A good AI?

[Seamus Fermanagh]

Has anyone encountered a strategy or battle game with a good AI?

I do NOT mean one that is tough with abilities/morale/etc ramped up, but one that "behaves" more challengingly when the difficulty is increased and can make a moderately experienced player work to succeed?

I have yet to encounter one that was flexible enough and solidly grounded enough to challenge me once I had become experienced enough at a game to know a few tricks.

E.G. Grigsby's 2nd Front. Computer lets you sneak a corps, unsupplied, along the northern edge of the map. You send a corps to cut the rail line West of Vologda (easy since the lines are drawn to defend Moscow) and have the northern corps plop itself on the archangel railroad. Leningrad is out of supply and falls in two turns as its troops lose strength.

Seamus

[womble]

I reckon that if some game did include an AI which (wrt playing the game, that is... :-), could be considered equivalent to a human, taking into account the perfect and instantaneous knowledge of tis own situation, that it would still need some sort of cheat-handicapping to give different difficulty levels.

If it could play at a perfect level, including its perfect self-knowledge and perfect understanding of the rules, then you could handicap it by censoring its knowledge or fuzzying it up somehow.

Don't think it's going to happen soon, though, for a game as complicated/complex as a TW title.

[Deus ret.]

With respect to AI, RTW was a big step back compared to previous TW games, at least insofar as the extent of the challenge is concerned. The AI is well able to pull off interesting maneuvres, they don't occur often and seem to occur rather randomly, though.

If you want a more nasty AI, try the Darth mod. It may make you rejoice even if you lose.

[SMZ]

I'd say 10 years from now we should expect a good challenge. These things seem to increase by leaps and bounds. Some things like the more automatic routines being used in games like Black & White 2, the Sims, etc... or the facial technologies developed with Half-Life 2, etc... computers are progressing quickly in what they can do and how well they can emulate ppl...

As for a game that challenges you without resorting to "cheating" or being assigned a superior position... can't really think of one

[professorspatula]

AI is something that hasn't improved as fast as other aspects of games. It takes time and lots of testing to get something close to good AI, and improving AI is an ever on-going process. I wonder if its time for the industry to consider employing specialist AI middleware or external AI developers to concentrate on the AI routines which are then adapted for the current title. I think CA has proved they haven't got the skills or time to produce exceptional AI yet, and really few games could boast about exceptional AI either. Perhaps middleware AI is an impossibility to implement, or overly expensive, but AI in games needs a kick up the backside in general. I'm fed up with idiot bots in games ignoring the player or doing utterly stupid things all the time.

[Red Harvest]

Civil War Bull Run has good AI (tactical game.)

You aren't likely to see good AI in many games for the foreseeable future.

Why? Each time you build a new game engine, you need a new AI. AI takes lots of time to perfect and its a lot more challenging than doing pretty artwork or 3D models. The money is in the flashy graphics and effects, not in gameplay.

[player1]

So basicaly, laws that protect copyrighted material, inlcuding AI source code, pretty much slow down AI development, since most programmers need to "reinvent the wheel" with every new game.

[Deus ret.]

F*** intellectual property! Not just for the sake of better gameplay.

[player1]

That why we should make petition for releasing of source code for STW. If not now, then at least in a few years.

[LittleRaven]

There haven't been too many titles lately with a really good AI.

In addition to what Red pointed out, it's worth mentioning that the more complex the game, the harder it is to do a good AI. Games are getting more complex, and AI programmers are struggling to keep up.

The TW series has it even worse, because you effectively need TWO AIs, one for the strategy map and one for the tactical map. Must be a nightmare.

[Red Harvest]

So basicaly, laws that protect copyrighted material, inlcuding AI source code, pretty much slow down AI development, since most programmers need to "reinvent the wheel" with every new game.

I don't think that is really the issue. The problem is when you create a new system for playing a game, you need to recreate the opponent. There are people who write AI and carry the basic concepts from game to game. People want more and more complex games, but the AI is not evolving at the same rate. Each time you get a new system to work with, you need to rebuild your tools. It is much the same in other software, which is why we tend to standardize.

Think of it like writing any tool developed to run on a specific app. As long as you are using the base app, you can continue to use the same tool with some modifications. As you add layers of complexity to the app, you can add complexity to the AI. When you start over with a new application (new engine, new programming language, etc.) you are starting over with the AI. So you can use prior general knowledge to rebuild. That gets you started, but it doesn't necessarily keep up, it just gets you back ot the same point...without the new features fully explored. The more complex a game is, the more attention that is needed to develop AI.