 04, 2003, 05:59 PM: Message edited by: MikeyD ]