Forum adverts like this one are shown to any user who is not logged in. Join us by filling out a tiny 3 field form and you will get your own, free, dakka user account which gives a good range of benefits to you: No adverts like this in the forums anymore. Times and dates in your local timezone. Full tracking of what you have read so you can skip to your first unread post, easily see what has changed since you last logged in, and easily see what is new at a glance. Email notifications for threads you want to watch closely. Being a part of the oldest wargaming community on the net. If you are already a member then feel free to login now.

As Ostrakon points out, there is a huge difference between doing a simulation of a few scenarios and "solving" a game with a very large (though theoretically finite) set of rules.

Also, I think you are mistaking the level of complexity involved in military simulations. It's not like they take all the platoons and have an AI figure out how they move all over Iraq. No, it is much more likely to be a scenario like:

"We have 6 brigades in this region and the enemy has 3 brigades here and 2 over there. Based on past experience, it will take us a 2:1 advantage to wipe out an enemy brigade. How do we allocate our 6 brigades?"

These things are waaay simplified and abstracted. And often-times you run a simulation with real people playing the OPFOR and the BLUFOR. Having real people play a "simulation" using some rules and random elements is far simpler than writing an AI to find an optimal solution. In fact... most "simulations" are likely to be closer to people actually playing a game of 40k than a super computer crunching numbers.

If you guys are wondering, this is how you do it. You take the units, the objectives and the deployment. Deployment is in some ways the hardest part but it is just the same as the rest of it, really.

You don't use any sort of a grid, what you do is consider the battlefield from the perspective of the units themselves, in sort of circular wedges. The six intervals are your approximation.

The hardest part is actually LoS but it is the same as the rest of it. What you would want to do is isolate that section of the code as a subset of other decision making.

What some of the previous posters may be missing, this isn't game theory, it is A.I. There is a difference. A.I. is a means of making decisions. All it has to do is be reasonably logical and make good calls, same as say modern sports video games. Heck, it is even turn based, in a game with a limited number of turns.

It's possible. It would probably take a great deal of specific coding for paper-rock-scissor match-ups but it is just the same as any other strategy game.

Tacobake wrote:What some of the previous posters may be missing, this isn't game theory, it is A.I. There is a difference. A.I. is a means of making decisions. All it has to do is be reasonably logical and make good calls, same as say modern sports video games. Heck, it is even turn based, in a game with a limited number of turns.

It's possible. It would probably take a great deal of specific coding for paper-rock-scissor match-ups but it is just the same as any other strategy game.

I think you are vastly overestimating the capability of AI to make proper decisions in open ended situations. To take into account the entire board with the implications of each target is a vast amount of decisions. Modern video games do not even come close to the sort of thing people are talking about. Often the AI has to follow a script or cheat to provide any sort of challenge.

Tacobake wrote:Any idiot could write a computer program that could beat humans at 40k, ESPECIALLY if they already knew how to write chess programs or other AI. Thinking turns ahead etc. Oh also it is only a six turn game with very clear rules of engagement. Most computer scientists could probably do it in an afternoon, other than the actual moving all the guys around. Everyone except Blackmoor who has that Brightlance list.

The computer's greatest advantage of course being the ability to guess ranges and measure with perfect precision. Besides knowing all the rules, not forgetting things etc etc.

Anyone who could write an AI that can plan ahead like that in an afternoon would be getting a Nobel Prize. Most computer scientists would say you are nuts. The choice of shooting a squad with a melta gun at a tank or infantry may seem simple, but start writing down everything influencing that decision and its effects in later turns, and you see how complex it can be for a computer.

A computer's ability to measure ranges wouldn't even enter into it since that is cheating and wouldn't be allowed. Ranges and LOS would be the easiest part.

Gestalt wrote:

Tacobake wrote:What some of the previous posters may be missing, this isn't game theory, it is A.I. There is a difference. A.I. is a means of making decisions. All it has to do is be reasonably logical and make good calls, same as say modern sports video games. Heck, it is even turn based, in a game with a limited number of turns.

It's possible. It would probably take a great deal of specific coding for paper-rock-scissor match-ups but it is just the same as any other strategy game.

I think you are vastly overestimating the capability of AI to make proper decisions in open ended situations. To take into account the entire board with the implications of each target is a vast amount of decisions. Modern video games do not even come close to the sort of thing people are talking about. Often the AI has to follow a script or cheat to provide any sort of challenge.

Tacobake wrote:Any idiot could write a computer program that could beat humans at 40k, ESPECIALLY if they already knew how to write chess programs or other AI. Thinking turns ahead etc. Oh also it is only a six turn game with very clear rules of engagement. Most computer scientists could probably do it in an afternoon, other than the actual moving all the guys around. Everyone except Blackmoor who has that Brightlance list.

The computer's greatest advantage of course being the ability to guess ranges and measure with perfect precision. Besides knowing all the rules, not forgetting things etc etc.

Anyone who could write an AI that can plan ahead like that in an afternoon would be getting a Nobel Prize. Most computer scientists would say you are nuts. The choice of shooting a squad with a melta gun at a tank or infantry may seem simple, but start writing down everything influencing that decision and its effects in later turns, and you see how complex it can be for a computer.

A computer's ability to measure ranges wouldn't even enter into it since that is cheating and wouldn't be allowed. Ranges and LOS would be the easiest part.

Once again QFT.

Tacobake wrote:
What some of the previous posters may be missing, this isn't game theory, it is A.I.

Thanks for proving your stillborn opinion invalid so I didn't have to.

Trying to find the absolute optimal army across all lists and tactics isn't a game theory issue? Stop calling yourself a computer scientist, it's frankly insulting.

here is my chart,

basically it gives the number of wounds/attacks as whole fractions. so for example 10/27 means 27 attacks will generate 10 unsaved wounds.


to hit 2/3
Str - T 6 5 4 3 2 PW
2 25/54 10/27 5/18 5/27 5/54 5/9
1 10/27 8/27 2/9 4/27 2/27 4/9
0 5/18 2/9 1/6 1/9 1/18 1/3
-1 5/27 4/27 1/9 2/27 1/27 2/9
-2 5/54 2/27 1/18 1/27 1/54 1/9


to hit 1/2
Str - T 6 5 4 3 2 PW
2 25/72 5/18 5/24 5/36 5/72 5/12
1 5/18 2/9 1/6 1/9 1/18 1/3
0 5/24 1/6 1/8 1/12 1/24 1/4
-1 5/36 1/9 1/12 1/18 1/36 1/6
-2 5/72 1/18 1/24 1/36 1/72 1/12


to hit 1/3
Str - T 6 5 4 3 2 PW
2 22/95 5/27 5/36 5/54 3/65 5/18
1 5/27 4/27 1/9 2/27 1/27 2/9
0 5/36 1/9 1/12 1/18 1/36 1/6
-1 5/54 2/27 1/18 1/27 1/54 1/9
-2 3/65 1/27 1/36 1/54 1/108 1/18

Tacobake wrote:What some of the previous posters may be missing, this isn't game theory, it is A.I. There is a difference. A.I. is a means of making decisions.

Granted not everything in AI is game theory, but in the end we're going to have to evaluate potential courses of action against one another, and I would
like to know how you propose to do so with out game theory in this context.

I wouldn't argue that developing a system that is capable of playing warhammer is impossible. In fact, I would have loved to do so if I had started playing
40k prior to starting my master's thesis, but don't allow hubris to fool you into believing that it could be done in an afternoon.