(Long Post) Where are games at?

[BoredStiff]

ORIGINAL: Joe 98

The content of you’re post is all about the AI

So I too will stick to the AI.

Playing against the AI is like playing a singles match of tennis against a brick wall

Playing against the AI is like having a date with a cardboard cut out woman

Playing an AI is only for practice so you can learn the game mechanics.

Once I learn a game I only play against human opponents.

And that's about all a person can say in regard to the constant whining about bad AI's.

The worst part of it is, that developers might sometimes not implement certain features of a game because it might be too difficult (or impossible) to program the AI for it. In other words, an aspect of a game that might work very well in regard to PBEM play might be compromised, or even left out altogether, because AI would not be able to adequately follow the routine.
IMO, wargames, especially those for more than two people, should be designed strictly with human-to-human play (PBEM, TCP/IP, hot seat) in mind, with the AI added as an afterthought and mainly as a teaching/practicing aid and nothing more.

[Terminus]

If only it was that easy... There's apparently still PLENTY of people who like playing the AI, and who will whine endlessly about being "second-class citizens", when the AI doesn't work as well as they think it should. Sounds stupid? Come down to the WitP forum, and see what we on the AE team have to deal with on that count...

[hazxan]

An excellent topic - I hope the (few remaining) designers are watching!

There are many good points made above. However, one old chestnut keeps cropping up: "AI is very very hard". Very true, but is it any harder than the maths used for 3d graphics? I really don't think so. Take a look at any book on 3D graphics and you'll see how scary the whole thing is. I program mission critical hard core engineering apps as my day job, yet 3D graphics maths frightens me!! The point is that the industry has spent HUGE amounts of time and resources on graphics in the past 20 years or so. And nothing at all on AI. Why do we have Graphics cards but not AI cards????

Now I don't think "the industry" us evil or daft. They're just fulfilling the consumer demand for more realistic visuals. Plus I guess it's easier for a techy to get money to develop flashy graphics, than for an AI.

I agree with Jim_H on that for all the complaining of AI, MOST gamers find the AI's in MOST games to be a perfectly good enough challenge. I suspect there's a small but noisy minority of virually 'full time' gamers who use every gamey device they can find to beat the game in a day. If you don't live on the forums and don't play 40 hours a week just play most games for pleasure, you'll get months before you can beat the AI on maximum.

Do we really want to be regularly thrashed by an AI, anyway? Continually being beaten would probably be a miserable experience and there are a lot of games that have flopped through being too difficult. And for historical games, do we really want the best AI or one that makes the same mistakes as the original commanders?

As for cheating AI's. Well the ultimate cheating AI would just shut the game down and reload when it was losing - 'cos thats what most humans do!

For wargames/strategy games, something Veldor said on another thread springs to mind. Too paraphrase: "developers spending 2000 hours researching the uniforms then cobble together an interface in 2 hours". Some truth in that, so again we get the games we deserve. The devs are doing that knowing that someone will nit pick the game (or should that be 'historical recreation') to pieces. So to achieve an OCD level of replication, the AI suffers, the GUI suffers and worst of all the whole game design suffers.

To end on an upbeat note. There really is demand for serious strategy games out here, the designers just have to tap in to it. For ages now games (of all genres) just get bigger and bloated yet 9 times out of ten, the original was the best. There's still millions of us who will probably never play an online game.

In contrast to overly complex, tedious, 'chore-ware' games with a customer base barely in 4 figures, there are many strategy games that are good quality and selling by the bucketload. Games like Advance Wars on the DS & GBA, Memoir '44 board game, Total War series, even the original Panzer General (wasn't this the biggest selling war game of all time?), many of the 'new wave' of Euro-boardgames - all of them good strategy games.

My point is that these games have serious strategic elements, but they succeed because they "Keep it Simple Stupid". They have accessability designed in from the start. IMHO too many of the strategy and especially wargames aim for 'bigger is better', because this is what their fans are shouting for. The most common repeated statement on AI in any forum is that "you shouldn't expect a decent AI because AI is too difficult to program and you should be playing online anyway". So why bother with AI.

It amazes me when I lurk on, say Paradox games forums, how many post repeatedly about making the game BIGGER!! More Provinces!! Bigger Timescale!! More detail - we need Tactical battles!!!! How much time are these people spending playing games? More importantly, are there enough of them to base your entire marketing strategy around targetting them?

And the real answer in 2 words:

RADIANT AI

[;)]

[hazxan]

ORIGINAL: Unhappy
5) Finally, and this may seem hard to believe, but I love World of Warcraft. Obviously it is extremely popular and monetarily lucrative. All of which leads me to wonder whether the strategy/war-gaming folks are examining this phenomenon with an eye to jumping on the bandwagon? And I don’t mean with first-person-shooter action games. If we must confront human instead of programmed adversaries I think WoW offers an excellent model. A persistent world, with action happening continuously – sounds perfect for a massively multiplayer online strategy game to me.

Interesting idea. Kinda like "Animal Crossing" with tanks and guns[8D]

[marcusm]

I think everyone needs to take a good look at what Brad Wardell(Galciv2 dev) does. He is without
doubt the best AI designer in the industry. As he told me, the AI has to be part of the core
design process and not something you slap on in the end. Usually bad AI is a simple question
of not dedicating enough resources for it.

I also agree that a game has to choose focus. Either single player or multiplayer. Not both. Most
try to both and fail.

marcus

[Kuokkanen]

5) Finally, and this may seem hard to believe, but I love World of Warcraft. Obviously it is extremely popular and monetarily lucrative. All of which leads me to wonder whether the strategy/war-gaming folks are examining this phenomenon with an eye to jumping on the bandwagon? And I don’t mean with first-person-shooter action games. If we must confront human instead of programmed adversaries I think WoW offers an excellent model. A persistent world, with action happening continuously – sounds perfect for a massively multiplayer online strategy game to me.

There are some MMO type strategy games. Not many, and some of them are off-line now, but they exist. From commercial games I can name Shattered Galaxy, from open source games I name MekWars.

How many computer wargames do you know of today that actually have a "winning" objective that they need to reach other than just keep you from reaching yours?? Name em.

Total War serie. And you're right about games being too difficult: I almost ripped my hair and stopped playing game when I couldn't get over with Shogun tutorials. I played and replayed 10+ times tutorial scenario where player's objective is to kill charging swordsmen with group of archers. In fact that seems like with reversed roles what ravinhood rants about: I can just delay swordsmen's victory.

BTW, can someone tell me how to win that scenario? I've tried shoot & scoot, different formations and stuff with no go...


As for digital wargames in general, it is my opinion that wargames are the cornerstone of computer games. Some call FPS games the corner stone, but nowadays those are converted and even exclusively made for consoles. If there ever comes time when games sell more for consoles than for computers, then wargames will reign on computers.

Finally, digital wargames can be much cheaper than... analog ones: I've spent over 300 €uros for Classic BattleTech and I'm going to spend much more for that one game. Some people call CBT cheaper wargame to get and keep up to date compared to some others, like Warhammer 40k. Comepared to these prices, $80 + shipping for SPWAW General's Edition is dirt cheap and game has enough game hours even for those who play 40 hours week. And if not, then make some more scnearios and campaigns [;)]

[leastonh1]

ORIGINAL: NotaGrog
I agree with Jim_H on that for all the complaining of AI, MOST gamers find the AI's in MOST games to be a perfectly good enough challenge. I suspect there's a small but noisy minority of virually 'full time' gamers who use every gamey device they can find to beat the game in a day. If you don't live on the forums and don't play 40 hours a week just play most games for pleasure, you'll get months before you can beat the AI on maximum.

Yes, that's exactly my point. I really do think (and would like to be persuaded otherwise) that the majority, not all, but most of those gamers who moan about AI use exploits of some description if they are being totally honest. They may play a scenario ten times and eventually win. Or quick save/quick load to get past a boss. Once they figure out that standing in a certain spot and firing at a set interval, they can kill it easily. I've done it lots of times and then thought to myself "well, that was easy. what crap AI this game has!". Take wargames. How many RW commanders have the luxury of sitting there with a cup of coffee and unlimited goes to take the next ridge or gain a beachead? We take it all for granted and what about intel? We are spoonfed all the info we need and probably vastly more than anyone really gets.
[Edit]Here's a perfect example...I know, because I've done this myself and other people have posted about it. At the beginning of a game of AT, you can use trucks or armoured cars as single units to expand your territory. You capture cities by just driving a single unit onto them and you get lots of production and research bonuses for doing that. Vic has addressed this with an option to have these existing town/cities already populated with a "neutral" AI opponent. But, to do this with trucks is simply exploiting the game mechanics in order to give yourself a better chance of winning. So, when you've beaten the AI on turn 15 after taking almost the whole map in the initial few turns through this exploit, do you remember this point? Or, do you sit back and think that it was just too easy to beat and that the AI must therefore be rubbish? Memory is very selective. Try beating the game without doing this and you have a much more difficult task on your hands. How many people, being totally honest, have done this at least once? I'm willing to bet the majority have.[/Edit]

ORIGINAL: NotaGrog
And for historical games, do we really want the best AI or one that makes the same mistakes as the original commanders?

That's a brilliant question and one I can't answer! The best AI on a computer with the amount of power we have today would beat most/all humans hands down every time. To make the same mistakes as the RW counterpart would probably be boring. What's the point of knowing in advance exactly what's going to happen? I wouldn't spend money on the game that does that.

ORIGINAL: NotaGrog
For wargames/strategy games, something Veldor said on another thread springs to mind. Too paraphrase: "developers spending 2000 hours researching the uniforms then cobble together an interface in 2 hours". Some truth in that, so again we get the games we deserve. The devs are doing that knowing that someone will nit pick the game (or should that be 'historical recreation') to pieces. So to achieve an OCD level of replication, the AI suffers, the GUI suffers and worst of all the whole game design suffers.

Which is my biggest bugbear with wargamers. I don't think there's anything wrong with trying for accuracy, but too many wargamers get all anally retentive about it and the devs would rather waste these 2000 hours researching the correct colour socks of a particular historical regiment than spend that time making a better game experience. Sure, some will argue that the wrong colour would spoil said experience, but where do you draw the line? It's a bloody game when all's said and done! You want real life? Go join the military. This aspect of wargaming really gets on my nerves. Can you tell? [;)]

ORIGINAL: NotaGrog
In contrast to overly complex, tedious, 'chore-ware' games with a customer base barely in 4 figures, there are many strategy games that are good quality and selling by the bucketload. Games like Advance Wars on the DS & GBA, Memoir '44 board game, Total War series, even the original Panzer General (wasn't this the biggest selling war game of all time?), many of the 'new wave' of Euro-boardgames - all of them good strategy games.
My point is that these games have serious strategic elements, but they succeed because they "Keep it Simple Stupid". They have accessability designed in from the start. IMHO too many of the strategy and especially wargames aim for 'bigger is better', because this is what their fans are shouting for. The most common repeated statement on AI in any forum is that "you shouldn't expect a decent AI because AI is too difficult to program and you should be playing online anyway". So why bother with AI.

Again, I agree with you. Panzer General was a great wargame, as is Memoir '44. Some say Memoir is too light and fluffy. But, at the same time it's games like this that keep the wargames on the map, so to speak. If there weren't lighter games than WitP, we'd have no wargame genre. Panzer General was very simple, but had some depth and appealed to players at all levels. Bigger isn't better, it's just more space with nothing in it.

I think one of the best AI's I've encountered in any wargame (recently or in the distant past) is Advanced Tactics. Set it to AI+ or AI++ and it's very, very hard to beat. I know Vic is putting a lot of emphasis and work into tweaking the AI for this game and that's precisely how it should be. If he can do it on his own, why can't other devs who have millions of pounds (dollars, whatever) and a team of other devs do it?? Because they go for eye candy, or a thousand scenarios, or perfect historical accuracy or whatever else the crowds are clamouring for, instead of concentrating on the rounded quality of the experience of trying to beat something with artifial intelligence.

Regards,
Jim

[ravinhood]

To me nowadays AI = Ain't Intelligent
 
Something I've discovered playing M2TW is that when you play on VH/VH settings the AI gets an abundance of resources more than you which for me is fine I like that. But, here's the problem with that. Because they do not protect their cities in the rear, once I get over there to them I'm getting 71000 gold from sacking them, thus giving me enough money to continue my rampage in their rear. Scotland had like 10 provinces to my 6 and I got a stack in their rear (up in Scotland and England area) and they only had 3 or 4 units in each city or castle. They had several marching stacks, but, weren't doing anything with them except piecemeal attacks with partial stacks against me. I made over 200k gold in just 3 city/castle attacks on them. So here is where CA forgot to look at the amount of gold the human player gets from sacking AI cities/castles. They should have programmed in that mathematically the human player could only get 1/4th the amount when the setting is on VH instead of the full amount. 50k would have been a nice chunk, but, it would have slowed me down considerably in my march against the Scots. Plus, I wouldn't have been able to rapidly build up my forces and deploy more stacks from that 200k sacking money I got. Of course it's not just Scotland, every AI that I sacked gave me large sums of gold to continue my global conquest. Once I have 3 full stacks out on the map rampaging there is just nothing that can stop me. It can only merely DELAY me after that. Once I have ten cities and/or castles it's the same. The AI can never muster enough strength or resources to out smart me or stop my advance it can only DELAY. I've seen this in 100 games and the developers don't make AI's that gang up on the human player when HE's WINNING. This is all I really want extra. When I'm ahead and/or have reached that 10 city/castle mark in the game I want to see the AI's ally with each other and come after me. I want to see it TRY to crush me with a HORDELIKE response. I mean the Golden Horde does it, so, why can't the regular AI do it also? I totally enjoy when the Horde arrives in MTW as it's the only ultimate challenge of the whole game. I also like the revolts and the return of the kings in the old MTW, I thought that was excellent to keep the game alive and challenging in the middle to end game. But, RTW and M2TW took that fun out of the game.
 
Jim calls it "EXPLOITING"? hahah how can you exploit what the game does itself as a feature? It's not exploiting it's bad beta testing and bad quality control finding these issues before the game is released. Exploiting is saving and reloading after you've lost a huge battle, city/castle or leader. Playing the game as it is made to be played is not exploiting it. How can you exploit that the dumb stupid AI sues for peace and offers you 10000 denari and then on the following turn or even on the SAME turn sometimes redeclares war on you? That's not the players fault that is bad game design, bad beta testing and bad quality control.
 
I know if I can take a game an in one week find so many flaws and exploits it's not funny that some beta tester and game programmers should find them also. Many are so blatant it's rediculous. These game companies need to HIRE beta testers who actually TEST the game for all it's quirks and exploits, not just these freeloaders who just want a free game in the end. I bet half or more of beta testers don't even send in very many bug reports. They just play and play and play and aren't doing their jobs. But, if beta testers were paid to test and then didn't do the service they could just FIRE them and get someone who would actually WORK as a beta tester. When I beta tested for SSI I did all kinds of things bassackwards. I tried to do things that really most people wouldn't think to do, but, still I did them because there could be a bug or exploit in it. I was always complimented by Dave Landrey for my detailed reports each week and show of dedication for finding bugs and flaws. THough they NEVER listened to my game design input. I didn't even beta test for the free games. I did it because I wanted a good solid bug free non-exploitable game when it was finished. What I see come out today out of the box just really amazes me and pisses me off at the same time that publishers allow this kind of crap to goto market. Beta testers aren't doing their jobs, quality control isn't doing theirs and publishers don't care anymore like they did back in the SSI days. Just more reasons the AI suks in most games today.

[Veldor]

ORIGINAL: marcusm
I think everyone needs to take a good look at what Brad Wardell(Galciv2 dev) does. He is without
doubt the best AI designer in the industry.

Thats a pretty bold statement unless your a leading AI expert (Don't know about you but I certainly am not one). The only flaw with that statement that I will point out at this time though is that the GALCIV2 AI is a cheating AI. Therefore the only real setting that is relevant is the base setting. And assuming you believe it doesnt cheat at that level (which though it doesnt in the way that the other leves do its still "aware" of things it probably shouldnt be) its a pretty average AI to beat (I always won at that level anyway). By definition this must be true or there would be no need to have so many other "cheating" levels. Who really cares that I never ever win on the toughest level (Such would be the same in any game if I gave the AI 100 times the resources I get). That said though I would personally rate the GALCIV2 AI very high on the sliding scale of strategy/wargame AIs and thoroughly recommend the game series to anyone. [:D]

As he told me, the AI has to be part of the core
design process and not something you slap on in the end. Usually bad AI is a simple question
of not dedicating enough resources for it.

Hey Ive been posting that for years! He must be a smart guy. Though one would hope things like that would be common sense. Apparantly such is not the case. Maybe thats because humans aren't really that good at multi-tasking and tend to program in a linear fashion (ie Interface first, then graphics, then logic, then network/pbem, then ai, then sound, then demo/tutorial) finishing most of one at a time before moving on to the other. This is probably where the resource thing comes in though in my case I'd mean people, like a dedicated AI programmer. And as pointed out elsewhere thats just not practical for most in todays wargamming world.

I also agree that a game has to choose focus. Either single player or multiplayer. Not both. Most
try to both and fail.

I don't think this applies to strategy/wargames in my opinion. Unless you mean two-player vs more than two-player. But in terms of either being a "one against the AI" game only vs a game with no AI altogether there is no way thats accurate. A HUGE number of people, even ones that would play multiplayer, want an AI in a strategy/wargame. Even those that don't care long term at least want to learn with one or experiment with various strategy techniques. Its not the same thing as with FPS games or something like that.

[Veldor]

ORIGINAL: Jim_H

At the beginning of a game of AT, you can use trucks or armoured cars as single units to expand your territory. You capture cities by just driving a single unit onto them and you get lots of production and research bonuses for doing that. Vic has addressed this with an option to have these existing town/cities already populated with a "neutral" AI opponent. But, to do this with trucks is simply exploiting the game mechanics in order to give yourself a better chance of winning. So, when you've beaten the AI on turn 15 after taking almost the whole map in the initial few turns through this exploit, do you remember this point? Or, do you sit back and think that it was just too easy to beat and that the AI must therefore be rubbish? Memory is very selective. Try beating the game without doing this and you have a much more difficult task on your hands. How many people, being totally honest, have done this at least once? I'm willing to bet the majority have.

Sorry but I respectfully think your wrong here. Since those are the rules by which the game was setup, and thats what every single other human player in games I've played has done, there is no reason why the AI shouldn't be capable of making the most intelligent first few turn moves also. In fact this is the easiest part of the game to make an AI good at (The first few turns). So its a double failure. While I will not argue that the game design is flawed in that regard (Its a big reason why I now refuse to play random games until its fixed) given that its the current rulesset the AI should understand it. That it wasn't caught sooner is just proof of poor playtesting since nearly everyone agrees this is a big issue. Or maybe its back to that most players/playtesters focus too much on historical nitpicks instead of the things that should really matter (like gameplay rules, UI, and so on). To take the otherside, briefly though, how do you get enough playtesters for games that aren't going to sell that many copies anyway?? You'll loose half your potential sales. [:D]

Which is my biggest bugbear with wargamers. I don't think there's anything wrong with trying for accuracy, but too many wargamers get all anally retentive about it and the devs would rather waste these 2000 hours researching the correct colour socks of a particular historical regiment than spend that time making a better game experience. Sure, some will argue that the wrong colour would spoil said experience, but where do you draw the line? It's a bloody game when all's said and done! You want real life? Go join the military. This aspect of wargaming really gets on my nerves. Can you tell? [;)]

The squeky wheel always gets the oil. Plus its a matter of standards being set in the wargame industry. The standard has been set that a poor UI and poor AI is "more ok" than "minor historical inaccuracies".

I think one of the best AI's I've encountered in any wargame (recently or in the distant past) is Advanced Tactics. Set it to AI+ or AI++ and it's very, very hard to beat.

CUZ IT CHEATS!!!!!!!!!
Give the computer 5 extra cards in a regular poker game and I guarantee you'll loose everytime. A great AI that does not make!
Specifically regarding AT I'll reserve my judgement until the patch. Given the AI currently ignores many of the units in the game entirely I don't think you can call it the "best AI" even if it may be "good".

I know Vic is putting a lot of emphasis and work into tweaking the AI for this game and that's precisely how it should be. If he can do it on his own, why can't other devs who have millions of pounds (dollars, whatever) and a team of other devs do it?? Because they go for eye candy, or a thousand scenarios, or perfect historical accuracy or whatever else the crowds are clamouring for, instead of concentrating on the rounded quality of the experience of trying to beat something with artifial intelligence.

Because the game isn't mucked down in layer upon layer of mundane historical complexities. Because the developer smells of having a personal pride in not just the historical aspects and gameplay but of the code itself. Because its probably not just all about the money (or lack thereof). It's not Advanced Tactics that has all the elements of success, nor any wargame, its the developer himself that has to (and in this case does) have the necessary components. Though the game is not without its flaws, its the primary reason I believe why the game is good in the first place and will only get even better in time. This unfortuneately to me, is the exception, not the rule. That AT can be looked at as such a great game in the first place is proof of where standards are at in computer wargaming. Though I can't really disagree with its stance as an exceptional product, I do on the otherhand kinda wish standards were higher (for everything from its graphics to its hard to manuever editor to its somewhat limited UI to the AI issues, lack of decent included tutorials, and so on). It takes more than one game to raise the bar though...

[marcusm]

I am not a leading AI expert but it is the general opinion that his AI design
is way ahead of anyone elses.

Giving AI + is not exactly cheating. It is a handicap. Handicaps are perfectly viable
in eg. Golf to even the field. Part of the game rules even. I think it's a great way to make the
match more interesting. The Civ4 BTS AI is good enough that i haven't even started with the bonuses to AI yet .

[hazxan]

ORIGINAL: Veldor
It's not Advanced Tactics that has all the elements of success, nor any wargame, its the developer himself that has to (and in this case does) have the necessary components. Though the game is not without its flaws, its the primary reason I believe why the game is good in the first place and will only get even better in time. This unfortuneately to me, is the exception, not the rule. That AT can be looked at as such a great game in the first place is proof of where standards are at in computer wargaming. Though I can't really disagree with its stance as an exceptional product, I do on the otherhand kinda wish standards were higher (for everything from its graphics to its hard to manuever editor to its somewhat limited UI to the AI issues, lack of decent included tutorials, and so on). It takes more than one game to raise the bar though...

I agree in that I find AT a very playable game, despite it's many quirks, not because of them. The long AI turn is my only real complaint, but hardly anyone else seems to care so......

On AI in games I have a question regarding the Panther Games HTTR and COTA. I've never played them (no demos) but they seem to have a lot of real ground breaking ideas. For a start, 'no hexes' looks excellent - surely all computer strategy games should do this?

However it's the command structure that I find most interesting. My understanding is that you take control at a higher level of the OOB and sub-formations use their own initiative (the AI) to achieve your goals. What I wonder is surely that means the AI is basically playing itself as the AI opponent will be fighting back using exactly the same AI routines. Fair enough the high level aim has been set by the human, but I'm still too doubtful to make the purchase. From the forums, it seems that opinion is that the AI is not too bad but not too good either.

So I'm wavering because to me it's a good thing that you don't micromanage your whole force - also very realistic. But then the AI takes over and well, maybe this isn't quite so satisfying. Any comments appreciated!

[leastonh1]

ORIGINAL: Veldor
Sorry but I respectfully think your wrong here. Since those are the rules by which the game was setup, and thats what every single other human player in games I've played has done, there is no reason why the AI shouldn't be capable of making the most intelligent first few turn moves also. In fact this is the easiest part of the game to make an AI good at (The first few turns). So its a double failure. While I will not argue that the game design is flawed in that regard (Its a big reason why I now refuse to play random games until its fixed) given that its the current rulesset the AI should understand it. That it wasn't caught sooner is just proof of poor playtesting since nearly everyone agrees this is a big issue. Or maybe its back to that most players/playtesters focus too much on historical nitpicks instead of the things that should really matter (like gameplay rules, UI, and so on). To take the otherside, briefly though, how do you get enough playtesters for games that aren't going to sell that many copies anyway?? You'll loose half your potential sales. [:D]

Sorry Veldor, I don't think I explained myself properly there. What you said above is absolutely correct, but not what I meant. What I was trying to say was that it's a game mechanic that can be exploited to win the game as opposed to a potentially real life type scenario. By exploited, I don't mean cheated, I mean a design feature or flaw or whatever in a game or AI that wouldn't normally/ever be replicated in real life. Take NotaGrog's point about the choice in programming AI to be very good, or to make the same mistakes as a historical person/leader/army etc. Do you go for an AI that has all the answers and just works by numbers to get the best outcome no matter what or one that is programmed to mess up a specific aspect of it's behaviour in order to replicate something historical, but is more "human" and fun to play? Chess programs can beat just about everyone on the planet except the top Super GM's, but you can change a setting to dumb them down to play with human-like flaws during play. Hmmm, I'm still not happy that I've explained that the right way...

ORIGINAL: Veldor
CUZ IT CHEATS!!!!!!!!!
Give the computer 5 extra cards in a regular poker game and I guarantee you'll loose everytime. A great AI that does not make!
Specifically regarding AT I'll reserve my judgement until the patch. Given the AI currently ignores many of the units in the game entirely I don't think you can call it the "best AI" even if it may be "good".

I don't know whether it cheats or not as I know nothing about programming AI or the algorithms behind it. It's one of the best that I personally have seen in a long time. It's just my opinion and I wasn't stating for a fact that it is THE best. Read my post again. [:)]

Regards,
Jim

[Arjuna]

Great discussion guys. As a dedicated AI programmer myself I do agree that the AI has to be an integral part of the game design, not something tacked on.
 
While there is a role for scenario specific scripting, relying on this alone can only result in a fairly predictable gaming experience. The real payoff comes with "generic" AI - ie AI that can make credible decisions regardless of the scenario. For this to work the AI needs to be situationally aware - ie it has to be able to interrogate the scenario data ( eg GIS/map, entities/units, weather, objectives etc ) and make its own assessments. It then must be able to develop its own plans, issue its own orders, respond to those orders, react to events and reassess its own plans.
 
Once you have these capabilities, the AI will be able to respond in any scenario without the need for scenario specific scripting. ( Note scenario specific scripting can still be useful. )The end result is a more dynamic game, where the AI responds to developments as and when they occur. The result is infinitely less predictable and this provides a more rewarding gaming experience.
 
For a truly generic AI to work you also need to model a command system, replete with a hierarchical structure.
 
With few exceptions ( including our products ) most wargames use a two tiered command structure. By this I mean that all decisions are made at two levels, the strategic level ( ie the side ) and the tactical level ( ie the unit ). The strategic level makes decisions on what objectives to go for - ie how best to win. The tactical level manages each and all units - eg where and when to move and fire. The overall plan for a side is done at the stratgic level. This allocates units to go to X etc. In effect all units report directly to the side boss. Now many wargames complicate this somewhat but ultimately they are characterised by the fact that all units are micromanaged by the side boss.
 
The trouble with this approach is twofold. First, the behaviour of sub-forces tends to be predictable, with no variation for the particular situation faced by the sub-force. They don't delvelop their own plan tailored for the terrain and enemy they face right now. Thus the behaviour is not as realistic as it could be. Second, it makes it extremely difficult to code a powerful AI of this type that can cater for the myriad of situations that are likely to be encountered.
 
However, if you at the outset design your system to model a hierarchical command structure where orders are passed down the chain of command and where at each level local AI controlled commanders assess their situation, develop their own plans, react to events and reassess their own plans, then you have scaleability. If you combine this with a system of doctrine that manages the development of plans, the reactions and reassessments then you can have AI controlled subordinates that behave in a realistic manner. Moreover, you have a code base that can be developed more easily.
 
That is not to say that developing such an AI is easy. If it were, everyone would be doing it. But there are better paths to success in this endeavour.
 
When we set out to design our Airborne Assault game engine, we echewed abstraction wherever possible. ( And a two tiered command structure is a massive abstraction. ) We endeavoured to simulate what happens in reality rather than come up with some "slick" methodology that "looked like" it was doing the right things. At the outset this meant a lot more work designing the system, but the payoff later on has meant we can model behaviours more easily and with less complications than had we gone down the quick and easy abstraction route. 
 
Developing AI is not easy and its not cheap. We have invested around 40 years of effort into our engine, over half of that on AI. That is a huge investment in AI by industry standards. Has it been worth it? Financially, probably not. But professionally it has been very rewarding.
 
In a risk averse industry, that the triple A titled game industry has become, they are not going to take the risk and invest that kind of money on something with that kind of development risk. So that's why you will continue to see only minor improvements to AI in the likes of your mainstream strategy games. In the miniscule niche wargames market it is financial stupidity to invest that sort of effort into AI. The fact that it happens at all is due to the love of the hobby by a few designers. Whether that will continue is problematic. For it to have any chance, wargamers must become more willing to pay significantly more for such product.

[ravinhood]

ORIGINAL: NotaGrog

ORIGINAL: Veldor
It's not Advanced Tactics that has all the elements of success, nor any wargame, its the developer himself that has to (and in this case does) have the necessary components. Though the game is not without its flaws, its the primary reason I believe why the game is good in the first place and will only get even better in time. This unfortuneately to me, is the exception, not the rule. That AT can be looked at as such a great game in the first place is proof of where standards are at in computer wargaming. Though I can't really disagree with its stance as an exceptional product, I do on the otherhand kinda wish standards were higher (for everything from its graphics to its hard to manuever editor to its somewhat limited UI to the AI issues, lack of decent included tutorials, and so on). It takes more than one game to raise the bar though...

I agree in that I find AT a very playable game, despite it's many quirks, not because of them. The long AI turn is my only real complaint, but hardly anyone else seems to care so......

On AI in games I have a question regarding the Panther Games HTTR and COTA. I've never played them (no demos) but they seem to have a lot of real ground breaking ideas. For a start, 'no hexes' looks excellent - surely all computer strategy games should do this?

However it's the command structure that I find most interesting. My understanding is that you take control at a higher level of the OOB and sub-formations use their own initiative (the AI) to achieve your goals. What I wonder is surely that means the AI is basically playing itself as the AI opponent will be fighting back using exactly the same AI routines. Fair enough the high level aim has been set by the human, but I'm still too doubtful to make the purchase. From the forums, it seems that opinion is that the AI is not too bad but not too good either.

So I'm wavering because to me it's a good thing that you don't micromanage your whole force - also very realistic. But then the AI takes over and well, maybe this isn't quite so satisfying. Any comments appreciated!

If I comment on it there will be a flood of negative feedback upon my soul and body. Let's just say you are quite correct in your theories an assumptions about this series. Namely "the game plays itself".


I also think though a game can have a fair/average AI, but, if there are difficulty levels that improve the "Challenge" of the game then I'll grade that game highly as well. Like the Civilization series, Master of Magic, Master of Orion, HOMM II and several more. I don't care that an ai get's handicaps an advantages. Just like someone else said, that's exactly how we make sports like bowling and golf fun and competitive amongst one another as one is better than the other there is a handicap system in place to bring them closer to equal. Same thing with games that give AI's handicaps and advantages. I don't see that as "cheating" at all.

I know when I played face to face with friends in board wargames we'd give one another handicaps an advantages depending on the sides we played because historical games are always out of balance. That's one reason I don't like historical simulations. I'd rather developers make "whatif wargames" with historical scenarios for the history grogs to play and a huge selection of whatifs and random generated maps and units for us whatif types. I don't really care to play the Civil War like it was or WWII like it was or Battles of Napoleon like it was. All I want are the components of those wars/engagements and then I want to be the one to decide who to attack first or who to try to ally with. Much like the Total War games and Combat Mission and even Steel Panthers WAW and Making History games. These are more fun and enjoyable to me than playing Stalingrad for the umpteenth time or the battle of Gettysburg or Waterloo.

We need more MOM and SCI-FI games instead of yet another WWII D-Day or Bulge game really. Though I like the Bulge games, they just get old and samey too fast. We need an X-COM GALAXY game, where you're not just fighting on one planet, but, 100's all over the galaxy and having not only X-COM battles on land and underwater, but, in ice and snow, and off planet ship to ship battles where you board or they board your ship and you fight it out in space as well as planets. Yeah I'd like that. Most 4x space games are so boring because they don't have a land based tactical game with them or a decent AI. X-Com is the exception though. I don't think it really has that great of an AI as it has a great challenging aspect to it as you increase in power so does it without having to build buildings x, y and z to do so.

[Veldor]

ORIGINAL: marcusm
I am not a leading AI expert but it is the general opinion that his AI design
is way ahead of anyone elses.

My only point in this is thinking a game has a great AI design because it slaughters you when its getting massive production bonuses, bonuses to combat results, and so on is hardly valid. I can't say I've ever seen any group of people agree on what was or wasn't a good AI in the first place. Well on any neutral ground that is. Given most people don't even realize the handicaps in place no majority opinion would really be valid even if it did exist.

I've still won at Advanced Tactics at AI++. That by itself doesn't have to be proof of anything more than needing an AI+++. But where do you stop? AI++++++ with 1000 times AI production, 10x the AI combat effectiveness, and so on?

Giving AI + is not exactly cheating. It is a handicap. Handicaps are perfectly viable
in eg. Golf to even the field. Part of the game rules even. I think it's a great way to make the
match more interesting.

Wargaming isn't golf. And since you don't have handicapping in ANY OF THESE games for player vs player its not needed for play against the A.I. in any other fashion than covering up (usually dishonestly) how bad the actual AI is in the first place. For starters it should be onscreen what the handicaps and bonuses to the various AI levels actually are. At least then all players would be aware of whats actually going on. Then I've no problem with the extra A.I. levels being there. As that provides a fairer way to make sure everyone understands exactly how good the base AI is or isn't.

[ravinhood]

Actually Combat Mission does put it on screen what the advantages are +10% more troups to +200% more troups. How hard is that to understand? It also shows you can give them veteran units to your average or green units. Real easy to see the handicaps the AI is getting. Plus, I don't agree with you that wargaming doesn't need handicaps or advantages. Perhaps a "simulation" shouldn't have them, but, everything in wargaming is NOT a simulation. Many are abstracts of wars or new wars that the games make themselves as you play like the Civilization series or Master of Orion or Master of Magic. These are wargames as well as strategy games. There are wars within them so they qualify as a wargame. Plus we are talking about AI's in general not the systems in which they are presented. AI's are bad in most ALL type of games from sports, to The Game of Life to wargames. AI's encompass all games not just wargames or wargame simulations.

[leastonh1]

ORIGINAL: Veldor
Wargaming isn't golf. And since you don't have handicapping in ANY OF THESE games for player vs player its not needed for play against the A.I. in any other fashion than covering up (usually dishonestly) how bad the actual AI is in the first place. For starters it should be onscreen what the handicaps and bonuses to the various AI levels actually are. At least then all players would be aware of whats actually going on. Then I've no problem with the extra A.I. levels being there. As that provides a fairer way to make sure everyone understands exactly how good the base AI is or isn't.

The AI will always need to be given bonuses, simply because it's not human. It can't really be compared to a person because by its very name, the intelligence in any game has to be artificially programmed to mimic us. Isn't it all a con when you really get down to it? You just can't replace people. That's probably why so many people prefer playing pbem or ftf rather than the pooter, because the unpredictability, creativity, subterfuge and sneakiness is all there, which you'll never really get with a PC game without allowing it to cheat. Again, I'm guessing based on what I've experienced and know zilch about AI programming.

The question is, does this cheating or exploiting or whatever you call it need to be implicit or explicit? Should the player need to know more than that the AI seems to play a "good" game or "bad" game? To me, AT has a good AI because that's my perception of the things it does, the way it moves units and because I get beaten a lot by it. That's also partly because I'm not a grognard and because of my own lack of skill/knowledge/understanding of military doctrine etc. You may think the AI is ok, good or awful because you can beat it on the highest setting. Then again, you may be lousy at chess/tiddlywinks/whatever and I could beat you hands down because I'm just more experienced and better at it. There is no straight answer.

Isn't this pasttime about immersion and the ability to suspend disbelief? If you enjoy the game, is it important what's going on underneath? Isn't the whole point to play the game and enjoy the experience rather than spend time comparing the AI's bonuses to yours? I don't go to the cinema and pick apart all the cgi stuff. I'm probably peripherally aware of it and sometimes spot it easily, but if the film is good, I don't care that much because I'm enjoying myself. Which is why I still keep going to the cinema and playing games.

Secondly, the problem with presenting the info you mention to the player is that you come back to the age old argument about whether to hide the stuff we don't really need to see or show it and put some players off who really don't care? I don't really want to be presented with that info on screen as it would just make the whole thing very sterile and really would defeat the object of playing in the first place. That's my opinion though and you may disagree.

Regards,
Jim

[Veldor]

ORIGINAL: Arjuna

...Developing AI is not easy and its not cheap. We have invested around 40 years of effort into our engine, over half of that on AI. ....

You've been working on your game engine since 1967? [&:]

Nah I know what you meant...

I agree with everything you said, though some of it is a bit specific to the style of game one might be making. For instance, at this point, some of the possibly more advanced styles of A.I. I've experimented with would probably be pretty difficult to implement into a game engine like that whereas much more applicable to say a more fixed hex-oriented turn-based game.

That is to say where you mention not doing any specific scenario scripting but rather a more what I'll call 2nd generation A.I I'm refering to a more 3rd generation "learning style" AI. The advantage to the "1st generation" scripted style is ease of programming and implementation, with all the negative effects you mentioned. The problem I think with the 2nd gen style you've mentioned is having to do so much of that work to seed the system with values and information upon which to make such decisions (The 20/40yrs of effort you mentioned to say best case achieve a 12yr old mind). the more 3rd gen style just starts your AI dumb and self-seeds and grows (So from an infant to hopefully more like a 16yr old ... vs the 12yr old). In testing that though the main problem is in order to get a certain base level of ability (Say to take your AI to age 8 in comparison) you either need to run it forever (not practical without tons and tons of computing horsepower) or seed it against a more standard AI opponent (also not so practical as then you have to develop a full 2nd gen AI to help seed the 3rd gen AI). The second is a guess as I've not tried that one and am having difficulty finding any real applicable reference that applies enough to wargaming needs. Maybe a scripted opponent could be enough for it to play against?

Even if that was all done then to take it from an 8yr old to 12yr old (So just to 2nd Gen AI levels) you'd probably need a good base of actual live players. I dont think wargames sell in enough quantity for that to actually ever be enough data back to the central system (assuming users would even agree to the data transmissions). And of course to then take it past 12yrs old might take years to get enough results. Which isn't really gonna help sales day one other than being really really cool (at least to me).

Though I'll probably never get such a thing working in a game or have 40yrs alone to spend on it perhaps someone else will use something along those lines to further advance wargame A.I.

I wonder if you think I am crazy?

[JudgeDredd]

So I'm wavering because to me it's a good thing that you don't micromanage your whole force - also very realistic. But then the AI takes over and well, maybe this isn't quite so satisfying. Any comments appreciated!

Well what would you like? Not to micromanage, but not have the AI do it?

It's quite simple with the HTTR/COTA engine...you can take command at platoon level (for some units (AT?)), company level, battalion level and regiment level (and higher still) and give orders at ANY of those levels.

You can give orders at your highest command level and then, as the battle unfolds, you will undoubtedly have to go down the levels to adjust your battle plan.

You most certainly can select your highest HQ and give it an order to attack and go out for a two hour walk. But you are not guaranteed anything there except missing out on one of the best command level games out there, bar none.

If you do play that way, you're missing the point of the game (as ravinhood always does - mainly because he doesn't own it, so makes wild, characteristic, crazy assumptions).

The beauty of the game is the level of flexibility you have with your units. I NEVER give a command to the highest command. I generally give commands at regimental level, often give commands at battalion level and very often take command of companies (when I want to hold a road closed whilst the rest of my units retreat). But I like to walk my units through combat. And micromanaging in this is not the same as micromanagement in WitP for example.

Unlike RH, I will not force my opinion on you. Each to their own. Horses for courses. I love it, you may hate it. I can only tell you what I know (and unlike some people, this is not an uneducated guess, it's from someone who owns and plays the game) and that is that the game and it's engine are ground breaking.

I was not at all interested in the battles in Greece and that region...but the game has some fantastic battles and the maps are beautiful. What's more, if you have the patience, it has a great editor.

Sup to you. I've said my piece.