Are Wargame AIs Fated to Suck?

[Mobeer]

ORIGINAL: loki100

ORIGINAL: Mobeer

1,2,3,4,5,6 is a terrible choice - it is as likely as any other combination, but huge numbers of people buy those numbers, so the shared payout is much lower than for other numbers

actually they don't.

at least for the UK national lottery they do:
http://news.bbc.co.uk/1/hi/sci/tech/240734.stm

and in more detail:
http://www.hpcc.ecs.soton.ac.uk/lottery/201.html

[Orm]

Numbers above 31 is much less used.

[loki100]

ORIGINAL: Mobeer

ORIGINAL: loki100

ORIGINAL: Mobeer

1,2,3,4,5,6 is a terrible choice - it is as likely as any other combination, but huge numbers of people buy those numbers, so the shared payout is much lower than for other numbers

actually they don't.

at least for the UK national lottery they do:
http://news.bbc.co.uk/1/hi/sci/tech/240734.stm

and in more detail:
http://www.hpcc.ecs.soton.ac.uk/lottery/201.html

its the sequence (non-random) not the actual numbers that makes that appear to be less likely - I've done the research believe me

[altipueri]

There is a difference between the chance of winning a 6/49 lottery which is about one in 13 million; and the number of people that you would share that winning with.

The chance of getting six numbers is the same whatever you choose. The chance of sharing them increases the more you choose low numbers because many people use birthdays (1-31) or house numbers which gives a slight bias to lower numbers; but 1,2,3,4,5,6 is as likely to occur as 49,48,47,46,45,44 assuming the number selection system is fair.

[Hattori Hanzo]

ORIGINAL: Ostwindflak

I've solved it, we should all sell our computers and virtual games and go back to board gaming. Especially the big board war games. You're welcome. [:D]

or you can fight against a human opponent with PBEM.

in my opinion the AI opponent is usually good enough when you start to learn a strategy-game or a wargame.

another good option to learn a game that I use sometimes, is to play PBEM against myself.

[MrsWargamer]

I guess I see AI from a multitude of angles.

The state of war gaming.
The state of online dating, and being able to find real humans on a dating site (they are a surprising amount of bots).
The state of OSs which are not a massive dose of goofs gaffes and just plain stupidly programmed software.
The state of robotics.

We won't be taken over any time soon. They can't think, they can't communicate intelligently, they're easy to hack into, and they won't outdo the fleshies for mobility.

[catwhoorg]

ORIGINAL: Orm

Numbers above 31 is much less used.

Not many people have birthday's on the 32nd or later.

[IainMcNeil]

The issue with AI is usually the rules. Until all the rules are coded the AI cant be made to work with them. Sometimes when changing a game rule it can be orders of magnitude more work to make the Ai deal with the change. Even as much as changing some stats or balancing without changing a game rule can break the AI (if you change how many VP are awarded for controlling a location on the map it could mean the AI no longer evaluates its options correctly). When playing a human player you adapt to their strategies. Making an adaptable AI at the moment is pretty much impossible.

And the biggest issue is that the AI is written to go with the specific game. It is single use and not reusable. The great AI in ATG is of absolutely no use in Vics next game unless it uses exactly the same mechanics, which games never do. You always end up writing the AI from pretty much scratch. 3D engines you can reuse, game engines you can reuse but rarely are you able to reuse AI so you're coding it from scratch each game. This is why you don't see a steady evolution in AI as you do in graphics, performance etc.

We do spend an awful lot of time working on AI but it is a challenge!

[Recognition]

I take the simple view regarding the AI in our gaming world.

In real life you can have a brilliant and competent General, lets say a 9, and a clueless incompetent General at 1, throw in the FOW and misplaced or badly interpreted orders and any end result can happen. So for me a well directed AI I put down to a well organised and implemented HQ system, and a lesser, sometimes incomprehensible AI down to a complete breakdown of the HQ system with orders misunderstood and incorrectly carried out.

So in my gaming world I always have a good excuse for the AI behavior when playing solo.


PS Generally in my experience the weaker AI gives the greatest gaming pleasure [:D]

[wings7]

ORIGINAL: Iain McNeil

The issue with AI is usually the rules. Until all the rules are coded the AI cant be made to work with them. Sometimes when changing a game rule it can be orders of magnitude more work to make the Ai deal with the change. Even as much as changing some stats or balancing without changing a game rule can break the AI (if you change how many VP are awarded for controlling a location on the map it could mean the AI no longer evaluates its options correctly). When playing a human player you adapt to their strategies. Making an adaptable AI at the moment is pretty much impossible.

And the biggest issue is that the AI is written to go with the specific game. It is single use and not reusable. The great AI in ATG is of absolutely no use in Vics next game unless it uses exactly the same mechanics, which games never do. You always end up writing the AI from pretty much scratch. 3D engines you can reuse, game engines you can reuse but rarely are you able to reuse AI so you're coding it from scratch each game. This is why you don't see a steady evolution in AI as you do in graphics, performance etc.

We do spend an awful lot of time working on AI but it is a challenge!

A good broad analogy and explanation!

[EwaldvonKleist]

General Topic:
To get a good AI, following requirements must be matched:
1)simple game rules, few possibilites (so e.g. few hexes)
2)sophisticated algorithms
3)computation power

1) is nothin a wargamer wants
2) Expensive to develop for small or even medium sized studios
3) Expensive for customers. Who want's to buy a workstation or a mac pro for thousands of dollars (and, depending on the situation, the cooling system) only to allow th AI to make turns in 10 minutes instead of a day?

2) Is probably the most interesting point. Just save a few million dollars/euros, go to an elite university and hire some specialists. Then give them two years and they will for sure produce a superior AI... But who has the money for this? A super AI unluckily does not increase the number of customers by 1000%[:(]

Lets take the Google Go project as an example: They gave very good specialists the time, money and computation power to design the GO AI. WIth simliar ressources, a wargame AI would be a hard opponent.

[IainMcNeil]

But create a superior AI for what? It has to be game specific and would be of no use in the next game. That's the biggest issue. You try getting the Go AI to play chess - I know what will happen!

[Hattori Hanzo]

without forgetting that both Chess and Go are infinitely more simple games, from the point of view of programming a valid AI for them, that any strategic game or wargame.

[sIg3b]

ORIGINAL: Matti Kuokkanen

ORIGINAL: sulla05

Are you playing historically or are you just trying to game the rules and system?

I have to agree with this. Those who understand how the game works, how the numbers work and how decimals are rounded, how AI works, they play Civilization IV on hardest deity level and win against all the AI. Those who don't have the knowledge what make the game tick and make the best moves at all times (like I) are best to play it on middle difficulty levels.

The gists of such arguments is: "If you play fast and loose and do not much watch and analyze your opponent, you may not even notice how much the AI sucks." [:-]

I was a horrible chessplayer, now I am a mediocre chessplayer. Just about everything I have learned I have learned from the Chessmaster AI.

So the *ideal* AI -I AM NOT SAYING IT CAN BE DONE!- would be an AI where you can tell beginners: "Just watch the AI to learn how to play well."

I am not expecting this to happen, though.

[sullafelix]

"Are you playing historically or are you just trying to game the rules and system?"

This part of my post can go for human players also. If I am playing a historical wargame, I am not going to use and game the rules just so I get a win.

[JWW]

ORIGINAL: Tesuji

ORIGINAL: Matti Kuokkanen

ORIGINAL: sulla05

Are you playing historically or are you just trying to game the rules and system?

I have to agree with this. Those who understand how the game works, how the numbers work and how decimals are rounded, how AI works, they play Civilization IV on hardest deity level and win against all the AI. Those who don't have the knowledge what make the game tick and make the best moves at all times (like I) are best to play it on middle difficulty levels.

The gists of such arguments is: "If you play fast and loose and do not much watch and analyze your opponent, you may not even notice how much the AI sucks." [:-]

I was a horrible chessplayer, now I am a mediocre chessplayer. Just about everything I have learned I have learned from the Chessmaster AI.

So the *ideal* AI -I AM NOT SAYING IT CAN BE DONE!- would be an AI where you can tell beginners: "Just watch the AI to learn how to play well."

I am not expecting this to happen, though.

Like I said before, it sounds like you migrated here from the Steam forums. I think the solution for you is to never play against the AI. That seems to be the only way for you to enjoy computer games. Nothing wrong with that.

One of my first computer games was Avalon Hill's Third Reich for PC. That was a long time ago. I quickly learned there were certain things I could do that the AI would simply not react to.

So I was left with a few choices. Give up the game because the AI "sucked." Use those exploits and easily defeat the AI. Not use those exploits and get a better game. I chose and still choose the third option in wargaming. And I find many games that have an adequate AI for MY enjoyment as long as I don't seek and use exploits. Some AIs indeed are horrible no matter how you try to play the game. And I get great enjoyment from this. You don't. Different strokes.

[sIg3b]

ORIGINAL: sulla05

"Are you playing historically or are you just trying to game the rules and system?"

This part of my post can go for human players also. If I am playing a historical wargame, I am not going to use and game the rules just so I get a win.

Well no, for me a "Historical Wargame" is a Strategy Game with a historical flavour. NOT analysing the system and optimizing my moves translates into "making obvious mistakes" for me. If the game wants me to use historical tactics, it should give me strong incentives to use them, not expect me to play the game subpar.

I mean I think I understand what you are saying: You prefer the historical chrome over everything else. Well, no: I prefer balance and challenge over everything else and would always sacrifice chrome to gameplay if they collide.

[wodin]

It's surprising and comforting to see so many posts that are in the same boat as myself considering nearly every game forum has a complaint about the AI. I just count myself fortunate to not be that good. Plus I never game a system nor like the others play the same game enough to learn how the AI works and then take advantage.

Now saying that I wont play any games I love against a human opponent. Thatw ay I never become aware of the AI's shorcomings. I used to love Squad Battles and played PBEM. However I harted being tied o a game so stopped, the things is though I could never go back and play the AI as it justw asn't the same. So what was obe of ym favourite games was taken of my hard drive. I don't want to do hat to other games I love so refuse online play now. Only shooters Verdun and Red orchestra 2 will I play online only.

[EwaldvonKleist]

ORIGINAL: Iain McNeil

But create a superior AI for what? It has to be game specific and would be of no use in the next game. That's the biggest issue. You try getting the Go AI to play chess - I know what will happen!

Well, if i were a Programmer, coding the AI would be the most interesting part of the game developement i think. I agree that it does not make sense to Put that much effort in a not recycable AI, and i dont contest that super AIs make no economic sense for wargames.
So it is more a scientific exercise. Maybe the basic methods like neuronal networks can be recycled.

To the game aspects: I just like to look for bugs and exploits in strategy games. I played a lot Android wargames from joni nuutinen and reported dozens of exploits. I also do not feel good until i squeeze out everything (exploit free) out of the system. That's why i have so much troubles with monster games cause you need a long time to unterstand them entirely.
I myself would like to play an ai which is a challenge with equal ressources after you have mastered the ruleset. But i also know this makes zero economic sense so i dont complain

[Ostwindflak]

@Hattori: PBEM is okay for what it is. I just prefer the face to face interaction over a good board game while sipping on a glass of scotch. It comes down to preference I suppose.

@Matti: I have never actually considered travelling to Finland, but wouldn't rule it out.