Are Wargame AIs Fated to Suck?

[sullafelix]

Okay, to me that is not an exploit and is an example of an AI that needs to be worked on. In many games the AI defends much better than it attacks so I usually attack.

To me when you describe an 'exploit' I am thinking about one certain case of using an odd tactic that the AI cannot cope with.

There was one game where the AI couldn't deal with units that used the outside of the map.

So you had some players who were massing all of their units either at the top or bottom of the map to take advantage of this.

Which leads into another part of the equation. Why would anyone boast online that they beat an AI?

I am not talking about you, just people in general. Because, that is what happened in the above scenario.

So you are talking more about things like this. There was a Napoleonic wargame that was made not too long ago, where the AIs artillery would charge out in front of it's troops and get slaughtered.

To be honest, that was fixed in the game.

[aaatoysandmore]

ORIGINAL: Tesuji

ORIGINAL: aaatoysandmore
Was that "hard mode"?[:)]

Honestly, I don´t remember. But as you insist it is as good as BN, I will see if I find the game somewhere and give it another try.

What exactly do you mean by hard mode? Any specific setting I should put up to 5?

When you get to the "New Game" selection screen there are some numbers for HighLord from 1 to 5 in two catagories. Change these from 3's to 5's on both. One is just difficulty and the other one is reinforcements. It's very challenging. Even on 3's though I had some very good games of it even though I won them eventually. But, it was a grind still.

Make sure they say "HighLord" because some of them are to give you advantages as the Whitestone player.

Some interesting articles on AI.

https://software.intel.com/en-us/articl ... mes-part-1

http://www.stat.columbia.edu/~jakulin/FT/

[sIg3b]

ORIGINAL: sulla05

Okay, to me that is not an exploit and is an example of an AI that needs to be worked on. In many games the AI defends much better than it attacks so I usually attack.

To me when you describe an 'exploit' I am thinking about one certain case of using an odd tactic that the AI cannot cope with.

There was one game where the AI couldn't deal with units that used the outside of the map.

So you had some players who were massing all of their units either at the top or bottom of the map to take advantage of this.

Which leads into another part of the equation. Why would anyone boast online that they beat an AI?

I am not talking about you, just people in general. Because, that is what happened in the above scenario.

So you are talking more about things like this. There was a Napoleonic wargame that was made not too long ago, where the AIs artillery would charge out in front of it's troops and get slaughtered.

To be honest, that was fixed in the game.

Wow, we are finally getting on the same page. [:)]

Yes, my problem is not with very odd tactics that no one would think to use.

Rather that most AIs cannot cope with obvious and ordinary tactics.

2 most important categories of crappy AI:

1. AI opponent is extremely predictable.

So you have these scissors/stone/paper situations where you *know* the AI will always use scissors. Now of course I will always use stone. I can´t convince myself I do not know what the AI is going to do.

2. AI does not use its units as an organized whole.

Now as insects do not have a very sophisticated CPU, it should be possible for an AI force to at least behave like an insect hive. Instead, more often than not, we have individual units that act like fire-and-forget missiles. If the AI designer viewed the whole force as *one* organism rather than a large number of individual entities, this mindset would be a large step toward a better AI.

So I want a hive-like AI with some high-level behaviour parameters randomized. This could be a worth-while opponent.

[EwaldvonKleist]

I don't doubt that some programmers could put more effort in their AIs and i also do not criticize a particular post here. But before blaming developers for beeing silly, think a bit how you would design an AI.
Lets assume a hexagon map with a front line, manned by a number of 60 divisions (both sides).
How would you tell the AI
-whether it is on the defense or attack over the whole front?
-when to attack/defend at a particular area even though the whole front strategy is different?
-whether units are threatened by encirclement over the nex 1-2 turns
-whether it should do a risky attack resulting in the encirclement of an enemy army, but also exposes the best own units to destruction
-how to find the best attacking shedule requiring as few MPs as possible, has minimal losses, still allows supply to flow to units during logistics phase etc.?
Now try to do not actually program it but formulate the Algorithm in Detail. You have no idea? I don't have one for every question too (i am not a programmer but interested in such problems). Maybe you will discover a solution after some time of intense thinking. But it won't be easy...

[sullafelix]

The other problem we have with AIs is if we do create one.

There is a post on the SimHQ forum that states a scientist saying that within an hour at most an artificial AI that starts out with the same brain power as us, would be a million times smarter.

So probably no pulling the plug on it once it starts going.

[EwaldvonKleist]

Well, for games we are speaking from a very very simple AI. It only does things the programmer has foreseen in some way and won't, in most cases, not learn from one game to the next one.
The AI the Scientist means will more like a neuronal network or something like this.
Two completely different things, the "intelligent" AI is way off topic (nothing which would bother me of course).

[aaatoysandmore]

ORIGINAL: Tesuji

ORIGINAL: sulla05

Okay, to me that is not an exploit and is an example of an AI that needs to be worked on. In many games the AI defends much better than it attacks so I usually attack.

To me when you describe an 'exploit' I am thinking about one certain case of using an odd tactic that the AI cannot cope with.

There was one game where the AI couldn't deal with units that used the outside of the map.

So you had some players who were massing all of their units either at the top or bottom of the map to take advantage of this.

Which leads into another part of the equation. Why would anyone boast online that they beat an AI?

I am not talking about you, just people in general. Because, that is what happened in the above scenario.

So you are talking more about things like this. There was a Napoleonic wargame that was made not too long ago, where the AIs artillery would charge out in front of it's troops and get slaughtered.

To be honest, that was fixed in the game.

Wow, we are finally getting on the same page. [:)]

Yes, my problem is not with very odd tactics that no one would think to use.

Rather that most AIs cannot cope with obvious and ordinary tactics.

2 most important categories of crappy AI:

1. AI opponent is extremely predictable.

So you have these scissors/stone/paper situations where you *know* the AI will always use scissors. Now of course I will always use stone. I can´t convince myself I do not know what the AI is going to do.

2. AI does not use its units as an organized whole.

Now as insects do not have a very sophisticated CPU, it should be possible for an AI force to at least behave like an insect hive. Instead, more often than not, we have individual units that act like fire-and-forget missiles. If the AI designer viewed the whole force as *one* organism rather than a large number of individual entities, this mindset would be a large step toward a better AI.

So I want a hive-like AI with some high-level behaviour parameters randomized. This could be a worth-while opponent.

You would probably like Combat Mission I or II then as that pretty much is what it has is a hive mind on offense. It's pretty predictable and runs up the same paths of cover nearly every game I've played. While they have improved the not guarding the edges a bit it's still very predictable. It rarely if ever splits its forces and once you find the nest the mortars and artillery can have a field day.

About the best thing with AI's today are when the game is brand new and you don't know the AI patterns. But, nearly every game just as soon as you see them and learn them then it's the same game after game. It's like playing your little 6 year old brother time after time game after game. He was such a sucker for 52 card pickup. [:'(]

Granted all games AI's aren't bad just the majority of them.

My top 10 AI games

1. War of the Lance
2. Civilization I-IV (this ai cheats abundantly though and we all know it).
3. Spartan 1.013 version (after I modded the AI build patterns)
4. Conquest Medieval
5. Centurian Defender of Rome
6. Battles of Napoleon
7. Panzer Command
8. Command Ops
9. Norbsofts Civil war series
0. Ageod's Ale Jete Est

[sIg3b]

ORIGINAL: sulla05

The other problem we have with AIs is if we do create one.

There is a post on the SimHQ forum that states a scientist saying that within an hour at most an artificial AI that starts out with the same brain power as us, would be a million times smarter.

So probably no pulling the plug on it once it starts going.

Nah, as Iain Mc Neil rightly pointed out, an AI is very game-specific. If it can beat me at Steel Panthers, doesn´t mean it can take over the world.

Fantasy Empires beats me, and still the Dungeon Master isn´t ruling us yet. (Or perhaps he is, and we are not intelligent enough to notice. [:)] )

[sIg3b]

ORIGINAL: aaatoysandmore

You would probably like Combat Mission I or II then as that pretty much is what it has is a hive mind on offense. It's pretty predictable and runs up the same paths of cover nearly every game I've played. While they have improved the not guarding the edges a bit it's still very predictable. It rarely if ever splits its forces and once you find the nest the mortars and artillery can have a field day.

About the best thing with AI's today are when the game is brand new and you don't know the AI patterns. But, nearly every game just as soon as you see them and learn them then it's the same game after game. It's like playing your little 6 year old brother time after time game after game. He was such a sucker for 52 card pickup. [:'(]

Granted all games AI's aren't bad just the majority of them.

My top 10 AI games

1. War of the Lance
2. Civilization I-IV (this ai cheats abundantly though and we all know it).
3. Spartan 1.013 version (after I modded the AI build patterns)
4. Conquest Medieval
5. Centurian Defender of Rome
6. Battles of Napoleon
7. Panzer Command
8. Command Ops
9. Norbsofts Civil war series
0. Ageod's Ale Jete Est

You missed the part where I said AI should be a Hive Mind, but also somewhat random and unpredictable.

I agree: Civ3 and 4 particularly have a reasonably good AI.

Most astonishing AI, however: Fantasy Empires!

After a game or two I suddenly noticed the AI was starting to counter my strategy. It learned from me! [X(] Playing late at night, that was a creepy experience. [:)]

Only game with a learning AI I have ever encountered.

[sullafelix]

I meant if we as humans can create real AI.[:D]

[catwhoorg]

ORIGINAL: Tesuji

ORIGINAL: sulla05

The other problem we have with AIs is if we do create one.

There is a post on the SimHQ forum that states a scientist saying that within an hour at most an artificial AI that starts out with the same brain power as us, would be a million times smarter.

So probably no pulling the plug on it once it starts going.

Nah, as Iain Mc Neil rightly pointed out, an AI is very game-specific. If it can beat me at Steel Panthers, doesn´t mean it can take over the world.

Fantasy Empires beats me, and still the Dungeon Master isn´t ruling us yet. (Or perhaps he is, and we are not intelligent enough to notice. [:)] )

That is a blast from the past. Loved that game

[sIg3b]

ORIGINAL: sulla05

I meant if we as humans can create real AI.[:D]

We should start with creating real I first. [8D]

[sIg3b]

ORIGINAL: catwhoorg
That is a blast from the past. Loved that game.

Still playing it (with Dosbox).

Being a purist, I roll my characters -no tinkering with the stats!- and play Ironman (no reloading except for crashes).

I only take on 1 or 2 opponents, though. Playing against 4 AIs is not a promising proposition, since they have learned to gang up on me and rush me before my defences are up. And if I try to pay for peace, they take my Gold, then still attack next turn! And then they paradrop Undead in my backyard. Gamey, cheating, backstabbing bastards! [:@]

[aaatoysandmore]

And if I try to pay for peace, they take my Gold, then still attack next turn! And then they paradrop Undead in my backyard. Gamey, cheating, backstabbing bastards! [:@]

nothing a human wouldn't so I'm sure. [:'(]

[sIg3b]

ORIGINAL: aaatoysandmore

And if I try to pay for peace, they take my Gold, then still attack next turn! And then they paradrop Undead in my backyard. Gamey, cheating, backstabbing bastards! [:@]

nothing a human wouldn't so I'm sure. [:'(]

That´s what´s bothering me. My psychiatrist says there are no evil Dwarves, Elves, Clerics and Mages in my notebook, or at least not many and they are harmless, but I know better!

[sIg3b]

Kinda like Brazil marching into Berlin in Paradox's HOI? Then on to Moscow. Is that gaming the game or just stupid AI?

Neither. It is a stupid game.
In a well designed game, something like that is simply impossible to happen. Period.

[sIg3b]

What I am talking about is using the rules against a human or an AI.

When you are playing someone a war boardgame and you notice a flagrant or stupid rule, you just ignore it. That is if you are playing with someone who is a friend or not out just to get a win and game the situation.

In a real classical boardgame there are no such rules.

Take Rise and Decline of the 3rd Reich for example. What you might call exploits is exactly, literally, the way the game is meant to be played.

And it´s not me who is saying this. It´s the Wargamer´s Guide to Thrd Reich -yes, the official guide!- that is saying this.

The real classic games are not meant to be played casually. They are meant to be played with the dedication of a professional chess player.

[sIg3b]

I never really considered Third Reich a wargame or simulation it is a game that really abstracts war.

Third Reich does an excellent job at limiting the outcomes on the strategic level to what was historically possible. It also does a good job of showing what was most important (the economy!). It does this by consistently throwing out the irrelevant little details. Such as one German infantry corps doing 10% better than another. They are all the same perfectly interchangeable 3-3s. And that´s what I like about Third Reich: It does away with the random anecdotical stuff and focuses on what was important.

[sIg3b]

" Better playtesters.

I can understand designers overlooking AI problems; but how come the playtesters don´t find exploits that the players find within 3 days after release?

Answer: Not enough time given for testers after AI completion, or wrong playtester activity focus, or the wrong playtesters.

Either nobody is listening to the testers, or the testers are fans who marvel at the graphics or the chrome or whatever gimmick of the day instead of single-mindedly trying to do this to the AI:"

I am a bit confused over this attitude. If you bought a game that you were interested in why would you look for exploits in it, and if you accidently found one what would that mean?

Would you uninstall the game and never play it again?

I have several games, perhaps a lot more I don't know about, that have game ending exploits. I just don't use them, and have great fun playing the games.

I will give you that playtesting is much more about checking WAD then about trying to find exploits in the AI.

My confusion comes because of this. Say I buy a truck, and I find out on the web or by myself that if I am in drive I can physically move the gear into reverse and strip the gears in the transmission.

Yes, it is a problem and should be rectified, but I don't park the truck and never drive it until it is fixed. I just don't put it into reverse while driving.

I have played against myself, as Mrs wargamer writes about, with board wargames for decades.

Of course, I could have always done stupid moves or nothing if I was on one side of the table. But, why would I do that?

To illustrate my point, I found out by accident how to beat a tough monster in witcher 3. I was reading about it and someone posted an exploit without putting it in a spoiler.

Okay, so now I knew. It really didn't matter, I didn't use it. It took me about ten tries but I beat it without it.

I can see your point.

However, my attitude to all this is: If a game does have (unintentional) exploits, it´s a sloppy design and I don´t like it.

That´s why I think the Heart of Iron series is basically crap.

[Kriegsspieler]

Umm . . . anyone up for some thread necrophilia?
It's kind of unusual to see the last poster in the thread before it went dormant 6 years ago also be the one to resurrect it.