Not to PBEM

[06 Maestro]

ORIGINAL: ravinhood

Thing is Maestro you can play 100 games against the same "human" opponent and reach that same knowledge and boredom. As humans we like things that are NEW...look how you all drool over the anticipation of a newly released game and it's not a day after release for a lot of gamers looking for the next NEW release and/or starting the usual WISHLIST thread on EVERY game. lol I see a couple of people here I think just buy games to be buying them to perhaps be the first to comment on them and then they are off into some other thread or website. But, with that human psychology I think that's why some prefer PBEM or Online opponents, they NEED NEW opponents so they can get a fix of something different each game. It all boils down to the same though, you can play against a lousy human opponent or a lousy AI or you can get lucky and play against a good/great opponent and also play some games with good/great AI's. I personally think they balance out. There's no perfect opponent human or AI.

Tis all true. Those are also all good reason for players to belong to a club with a ladder; you can see who a real player is and who is the yapper, who the rookie is and who is the pro. The War Gamer ladder is a very, very basic ladder. It could be modernized to include comments about a particular scenario and the opponent for the game played.

Also chatting while trying to play a war game can interupt my concentration of my turn. Last thing I want is someone chatterboxing while I'm trying to think out my turn. Thus, for some of us even face to face encounters via TCP/IP must be sort of anti social. I prefer to talk about the game and history of the game and whatnot AFTER or BEFORE the actual game.

Check; in game talk is game related and limited. Normal banter is for afterwords.

[Brigz]

ORIGINAL: Son_of_Montfort

Might I remind you Oleg, Deep Blue beat Kasparov. These rants about "AI being worse than a human" (which implies that AI is for "n00bs" and "dumb-folk") are old and tired. AI can, and has, been produced that outpaces and out-thinks many human players. Properly done AI can think through permutations, think more moves ahead, and "crunch numbers" in odds and casualties that infinitely faster than the human brain. Get off that PBEM high horse and be realistic.

It is a good thing that solo-players have demanded faster, better, and smarter AI. Many of the "bad" AI's out there do not come from overly complex games or "too many variables" but from a lack of developer resources and AI development time. Thus, more of both means AI that comes close to or overcomes, human calculating capacity. Because most wargames are really hex maps (or the like) and number based rules, the AI for wargames should be able to shine far better than RTS or RPG type AI, were AI must emulate unpredictable creatures acting on instinct or emotion.

I think your metaphorical hamster has been, and often is, smarter than you (and most of us) when playing a game.

SoM

Too many make the mistake of assuming because a chess AI can be designed to beat nearly any human that the same applies to an AI for a typical wargame. Programing for chess and programing for a typical wargame are two completely different things.

Factors to consider when programing a chess AI:
Game board - 64 squares - all are open ground.
Pieces - 6 different pieces with 6 specific moves.
Rules - Simple rules that can be summed up on a single page.
Turn sequence - Player A moves one piece; playe B moves one piece. Turn complete.

Factors for a typical wargame AI:
Game board - hundreds but most likely thousands of "squares" composed of dozens of different terrain/weather/supply values.
Pieces - Dozens to hundreds, possibly thousands, of specific pieces with multiple characteristics, including movement, offensive value, defensive value, supply, and organizational sturcture to just name a few possiblilities.
Rules - Usually extensive and detailed and highly interactive. Usually dozens if not hundreds of pages.
Turn sequence - Player A performs multiple phases with possible player B phases interspersed. Player B performs multiple phases with possible player A phases interspersed. Possible book keeping phases: supply, disruption recovery, command structure, unit building phase, etc.

Obviously designing an AI for a typical wargame, especially an accurate, immersive type of wargame that most people on this forum seem to want, would require a completely different approach than designing an AI for chess. I doubt if there is any typical wargame AI that could always beat a human opponent, but there are AI opponents in chess that can always beat a human. You can't compare the two. It's going to be a long, long time before a typical wargame AI will be as capable as a chess AI.

I'm not saying there aren't some good wargame AI's out there and I'm not putting down AI play or the people that play them. I'm just putting things into perspective and pointing out that you can't compare programing for chess with programing for a wargame. They are two completely different things.

[Veldor]

ORIGINAL: Dave Briggs

Too many make the mistake of assuming because a chess AI can be designed to beat nearly any human that the same applies to an AI for a typical wargame. Programing for chess and programing for a typical wargame are two completely different things.

Go ahead and go down to your local bookstore and pick up any good book on programming an AI. You'll see tons of comparisons to a chess A.I. for any type of AI Engine (RTS, Wargame or otherwise).
In some ways a chess AI is actually HARDER to write because there is no randomization in chess. One wrong move early in the game and you are guaranteed to loose against a good player... Thus the need to do all those forward calculations to probable end games and later moves. Thats not necessary with other AIs as randomness erases those mistakes. While a good AI still needs to look ahead it doesn't need to look as far. Just as a human player wouldn't in a wargame but would in Chess.

Paradox's Diplomacy AI is a good representation of a difficult AI to make, yet one that parallels chess in several ways. There is no randomization in Diplomacy, no dice rolls etc, just like chess. Yet due to its extra complications such a mistake still isn't nearly as costly as in Chess, thus far easier to write the AI very very well.

In Diplomacy's case where the AI has trouble is in negotiating treaties and the like with other players. This is a very non-algorithmic type of situation and one that most AIs cannot handle very well. But this is an exception from most wargames. Its normally just a matter of reducing everything to the proper math. Even in the case of negotiating this can be done, its just significantly more complicated as the proper alliances/backstabbing/etc has to then take into account a much longer period of gameplay time (akin to chess again) which wasn't otherwise a requirement of the AI for normal moves.

No its actually no more difficult to make a good AI then it is to make a good GUI or any other element. Its simply a matter of where the time and effort is spent which in wargames is not on either of them. Even with Advanced Tactics its obvious the AI was never really finished as they only now got as far as including routines to even account for each of the unit types in the first place. Obviously it only continues to need to evolve from there. Whether it needs to or should is perhaps part of this debate (pbem or AI?) but the simple fact is if the AI in a game such as Advanced Tactics isn't good enough its not because it CAN'T be.. its just because the time and effort have not yet been put into it (And again maybe doesnt even need to be).

Again with the comparison to the chess AI go look at how much time went into the development of those AIs.. and then compare to something like AT.... That's the real difference and not anything inherent to the game itself.

Yet another difference is those people were absolute AI experts and pioneers. That can hardly be said for the average wargame developer who instead has to be a master of all things (game design, programming, graphics, gui, etc.).

[Fred98]

Deep Blue never had an AI

Instead, given the position of the pieces, the perfect move is this: ****

And given a different position the perfect move is that: ####

All the possible positions of the pieces were considered and the perfect move input for each one.

It was an exercise in accurate data input.
-

[BoredStiff]

ORIGINAL: Dave Briggs

Factors to consider when programing a chess AI:

Pieces - 6 different pieces with 6 specific moves.

OK, although there are a couple of special moves. Firstly, although pawns move in a straight line, they capture diagonally. Pawns also can never move backward, unlike all other pieces.
Then there is the castling move involving the King and either Rook, (neither of which must have been moved prior to castling).
But I'm being anally retentive.
I get your point and agree with it.

Come to think of it, that's what almost all wargames seem to be lacking - a good castling option.

[Brigz]

Veldor, we are just going to have to dissagree on this. It's obvious that you and I see chess and wargaming in a totally different way. I never said that programing a chess AI is easy, but I do think that programing a wargame AI is much more difficult. That doesn't mean that wargame programing is not going to get better, it will. No matter how many books say that chess programing and wargame programing are similar does not mean that programing an AI with all the variables in a state of the art computer wargame is the same as programing a chess AI. I still stand by my statement that wargames and chess are not the same thing and I accept the fact that you do not agree with that.

[Brigz]

ORIGINAL: BoredStiff

Now that's what I call a good chess AI!

[Veldor]

ORIGINAL: Dave Briggs

No matter how many books say that chess programing and wargame programing are similar does not mean that programing an AI with all the variables in a state of the art computer wargame is the same as programing a chess AI.

I didnt say "programming" them was similar, never once, even though it can be. But the way you've compared them doesn't entirely make sense in A.I. terms as the reasons you stated are not the real reasons for the differences. And this is why almost any good A.I. book will, in fact, at some level begin with an appropriate Chess A.I. comparison.

I happen to be fairly well read on this topic, and working on an A.I. project of my own.

But rather than argue what is an infinitely long set of points, I'll simply refer you to a much longer analysis and explanation that is free on the web:

The Current State of Human-Level Artificial Intelligence in Computer Simulations and Wargames.
http://www.gilgameshcontrite.com/Computer_AI/

While I don't agree with everything he states, I do agree with much of the conclusions in terms of why wargames are failing at A.I. and that what we are all describing is pretty easily achievable, if not now then in only a few years. I find him to be a plenty credible source on the topic.

You can skip to the conclusions at:

http://www.gilgameshcontrite.com/Comput ... ge_47.html

Also he himself begins with a comparison to Chess AI, though he doesn't really properly explain its differences in the commercial wargame space as the focus of his statements is on the AI such as the military desires to simulate out of the box human thinking vs in a more commerical application where the "Discrete State Game" has much more use as a basis for comparison.

In any case he does an excellent job of quickly summarizing the various generic types of A.I.s and goes so far on the summary page to state that the "FSM/DSM" is the basic building block of most AIs. This again refering to the technique used in many Chess AIs.

Thus confirming the accuracy of my statements. [:D]

[Veldor]

Oh, while we are on the topic, the researcher also correctly illustrates the big flaw with A.I. cheating which I brought up I think in a different thread.

"Furthermore, there is a long history of cheating in commercial computer game AI design. Even when it is not overtly cheating many of the AI routines that are now being created have no real value outside the original application. For example, one recent technique is for the level designer or artist to pre-plot waypoints to be used in lieu of heuristic pathfinding."

The tendancy to do this is a lazy approach that doesn't evolve the industry. Thus another simple reason for lack of A.I. advancement.

[Oleg Mastruko]

Might I remind you Oleg, Deep Blue beat Kasparov.

Ah I wondered when will the DB get into this thread [;)]

Deep Blue was built with practically unlimited resources from IBM, who saw it as premier propaganda project. It took endless gabazillions of dollars and kabazillions of engineer hours to get built.

Plus most wargames - the kind of like WITP and EIA - are more complex than chess.

Still Deep Blue is an interesting, perhaps even epic achievement, no doubt. A machine beat the best human in quintesential game of chess!

However most wargame developers don't have 0,0000001 of IBM resources, to build an AI for a game more complex than chess (complex, because players want complex!)

For me it's also a psychological question (the proverbial hamster). It is impossible for me to feel emotions when playing the AI, regardless whether I am winning or losing.

[Veldor]

?!?!?

[Fred98]

ORIGINAL: Oleg Mastruko
It is impossible for me to feel emotions when playing the AI, regardless whether I am winning or losing.

I agree. Playing a human has a certain Edge to it.

Playing an AI is boring!!!!!!

-

[Brigz]

ORIGINAL: Veldor


I happen to be fairly well read on this topic, and working on an A.I. project of my own.

Well, you got me there. I concede.

Thus confirming the accuracy of my statements. [:D]

If you're happy, I'm happy.

[Veldor]

ORIGINAL: Dave Briggs
If you're happy, I'm happy.

Well I'll be happy if you read that article, or given that its 47 Pages if you at least read the last page. [:D]

I think its also worth mentioning that any modern wargame should have an AI helper for you even if you never play the AI itself. This same AI helper should/would understand many of the very things that an AI opponent would have too.

At a very generic level thats not just for things some would rather do themselves (Like micromanage every single subtle factor) but those things such as pathfinding (And pathfinding doesnt necessarily have to mean that which you'd find in a realtime game but rather any game where pathing is a factor maybe even Advanced Tactics and how it calculates supply routes for instance). You wouldn't want to pick an EXACT supply route for each unit if you could choose would you?

The more the AI can assist you the more programming the actual AI "opponent" becomes a practice in creating true "artificial intelligence" vs. simply teaching it to behave within a certain fixed set of rules.

It also provides value to the game and the AI programming investment to even those who don't wish to play a true AI opponent.

[Brigz]

ORIGINAL: Veldor

Well I'll be happy if you read that article, or given that its 47 Pages if you at least read the last page. [:D]

Out of curiosity I read the article. I came across this (my emphasis in bold):

"The Environment

What military and sporting simulations have in common is that they are not Discrete State Games.

Chess is a Discrete State Game. There are a finite number of possible positions; it is an extraordinarily large number but it is a definitive number. If one had a computer, or a book or a file cabinet that was capable of storing every conceivable position in chess with the best possible move to make at every discrete state in the game then you would have the ultimate, unbeatable chess master. Call it AI, call it an expert system, or just call it a giant look up table but it would be definitive and unbeatable.

Indeed, any Discrete State Game strategy or Discrete State problem can and will be solved. It does not matter what methods are used in the short term — neural nets, genetic algorithms, or brute force — there is a definitive solution waiting out there.

This is not the case of simulations. Computer simulations, by definition, are emulations of the real world. The real world rarely rests in discrete states."
Here's the reference: http://www.gilgameshcontrite.com/Comput ... ge_02.html

Sounds to me like the guy is saying that chess is distinct from a simulation. Being as well versed in game theory as you seem to be, you know that a wargame is by definition a simulation, it is not a discrete state game. Chess is not a simulation and by the above definition is a discrete state game. My original point was that chess and wargames are two different things. You can easily go back and read this in my original post. This quote from the article you provided seems to back that up.

[Veldor]

ORIGINAL: Dave Briggs

Sounds to me like the guy is saying that chess is distinct from a simulation. Being as well versed in game theory as you seem to be, you know that a wargame is by definition a simulation, it is not a discrete state game. Chess is not a simulation and by the above definition is a discrete state game. My original point was that chess and wargames are two different things. You can easily go back and read this in my original post. This quote from the article you provided seems to back that up.

Right. But now skip to the last page of conclusions and read how that is still the basis for and/or a large component of almost every A.I. currently developed.

So the techniques used in a chess AI are viable enough to wargame simulations, they just aren't necessarily the entire answer in and of themselves to ultimately achieve the same human-like level that they may with chess.

Which is why I've been a big proponent of experimenting with alternative A.I. styles and techniques but there doesn't seem to be much interest in other wargame developers in doing the same (As again pointed out in the article by the researcher).

It's far easier for devs to just skip to cheating.

In the vein of a single game maybe no price is paid. A decade later though we wonder why other genres have advanced past ours.....

[Brigz]

ORIGINAL: Veldor

So the techniques used in a chess AI are viable enough to wargame simulations, they just aren't necessarily the entire answer in and of themselves to ultimately achieve the same human-like level that they may with chess.

That's pretty much my initial argument. Chess AI programing and wargame AI programing are two different things.

"Thus confirming the accuracy of my statements.[:D]"

[Veldor]

That's pretty much my initial argument. Chess AI programing and wargame AI programing are two different things.

"Thus confirming the accuracy of my statements.[:D]"

Well if your happy then I'm happy. [:D]

But they are different like Peanut Butter is to Cashew Butter and not like Peanut Butter is to Jelly... Which is what it would seem your implying...

[Brigz]

ORIGINAL: Veldor

But they are different like Peanut Butter is to Cashew Butter and not like Peanut Butter is to Jelly... Which is what it would seem your implying...

More like a peanut butter and jelly sandwich. But of course this adds the complexity of bread.[;)]

[Zap]

Dave Briggs
Been reading this thread and I'm just shaking my head. Who gives a rat's ass who plays email and who plays strictly AI? It's the same unending, unanswerable argument as "do you like hexes or no hexes", "do you like 3D graphics or 2D graphics". It's not a matter of which is better, it's a matter of personal taste. I play both AI and pbem. Having said that, I have to admit, Oleg has some good points about playing a human opponent and I prefer playing another person than the AI. When playing a human opponent I will always play out to the end, but many times playing against the AI is boring and I am more apt to lose interest. Just me, just a matter of personal taste and I hold no measure up to anyone. Vive la differance. (Pardon my French)

Hamster chess does sound pretty interesting though.

ORIGINAL: Zap

It was'nt supposed to be a thread of I like PBEM(better gaming) compared to Not PBEM(best way to game). That was done by Oleg when he had disdain for my style of presentaion. And revealed that he looks down on players of the AI(only).
Actually I was surprised to see so much venom for my post. Which, gave no reason for the response it got.

Anyhow, From what I understand right now, with my job requirments, I would not make a good PBEM opponent. Since most want continuity of play. I could not set up a strict time to return a turn.

Well I'm glad that the AI will be continued to be developed. Although not as fast as most AI players would like it.