Artifical Intelligence for World in Flames

[Shannon V. OKeets]

ORIGINAL: ravinhood
I'd like to give a little more input of why I don't mind the ai having advantages I don't have or as some call them "cheats". When you play a human individual who is of lesser experience than you are you can "teach" them to improve their game. This is something the player cannot do with a computer ai other than at least "options" to give it some distinct advantages. Most often it is in numbers of units, but, it is rarely Ever improved gameplay (slitherine uses a feature where the ai doesn't actually do anything for a certain amount of turns per difficulty level, I don't really consider this improved gameplay) whereas another human player usually can and will improve. Also since the majority of computer games sold are for the "solo" play (at least by all the polls I've seen), there definitely needs to be things that can at least improve the "challenge" of the game even for those that play hardcore highest difficulty from the start.

I've just never understood this philosophy of not adding "options" to increase difficulty beyond the norm. Why is it so hard to have something in the options that the player can adjust to make any game extremely challenging to them on an individual basis? Why is it so hard to add an option to increase the offense/defense stats of the ai units in % increments?? I've always thought a slider from 10% to 100% increase to offense/defense (one for each) stats would be perfect for computer wargames. Die roll modifier options as well. Maybe I'm thinking too much in board wargame play, but, it certainly was easy to do when giving a lesser experienced player some advantages when I played them.

Your point is well made. I'l keep it in mind.

Clearly the implementation in software of what you are requesting is relatively easy. So it really is a philosophy and ego issue at this point. Of the two, the ego thing will be harder for me to overcome.

[SamuraiProgrmmr]

ORIGINAL: ravinhood

I'd like to give a little more input of why I don't mind the ai having advantages I don't have or as some call them "cheats". When you play a human individual who is of lesser experience than you are you can "teach" them to improve their game. This is something the player cannot do with a computer ai other than at least "options" to give it some distinct advantages. Most often it is in numbers of units, but, it is rarely Ever improved gameplay (slitherine uses a feature where the ai doesn't actually do anything for a certain amount of turns per difficulty level, I don't really consider this improved gameplay) whereas another human player usually can and will improve. Also since the majority of computer games sold are for the "solo" play (at least by all the polls I've seen), there definitely needs to be things that can at least improve the "challenge" of the game even for those that play hardcore highest difficulty from the start.

I've just never understood this philosophy of not adding "options" to increase difficulty beyond the norm. Why is it so hard to have something in the options that the player can adjust to make any game extremely challenging to them on an individual basis? Why is it so hard to add an option to increase the offense/defense stats of the ai units in % increments?? I've always thought a slider from 10% to 100% increase to offense/defense (one for each) stats would be perfect for computer wargames. Die roll modifier options as well. Maybe I'm thinking too much in board wargame play, but, it certainly was easy to do when giving a lesser experienced player some advantages when I played them.

While I agree with your statements in principle, there is one part to that argument that is distateful. It seems to me that when the computer is given bonuses to make up for inadequate programming that the game then becomes figuring out how to beat the computer rather than figuring out how to play the game the best way.

This may seem like a small difference. I think it is more important than it seems on the surface for the following reason: What we seem to need is a way to train new players so they will be able to be competitive playing against other human players. In order to learn these skills from the AI, it is important that the AI play well rather than rely on advantages to be 'competitive'.

Writing the AI is not the problem. The problems are :

• 
  [* accomplishing it on a budget (both time and money)]

I have great faith in the ability of this programmer. He is the most organized developer I have ever been around. If it can be done, he is as likely to get it done as anyone I can think of.

[Lothos]

Shannon I sent you a PM and email please drop me a line via IM or email about AI stuff.

[Shannon V. OKeets]

ORIGINAL: Lothos

Shannon I sent you a PM and email please drop me a line via IM or email about AI stuff.

Done.

[PanzerMike]

Wow, Lothos the Hoi2/DD AI Master is in the house. This is gonna be interesting. Yay [:D]

[DerekP]

If people seriously think the AI can be challenging for the average / superior player they will be kidding themselves.

The biggest problem in all startegy games is that the players "cheat". Not by breaking the rules (although some will do that[&:]) but by exploiting the rules. If an AI is programmed to execute a multiple choice type attack (if conquer France then choose Spain, Britain or Balkans depending on certain factors) then the players will know this and plan accordingly or even dupe the AI into choosing a favourable course of action for the player by artificially strengthening or weakening fronts.

Similarly any knowledge of the AI's rules (if Britain does this then Germany will probably do that) will lead to exploits

You can randomise the AI's actions to a degree but this leads into problems with sub-optimum (in the long run) strategies being chosen even if they shock the player in the short term.

The best we can hope for is an adequate AI for learners and a solid MP interface (and yes, Lothos is the best around [&o]and if he can help manage that for WiF then he will have excelled himself).

[wodin]

I hope this thread stays helpfull to Steve rather than turning into "We dont want an AI, we want the game" as usual in past threads.

[hakon]

ORIGINAL: DerekP

If people seriously think the AI can be challenging for the average / superior player they will be kidding themselves.

The biggest problem in all startegy games is that the players "cheat". Not by breaking the rules (although some will do that[&:]) but by exploiting the rules. If an AI is programmed to execute a multiple choice type attack (if conquer France then choose Spain, Britain or Balkans depending on certain factors) then the players will know this and plan accordingly or even dupe the AI into choosing a favourable course of action for the player by artificially strengthening or weakening fronts.

Similarly any knowledge of the AI's rules (if Britain does this then Germany will probably do that) will lead to exploits

You can randomise the AI's actions to a degree but this leads into problems with sub-optimum (in the long run) strategies being chosen even if they shock the player in the short term.

The best we can hope for is an adequate AI for learners and a solid MP interface (and yes, Lothos is the best around [&o]and if he can help manage that for WiF then he will have excelled himself).

I agree with the above.

When humans play vs an AI, they will always (my thesis) tend towards playing an anti-AI game. This means figuring what kind of decisions the AI is particularily weak at, and exploiting that.

This is even true for the strongest AI's made. For instance, when top human chess players meet the best AI chess programs, they will use different playing styles from what they use against human opponents.

Once you "figure out" the AI, the game is worthless for single player, unless you have the option to give the AI some advantage.

Anyway, AI advantages can be put into a game very late in the development cycle, so can be put in when it is discovered just how easy it will be to beat the AI. If Shannon is confident enough to not plan for AI advantages from the start, I dont mind, unless it becomes an Ego thing that prevents him from putting such features in when/if it is discoverd that the AI will not provide a real challenge to a competent human wif player.

[mlees]

Not everyone is an elite player.

I am a noob WiF player. By that, I mean I never played a human opponent. Therefore, any AI will be a challenge for me for some time. And if I decide to get my feet wet by playing a human, I will get spanked, and run back to playing the AI for emotional support.

Those customers who buy this computer game (and who do not own the boardgame) will need some kind of AI to play against while learning the mechanics of the game.

If they want to find opponents, or discuss strategy and become better players, then they can be directed (by info that is in the "about" section of MWiF) to these forums.

[Greyshaft]

ORIGINAL: hakon
If Shannon is confident enough to not plan for AI advantages from the start, I dont mind, unless it becomes an Ego thing that prevents him from putting such features in when/if it is discoverd that the AI will not provide a real challenge to a competent human wif player.

Perhaps it would be gracious to give the Man room to show us what he can do before we condemn him. Suggesting he has an 'ego' if he does certain things sets the two of you up for an adversarial relationship which will not help MWiF get out the door. Your comment may be valid but there are softer ways of saying these things.

[Zorachus99]

ORIGINAL: mlees

Not everyone is an elite player.

I am a noob WiF player. By that, I mean I never played a human opponent. Therefore, any AI will be a challenge for me for some time. And if I decide to get my feet wet by playing a human, I will get spanked, and run back to playing the AI for emotional support.

Those customers who buy this computer game (and who do not own the boardgame) will need some kind of AI to play against while learning the mechanics of the game.

If they want to find opponents, or discuss strategy and become better players, then they can be directed (by info that is in the "about" section of MWiF) to these forums.

Somewhat reluctantly I'm forced to agree. I'm exceedingly cruel exploiting new players weaknesses. The simple reason is: if I don't play at my best level, the other person will never learn to beat me. I've taught three (3!) people how to play the game, it was a work of patience and a good return after a VERY long time. This is the kind of game that can take years to master.

I say let the AI do the best it can to prepare new-comers to game mechanics, because I would say the majority of us prefer to be challenged by a human opponent. It's very satisfying to have a 'real' someone on the receiving end of a lucky die roll.

One thing bothers me: when playing over a game you roll and handle the dice physically, hoping you can impart some kind of luck to your die rolls. My experience is that sometimes die rolls made by a computer seem less... organic. I can see people blaming the random number generator already

Crazy luck can turn a game upside down... (particularly with naval combat) I see no way to avoid it. It's also crazy luck that ensures games don't play out identically, thereby making WIF a work of art. The number of possibilites are so much closer to infinity than a game of chess.

[Shannon V. OKeets]

The number of possibilites are so much closer to infinity than a game of chess.

Mathematicians define two kinds of infinity: countable and uncountable. I believe both of these would be considered finite though.

[Neilster]

ORIGINAL: mlees

Not everyone is an elite player.

I am a noob WiF player. By that, I mean I never played a human opponent. Therefore, any AI will be a challenge for me for some time. And if I decide to get my feet wet by playing a human, I will get spanked, and run back to playing the AI for emotional support.

Those customers who buy this computer game (and who do not own the boardgame) will need some kind of AI to play against while learning the mechanics of the game.

If they want to find opponents, or discuss strategy and become better players, then they can be directed (by info that is in the "about" section of MWiF) to these forums.

Exactly. Expert players will play each other. No wucking forries.

Cheers, Neilster

[Klingon]

ORIGINAL: Shannon V. OKeets

The number of possibilites are so much closer to infinity than a game of chess.

Mathematicians define two kinds of infinity: countable and uncountable. I believe both of these would be considered finite though.

(reminisces)
Ahh; yes. One of the few times that x-1=x is indeed possible; when x is infinity.

[Anendrue]

As I see it an Ai is indespensible.

1) It helps tutor new players.
2) An advanced played can test various strategies while making assumptions on what a human player might do. Which could be relayed to Steve for MWiF 2 [:D].
3) If there was no Ai then no Ai would ever get better. Only by building something do we see the flaws that can then be improved upon.
4) Network down and you need a quick fix of strategy gaming.

[chopper66]

I expect this is too late, but if you are still interested in having someone with a strong background in AI then let me know.

My Staff Page

[Shannon V. OKeets]

ORIGINAL: idwilson66

I expect this is too late, but if you are still interested in having someone with a strong background in AI then let me know.

My Staff Page

I am.

Send me an email on your interest: Steve@PatternDiscovery.us

[macgregor]

Go team! Go!

[ravinhood]

Don't forget though there is something beyond infinity....have you never heard of infinity plus shipping and handling??
 
Also, I'm just very glad to read that Shannon/Steve appears very interested in creating a challenging decent ai. It's something that in most games just gets tossed in at the last minute and gets very little support afterwards.

[Neilster]

This link to research at the University of Tasmania (UTAS) on Multiple Classification Ripple Down Rules (MCRDR) may come in handy when designing the AI....

http://www.comp.utas.edu.au/users/bhkang/

I'll try to get Sprocc involved again too as his research work built on MCRDR. Something went wrong with him getting his NDA and then he was so busy with his PhD that he lost interest.

Cheers, Neilster