Question on AI difficulty

[ZOOMIE1980]

Long time reader, seldom poster. But I have one question concerning AI difficulty settings.

Having played countless computer wargames over the past 20 years, it seems that increasing AI difficulty usually means the computer player's weapon systems simply get a greater than 1 multiplier. Even in Uncommon Valor it seems that the AI really isn't any smarter or uses more aggressive tactics or uses smarter tactics, but just has more effective combat resolution.

So what's the deal with this game's AI. If I smack the hell out of the computer AI player at "historical" difficulty, will increasing the difficulty just make the AI's weapon systems more effective? Or will it play "smarter" or more aggressively?

As a case in point in UV. I noticed, for instance, the AI seldom places more than two carriers and a light carrier in a TF. All I have to do is mass my carriers in a 4 heavy carrier TF and I'll knock out any carrier TF the computer ever throws at me! And the Japanese used six heavy carriers in the Pearl Harbor raid and had 4 at Midway, so that is a historical Japanese tactic.

I noticed playing the #18 scenario in UV, the one where the Japanese seem to have the most forces available the quickest, all I have to do as the Japanese player is wait until I have at least four main carrier and I can wipe out any American carrier TF they ever send at me, regardless of AI difficulty setting. Same thing the other way around. As the American, I wait until I have four or five carriers, mass them together, and go after the Japanese because they never seem to send more than two heavies and one or two lights at me, not matter what.

So I just mass my carriers, and then pick off the enemy carrier TFs one at a time until the enemy had none left, and then start my landing assaults with full air superiority. And the only thing increasing the AI difficulty seems to do is to make their weapons more effective. Pretty lame actually.

And the AI almost NEVER faints? What's with that?

[Mike Scholl]

Before you get too depressed, take a good look at some of the other AI's out there.
I've played ground combat games where the dumb thing couldn't even follow the
game's own road net. Given that this one has to try and emulate air, sea, land,
supply, and multi-tasking cooperation between them, it's amazing it even works
at all. Short of a major breakthrough in computer technology, it's not going to get
a lot "smarter".

Just view it as a learning aid for games with real opponants. It will be a better
opponant in smaller scenarios where it can have a more complete "pre-programmed"
set of options, but as a strategic opponant it will never be the kind of challange you
want.

[ZOOMIE1980]

Actually, I have written combat simulations for all three service branches over the years. The AI code in almost every case was over 80% of the total lines of code in each project.

But like the long winded discussion about the various portages at Pearl and what was "realistic" and what was not about the results, you only have so much room in the budget for stuff if you ever want to release something.

But for me, I'd give up graphics, detail, and some depth for a really intelligent AI. But then, I seldom play e-mail or hot-seat games.

[DoomedMantis]

I highly recomment playing PBEM. If time is an issue there are still players willing to play, just state it at the start

[pertsajakilu]

ORIGINAL: DoomedMantis

I highly recomment playing PBEM. If time is an issue there are still players willing to play, just state it at the start

But that is not the point. I personally want to play whenever I like or can, online ( or pbem ) playing is not same thing. I have a real life outside and I don't like to be tied at the puter because of games or because there is a precice time when You have to play because it suits Your pbem companion.. these are only games. Not life.

What I am afraid of is because of online there will be much less resources in development of AI in future games.

TV.
Pertsajakilu

[crusher]

i really hope that it will have at least a fair A.I. because here in the P.I. it is almost impossible to play on line or pbem due to internet connections are bad or even impossible for long periods of time due to power failures and lousy dial up connections

[Mike Scholl]

ORIGINAL: pertsajakilu
But that is not the point. I personally want to play whenever I like or can, online ( or pbem ) playing is not same thing. I have a real life outside and I don't like to be tied at the puter because of games or because there is a precice time when You have to play because it suits Your pbem companion.. these are only games. Not life.

What I am afraid of is because of online there will be much less resources in development of AI in future games.

Pertsajakilu

Find an email opponant who doesn't demand instant turn-around. Then you can do
your turn at your convenience. But in a game this complex, the AI would have to be
programed with a response for each of millions of potential actions on your part to
provide the kind of game opponant you want---and the CD would need to be 4 feet
in diamater. It's a nice dream..., but it ain't gonna happen any time soon.

[pasternakski]

ORIGINAL: pertsajakilu
What I am afraid of is because of online there will be much less resources in development of AI in future games.

It remains true that the vast majority of computer wargames are played solitaire. Someone on one of the UV threads mentioned that he starts scenario 17 as the Japanese, then switches to the Allies later on when the force levels begin to shift in their favor. I suspect, also, that there is a lot of head-to-head solitaire play out there in parallel to what guys used to do (and probably still do) with the old paper-and-cardboard games.

If AI development does not parallel improvement in game design, it may well kill this hobby. The market for PBEM and online players is just not that large. Some developers have decided that this is so (most notably Brad Wardell over at Stardock) and are trying to develop ways of making the AI more creative and competent.

I can only say this. Most of the successful games in terms of sales in recent years have been the ones that were designed specifically with the solitaire player in mind (the best examples being Galactic Civilizations and the Civilization series). A concerted effort by developers may lead to breakthroughs in AI design. While my preference is playing against humans, I would welcome such a development and would probably increase my solo playing time as a result (not to mention spending more money on games that pay attention to serious AI design than on games that do not).

There are times when your own head is the only place to be. This is one of the reasons the world became so computer-centric: it used to be a boy and his dog could get away from the world, then it became a boy and his pooter. A hobby that seeks to survive needs to take into account the characteristics of its intended audience and satisfy the desires of those potential customers.

I know that a lot of this is non-verifiable conclusory crapola, but it's what I think I see out there (of course my high school English teacher warned me that I would always have to be careful about seeing the emperor's new clothes, too).

[Mike Scholl]

PASTERNAKSKI I would guess that in the long run you are basically right in
many of your conclusions..., but it's going to limit the kind of games you can
play for a while yet. A grand strategic effort like WITP just has too many
variables and interactions for the level of programming and capabilities of
home computers. Even the ones you mentioned that were really good at
providing competition (Thank God for Sid Meier) were still eventually "sol-
vable" and became a "rote exercise". Only a human opponant can currently
provide the "unexpected, off-the-wall" efforts that really tax your brain. So
my feeling is that email is still the best available choice. You don't have to
"schedule" playing, and you can have several games going at once if you
want. And the "connection and speed problems" that plague online games
can usually be gotten around through compression software. I'm all for
designers striving to create better AI's (and down on some that never seem to
try to improve (John Tiller comes to mind), but if you want big intricate games
we're just going to have to settle currently for an "acceptable" AI---and we all
have to define that term for ourselves.

[byron13]

Pasternatski, what you say about the AI is absolutely a bunch of non-verifiable conclusory crapola, and I absolutely agree with every word of it. Playing solo is about the only thing you can do these days that is not dictated by some kind of schedule. Schedules at work, schedules at movie theaters - heck, even tv is dictated by a schedule. It's nice just to play at my own pace.

PBEMs are unbeatable for challenge, but there is inevitably a divergence between the (gack!) schedules of the players. There is nothing so frustrating as being one turn away from delivering the master stroke attack and your opponent suddenly has a large project at work that will take two weeks to complete. Or, you lose interest in a game - not because things are going poorly, but just because you're suddenly in the mood to play a modern or classical strategy, sports, or fantasy game. Against the AI, you just shelve the game for a year; against a PBEM opponent, you're pretty much forced to continue against your druthers.

Pasternatski's right. The AI industry-wide must improve. PBEM or on-line gaming will surely increase, but solo gaming will always be there.

And, yes, Tiller's AI is awful. I'm not sure his Panzer Campaigns have an AI much more sophisticated than moving units to a predetermined location. You can play those games about twice on each side before the only challenge is to beat the time it takes to obtain an absolute victory.

[pasternakski]

I think what you say is exactly true, given the current state of AI art. Still, it seems to me that fruitful avenues of further development have yet to be taken. In UV, for example, the "level of difficulty" setting hasn't advanced at all from the days of Carrier Force or Solomons Campaign.

I can't believe that, with all the memory available, more flexibility and at lest simulated creativity ("Hey, let's build in a surprise twist that triggers on certain predetermined circumstances").

I just hate to think that AI design is a dead end, that's all.

[Damien Thorn]

Does anybody remember the game "Archon" back in the Atari 800 (and maybe Apple II) days? It was a chess-like gmae with tactical combat. The AI was very good, both in where to move and the execution ofthe combat phase as well. All that in a 48k game.

[Joel Billings]

We have removed combat cheats from the hard level. At the "historical level", the AI has a few rules it does not have to follow (due to it doing things differently than the human player. At the hard level, it gets various benefits like increased production, better pilots, etc., but it does not get a combat formula cheat. At the very hard level, it gets everything at hard level (sometimes with more benefit) and it gets some combat cheats (but less than in UV). I suggest everyone play at hard level after their first game (very hard once you really know the game well). You will not notice the benefits the AI gets, but it will help the AI out. The AI does not get smarter at any level. These games are so complex that the smartest we can make the AI is still dumb compared to any decent player after their first game.

[Capt. Harlock]

A suggestion: at higher difficulty levels, the AI should get better intelligence. Since giving the computer more in quantity and quality is non-historical anyway, why not give more insight into what the human player is doing and where? I don't think there needs to be perfect knowledge of all TF's (although some past games have had that), but making the AI harder to surprise might well have the apparent effect of making seem more intelligent.

[Damien Thorn]

ORIGINAL: Capt. Harlock

A suggestion: at higher difficulty levels, the AI should get better intelligence.

I'd rather give the computer more stuff than get jumped everytime I turn around because the computer knew exactly where I was going.

[pasternakski]

ORIGINAL: Damien Thorn
I'd rather give the computer more stuff than get jumped everytime I turn around because the computer knew exactly where I was going.

But, see, that's the whole problem. If there is an AI, it shouldn't need more stuff or better intelligence. It should work better with what it has.

[ZOOMIE1980]

ORIGINAL: pasternakski

I think what you say is exactly true, given the current state of AI art. Still, it seems to me that fruitful avenues of further development have yet to be taken. In UV, for example, the "level of difficulty" setting hasn't advanced at all from the days of Carrier Force or Solomons Campaign.

I can't believe that, with all the memory available, more flexibility and at lest simulated creativity ("Hey, let's build in a surprise twist that triggers on certain predetermined circumstances").

I just hate to think that AI design is a dead end, that's all.

Problem with AI advancement is the current economic model is centered on "eye candy" even in turn based wargames. A lot of bleed over from the RTS and First Person Shooter genres. Very "pretty" maps, animations, obsessiveness with minutia and historical accuracy. Plain and simple, AI's don't sell games these days. And you start a company to write wargames and make money, not as a hobby.

There actually is quite a lot going on in the world of artificial intelligence in places like NASA, the US Military, home electronics ("smart applicances"), and on college campuses, and a lot of it is available in the public domain and runs on higher end PC's a lot of gamers already have. 500GB of HD space and 1GB+ RAM are all fairly common these days with gamers. So computing capacity isn't the problem. Maybe when game shop finally realize top end AI's can sell as good or better than the eye candy, they'll actually hire the expertise and put in the man hours.

I can see from this thread I am not alone in wishing that a new state of the art wargame come with a state of the art AI engine.

[byron13]

Yes, high AI levels ideally would do nothing more than dial up the competence of the opponent so that it makes better use of its resources; make it more experienced, if you will. The problem is developing relatively cheap AI that either (i) learns by itself or (ii) is provided with "intelligence" by providing an enormous amount of logical decisions. Hasn't been done yet, and Matrix is not going to be the one to do it.

I think giving the AI more intelligence on your movements at a higher setting would also be relatively expensive to code. Of course, I have no idea how the computer "sees" you now or reacts to units seen. But I'm not sure it is as easy as doubling its search range. In any event, just "seeing" that you are massing troops and carriers in the South Pacific as opposed to the Central Pacific may or may not be sufficient to trigger a response - depends on how the AI is already coded. Or, if it "sees" that all of your carriers are damaged and sitting in San Diego, does this trigger a response? I simply don't know how they've limited the AI when it is suffering from the Fog of War. This would determine what would happen when the fog is lifted.

The easiest way to dial up the competition is simply to give the computer more stuff or cheat on combat results, and that's what they do.

[pasternakski]

ORIGINAL: byron13
The easiest way to dial up the competition is simply to give the computer more stuff or cheat on combat results, and that's what they do.

Laziness - that's how Detroit lost the battle of the sedan to the Japanese.

[byron13]

Yeah, but you really can't expect a company with resources as limited as Matrix has to be the company that designs the first intelligent AI. I'm just wondering if there is a kind of generic intelligent AI that can be developed that could be adapted to almost any game - or any strategic game. That way, one company could develop it and then license it. But, I admit I don't know - unless I'm arguing with mdiehl. [:'(]