Will the AI ever be improved?

[ravinhood]

Ahhh yes, now Fantasy General was pretty good. I enjoyed that game a lot. Cmon now that's the right idea create some more games like that, but, prehaps with a bit more tactical combat. I think I have most all the fantasy/sci-fi type wargames/civilization/Homm (except for IV & V) that were made. Did you work on "Sword of Aragon" as well?? I had a blast with that, but, wished it had a more randomized map so it would have had some longevity. Also do you remember "Sword of Fargoal"?? lol that game is soooooo old, but, I still play it every once in awhile. You had to find the Sword of Fargoal which was usually on the oh 17th to 20th level of the dungeon and then you had 30 minutes to find your way back out. They just don't make them like that anymore. Talk about an adrinilin rush.

[JanSorensen]

ORIGINAL: ravinhood

Well I'm here to say the ai in Master of Magic is pretty good. I'm at the gates of losing a game setup on Hard w/large map, 4 other computer AI and powerful magic.

 
Lets not confuse "good ai" with "the computer opponents are playing by such different rules that they become hard to beat".
Lets also not confuse games of "just have multiple ai opponents gangs up on the player" with actually strategy games.
 
I know you mention those things yourself - but they simply cannot be stressed enough. I already said it in a prior post - Civ type games are much, much, much  easier to code a reasonable AI for simply because its acceptable to have the AI not only cheat in starting forces, production, and by knowing exactly what you are doing but also by having the AI players gang up nomatter what. Those tricks, while smart, are not really AI.

[Grotius]

I think the GGWaW AI does a decent job, considering the complexity of the game as compared to chess.  And I think the charge that devs aren't interested in AI is unfair.  The people I know who code games actually say the AI is their *favorite* task.  Certainly Shannon, over in the WIF forum, has said that repeatedly.

In any case, whatever the AI's flaws, they seem to me mitigated by the ease of playing PBEM with this game.  It's the most PBEM-friendly wargame I own.  I've played a fair number of PBEM games with my friends, and no one has quit on me yet.  The game moves fast, and even if the war isn't going well, the victory conditions are balanced well enough that you have a shot at some sort of decent outcome.

One nitpick: the strong AI in "Galactic Civilizations II" doesn't cheat except on the hardest level.  I have yet to beat it on "Normal," in which the AI doesn't cheat.  So yes, the devs did make AI a big priority in that game.  But the downside is that it has no multiplayer, and as a result I've played it far less than GGWaW.

[JanSorensen]

One nitpick: the strong AI in "Galactic Civilizations II" doesn't cheat except on the hardest level.  I have yet to beat it on "Normal," in which the AI doesn't cheat.  So yes, the devs did make AI a big priority in that game.  But the downside is that it has no multiplayer, and as a result I've played it far less than GGWaW.

 
I took a brief look at the game you mention. Out of curiosity. Do you know for a fact that the AI does not know at all which race is being played by a real player and which are player by the AI and uses that to gang up? Thats the most common "trick" in games of those kinds to make things hard for the single player without the AI appearing to be cheating.
 
Another thing I noticed is that the game boosts a very steep learning curve. I obviously do not know how much you have played it - but a possible explanation to not having won yet might be that its just a complicated game rather than the AI being exceptionally strong. I venture that a mediocre or even poor AI would kick me silly in Go the first many times I was to play not because the AI is good but because I would take a long while to learn the ropes.
 
Obviously, the AI may just be strong - in which case I applaud the designers

[ravinhood]

Frog-guy the designer/developer of the game assurd us the ai doesn't use any forms of cheats in GalCiv II until one goes beyond "tough" challenge level. Even then it is in resources. There isn't really a "gangup" effect in any level of difficulty except under certain conditions, the player is attacked according to his weaknesses. If he gets too economy oriented and falls behind on military, yeah boy some of them will come running. With the added fear factor, even the aggressive ai's that attack the player and vice versa the player gets aggressive can find themselves in a several front war because the game is programmed to go after aggressive players (this to enhance the challenge of the game in the middle to end game).

I've played a game with total peace and used influence and/or diplomacy to win. The newer patch has the ai's becoming more aggressive to this and also if the player is going for a technical victory (kinda like building the spaceship in civilization and the ai's become more aggressive to stop you).

At any rate whether it's cheating or not the ultimate point about AI is "challenge" and being able to challenge even some of the most hardcore gamers/wargamers out there. This is something standard wargames need to come up to par on. I like HPS type games, but, the AI is pitiful. All I'm asking and saying, is fine make the game as you intended, but, just add more "challenge options" to it. A few combat modifier sliders/options, die roll options/modifiers that go beyond the games hardest setting difficulty would be welcomed I'm sure.
 
Also putting the AI aside. What I buy games for is a good challenging game that "can beat me at some point". Most every game I get I start at the hardest difficulty and work my way down should I find myself unable to beat it. Games that I start at the hardest difficulty and beat the very first time I really get annoyed with and the developers and designers for not considering some of us out here are pretty good players and need more of a challenge beyond the max set difficulty. I don't buy games for multiplayer and/or ISP, though I do occassionally play some PBEM games even those must be something that can be finished within a couple of weeks, maybe a month max. I lose concentration of what I am doing playing pbem games when the opponent takes days to make a move.
 
Lol I remember when I took a couple of these online IQ tests, I came up with Einstein and Computer Analitical wiz. Heh, I should be the AI inside of peoples games. lol My numbers were 140+ I guess that's good eh?

[pvthudson01]

Another thing I noticed is that the game boosts a very steep learning curve

LOL so does World At War!

[Warfare1]

Good AI is important to strive for.

However, what can help an AI greatly are very user friendly and well-crafted unit, scenario, campaign, and map editors for any game. The players themselves can then tweak a scenario; add more units; and give the AI more resources and experience; etc....

The Civ games all come with excellent editors - you can modify almost anything. I have had my butt handed to me a number of times.

It is amazing how difficult experienced user-made scenarios can be.

For example, in Strategic Command 1, I am currently using the in-game editor to make my games very difficult as the Axis. Believe me, my games are now extremely tough, yet they also seem to be historical.

Provide great editors with every game; make the AI good; and games can be made tough, even for veteran players.

[Lebatron]

The thing about GalCiv II is that its not a game of positional movement. Creating AI that can recongnize patterns and see a few moves ahead is a very difficult thing to create. Notice how simple Chess is? A few pieces, not many rules, no build que to speak of, and it took years with the best minds in AI programming to make a Grand Master equivalent. I know you may want to point out that GalCiv does have positional movement. What you got to realize is that it's an entirely different form and in my opinion does not really count as the type of positional play that is the very hard kind to program for. The game space is entirely open to move around in. So all the AI really has to do to seem really good is just have a decent build strategy(very easy to script BTW) and seek out and attack you wherever your weaker than him. He may hit some world of yours that's lightly defended and you say to yourself damn that was smart of it. Not really. You could if you where in the right position return the favor. And this slugging match would go on until someone gets the upper hand. It basically boils down to who has the larger economy. If your looking for a space game (4x game) that does have a bit more strategy in it try Space Empires IV. In that game there are warp holes that create a real game map feel instead of the pure open space that you get in GalCiv II.

[Lebatron]

ORIGINAL: ravinhood

Lol I remember when I took a couple of these online IQ tests, I came up with Einstein and Computer Analitical wiz. Heh, I should be the AI inside of peoples games. lol My numbers were 140+ I guess that's good eh?

I did too. When your in the top 1% you can hit the max IQ score these tests are designed to handle. When you are that intelligent only a real full blown IQ test can tell you where your at when your over the 140 range.

[Hycanth]

ORIGINAL: JanSorensen
[...] Civ type games are much, much, much  easier to code a reasonable AI for simply because its acceptable to have the AI not only cheat in starting forces, production, and by knowing exactly what you are doing but also by having the AI players gang up nomatter what. Those tricks, while smart, are not really AI.

I agree with the spirit of what you are saying, but I wouldn't say that Civ AI's are easier to code. Civ AI's are complicated by large numbers of pieces, spaces, and turns, as well as complicated production models. Picking the good moves from the bad with all of these factors can be difficult for the AI. Historical games can be somewhat more scripted (attack WA in Fall 1941, CAGs against Pearl Harbor, snag Singapore, etc...) into a "playbook". A human player is at a disadvantage until he can develop his own such playbook.

Now, yes, it is easier to create the illusion of reasonable AI in a Civ game. Because you can make it cheat. A lot. And it is hard for the player to know just how badly it is cheating. Master of Magic is a great example of abyssmal AI that cheats to no end. On hard difficulty, it doesn't just cheat somewhat; it cheats massively. (If you want to get an idea, look at an AI's mana income when they control just their tower!!) And still a human player can crush the AI with just about any reasonable core strategy. Even on impossible, (where the AI enjoys something like x10+ mana production!?!) a human can routinely defeat the AI with certain "uber" strategies (All life picks + halflings is one). That is because the AI is so poor. It will mill massive numbers of troops about without any sense of purpose. (Just imagine the Russian AI circling 20 tanks pointlessly around Siberia in 1941...)

Yes, I know that GalCiv (1 or 2) is not supposed to cheat (at all) until the upper difficulty levels. But, quite frankly, GalCiv is a cakewalk until the Crippling+ difficulty levels where it does cheat. I think that most people who consistently have difficulty at lower difficulties are struggling with the bizarre economic model and the fact that they do not enjoy micromanaging *every* turn. This doesn't mean they are bad players, just that their sense of fun doesn't permit them to stoop to certain chores. The AI, of course, will.

Anyway, given all of this, I would say World at War has quite good AI. It is capable of giving new players a decent challenge without even cheating. And then, at higher difficulties, you know exactly where it will cheat. Its major weakness is being timid. The Axis just never seem to go for the throat, even when it is exposed.

Hycanth

[ravinhood]

All I am saying on this issue and I think I've said it before. Give us the ability an options to make the game harder than the hard coded hardest difficulty. I don't care if the AI cheats. I welcome the challenge no matter what it does as long as it is a challenge. There is no use having a computer opponent if it can't win or at least scare the pants off of yah that it might just actually win.
 
Even playing another human being, if you beat them time after time after time you eventually become bored with playing against them. (you can only needle them so much anyways lol). A computer opponent though with the proper programming could be made to be as hard as someone wanted it to be with "cheats" though they aren't cheats at all they are really advantages and handicaps. Look up the word "cheat" it has nothing to do with what computer AI's get on higher difficulties. A computer opponent doesn't even know how to cheat. It's not programmed to cheat.

[pzgndr]

Give us the ability an options to make the game harder than the hard coded hardest difficulty.

There is an option. Different Axis and Allied mods could be developed for AI play which adjust OOBs and research/production priorities to compensate somewhat for AI weaknesses. I mentioned this a while back and Lebatron initially balked at the suggestion but then went in and made some decent adjustments to the editable AI files. So it's possible to make some improvements with what is allowed to be edited. Also, customizing later scenarios such as 1941 or 1942 for the AI could result in more challenging gameplay. Since it's unlikely the generic AI will be enhanced, it would be interesting to see more custom mods developed that focus on either the Axis or Allies with reasonable bonuses and handicaps implemented.

[Hycanth]

ORIGINAL: ravinhood

All I am saying on this issue and I think I've said it before. Give us the ability an options to make the game harder than the hard coded hardest difficulty. I don't care if the AI cheats. I welcome the challenge no matter what it does as long as it is a challenge. There is no use having a computer opponent if it can't win or at least scare the pants off of yah that it might just actually win.

I'm not sure what you are getting at here. You can give all of your opponents up to +9 combat help and +999 supplies/turn. Is that not challenging?

At a certain point, just giving advantages to the computer to increase difficulty ceases to be interesting. You (the human) end up playing not the original strategy game, but a very different "puzzle" where you must take advantage of the "AI" tendencies to achieve victory under absurd circumstances. Imagine giving your opponents 10 free tanks per turn. If you still stand a chance of beating them, something is wrong.

Even playing another human being, if you beat them time after time after time you eventually become bored with playing against them. (you can only needle them so much anyways lol). A computer opponent though with the proper programming could be made to be as hard as someone wanted it to be with "cheats" though they aren't cheats at all they are really advantages and handicaps. Look up the word "cheat" it has nothing to do with what computer AI's get on higher difficulties. A computer opponent doesn't even know how to cheat. It's not programmed to cheat.

That is a good point about the semantics. I believe an AI "cheats" when it has an advantage of which you are unaware. In Civ games, you are usually quite uniformed of all the advatages given to the computer (and there are many). Hence, this is why I say MoM does in fact cheat. Your opponents are actually playing a slightly different game, and you don't even know their rules. How can you be sure your actions will help in defeating them? Does it even really *matter* that you razed one of their cities to the ground?

In WaW, at least I know the advantages my opponents have (or so I believe), and can plan accordingly. The same is true, more or less, in GalCiv.

Hycanth

[Aloid]

Joel, why not take advantage of communities like this to develop AI? Give us the tools to mod sections of AI routines, as well as the over-arching context of a language, tags, parameters, etc. My god, there are enough freaks willing to spend hours for free developing mods like this.

You'd have a lot of work to do to externalize something like that, but think of the possibilities...

It would be something like the seti project where you leverage other people's processing power (brain power). Who knows, someone would probably write such amazing AI that we'd need to share CPU time just to use it.

Aloid

[Rocko911]

Joel & Ravinhood, I found where you can still get Fantasy General as a download http://www.panzergeneral.org/SSI%20Downloads.htm so see there is still a following your good work , Joel. I guess its been considered abandoned ware.

[ravinhood]

I have Fantasy General so I don't need to download it. It's one heck of a game if you don't play a certain way as you get to the end you'll get your butt handed to you.
 
And yeah letting us into the AI scripts or whatever in the program is an excellent idea. I worked on Spartan's Ai and on impossible now and evens sometimes on hardest it's impossible to win, but, CHALLENGING to try.
 
I also finally got so sick of beating RTW's sorry excuse for an AI that I turned off MORALE before starting a campaign and I mean to tell you on VH/medium is kickin ass hard now. Mainly due to no routing and losing lots of men on both sides. It's a slugfest, but, the battles last a long time, some past 30 minutes and after it's over I don't have these 18 men lost I have 1800 men lost sometimes. lol And if you use HUGE units you'll be out of an army fast and sucking for air and begging your people to populate. It's actually turned into the game I wanted. I just never knew turning morale off would have such an effect. Even battles with rebels can be devastating to ones army. If you've never tried playing it this way I suggest giving it a try. That is IF you like CHALLENGE over historical accuracy.
 
And I guess I like PUZZLES as someone mentioned earlier. Trying to stop a real horde is more fun than just plowing thru every single one of them and watching them rout the moment I hit one single flank. Wish we could save the log files as movies like MTW. I've had some that were just epic. Me 3000 men vs the Barbarians 6000 men. That was the best battle I ever had (with morale off). I got beat but I took 4800 of em wit me. Suks they didn't put movie logs in RTW.