PACMAN Has Superior A.I.

[derF]

Yes thats right...Pacman.

I cry in hysteria when i try and remember what happened last night. It was about 0130 So technically my brain may have conjured the image in a fit of sheer exhaustion.

Okay, i was playing as the Roman Scipii. The turn before, i had eliminated the Greeks and there was only one legitemate Greek unit left, the (wouldve-been Greek heir) so in a gesture of recongnition of an honorouble foe, I attacked him and his only 19 bodyguard cavilry with my general and his only 20 cavilry.

So it was 20 Vs 20 in this battle of mighty honour. The problem was that i HAD to have reinforcements because although i had detached my general from army, the army was still nearby as "reinforcemnts".

I wanted to leave them behind, but strangly, i had no choice. No biggie i thought, id just have to make my general hurry up right? WRONG!

The enemy general rightfully started the battle in a nice uphill position. So i began marching my 20 horsemen beside him in a way that none of us would have advantage from elevation. Out of the corner of my left eye, what do i see? my entire 800 strong reinforcements SPRINTING up the hill in a stupid line formation against the general. I would not make it in time for the fight, so all i could do was watch the foolish general smite hisself on my elite army chargin after him - uphill!

The next bit, is the meat of the thread. The enemy general cav began charging downhill toward that unorganised line of mixed troops. Then i said to myself, "hold on, shouldnt they put some pikes in front?" too late. The general broke through the line, hitting and killing 30 Archer Auxilia in a frontal charge. Now the line was in chaos. Onagers were shooting wildly into my own horde, killing my own men, as the valiant enemy cavilry enveloped themselves in the swarm of troops incapable to killing them. Archers were nonetheless firing into the crowd, inflicting more friendly fire, and the onagers kept opening up.

I didnt see the final results, i had won, yes, but had hit ALT+F4 in utter disgust.

[Spino]

Wakka wakka wakka

http://www.atariguide.com/b1/144.jpg

http://www.1101.com/nintendo/pac_man_vs/images/PAC-MAN.jpg

[Gregoshi]

Pacman never had to deal with reinforcements and he was only allowed to move along a fixed path. Now the ghost AI on the other hand... ~:)

[Mikeus Caesar]

It's sad but true.....sad but true....i have more difficulty winning at pacman then RTW. I've spent the last 5 years playing pacman for half an hour every night and haven't managed to get that far into it. I beat RTW within a week of getting it. It also shows that pacman is more addictive than RTW with it's glossy graphics.

[Wrangel]

Havent we all been there.. at the very brink of suicide... when your superior reinforcement spread themselves and gets killed one by one after their general "honorably" rushed ahead of his army and got himself a ticket to the afterlife. Well this goes back to the problem of the AI controlling your reinforcements. If you removed that feature this would never happen. And you could actually go ahead and take that fight without having to move all your reinforcements AWAY from the battle, in fear of loosing all your generals and a lot of troops as well.

[drone]

After a couple of battles with the suicidal reinforcements, the lesson was burned into my brain, "Take generals out of all reinforcing stacks!" Not a problem now...

It always seemed to me that CA got this backwards. Reinforcing stacks with generals should always be controllable by the human player, stacks without should be AI controlled.

[Delwack]

Yes, CA should have done it like that, or allow you to choose which army you wish to control. Hopefully they will fix the suicidal general AI problems for both reinforcements and enemies, although I still wouldn't use AI-controlled reinforcements anyway.

what also really ticks me off is how the game decides timers. the game gave me 12 (yes that's right 12) minutes to fight a stone-wall siege with equal numbers (about half a stack on each side). I rushed in as fast as I possibly could, and got to the town square at 2 and a half minutes left, vs a general and hoplites and of course the great pathfinding at the boundaries of the town square got my general killed against hoplites, routing what was left of my army. If I didn't have to rush like a maniac in order to beat the timer...

[Ar7]

Come to think of it, you're right, I have never managed to beat PacMan, maybe because I was like 7 hmmm...*goes off to look for the game*

[Delwack]

the simplest ideas always produce the hardest games...

[EGO]

Maybe somebody should mod it so that all the AI units look like Pacman then maybe they'll fight better :charge:

[hoof]

It's very easy to make a hard game. It's much more difficult to make a fun game.

Harder does not equal more fun in my book. Sometimes making something harder can make it more fun, but in my experience, that's the exception, not the rule.

Sometimes when I play certain games, I get the sense that the designers have a deathly fear of their games being seen as "easy". It's sad, really, as I wish those that do would spend more time making a game "fun" than simply "harder".

[Mountaindew]

lets just hope that when CA releases its lastest RTW patches, that this stupid AI problem would be fixed...

come to think of it, if the sieges are so bad, then CA would probably counter with "this style of seige has never been done before, we are the pioneers in RTS realistic siege warfare...blah blah blah..."

eh. its bad. it needs to be fixed.

[KiOwA]

"Pacman has better AI than RTW!"

"Wait, Pacman doesn't HAVE any AI..."

"Exactly."

[Mikeus Caesar]

pioneers in RTS realistic siege warfare...blah blah blah..."

Real siege warfare didn't have a limited amount of time. Real siege warfare had days and days to fight battles.

[Blitz]

Well i just don't get it. Everytime when i am fighting small army maybe peasants or army that is smaler than me i usually get like 35 minutes and this is like 3/4 of the time. While sieging i never get higher than 21 minutes. I just don't understand.

Is it because if you fight small army you take more time then when you attack a bigger enemy?
Is it because sieging take less time then fighting on land without any walls stopping you from marching?
~:confused: :dizzy2:

:help: Please

:bow: :bow: :bow: :bow: :bow:

[SOSamurai]

I haven't seen AI in any game that could properly handle RTW. Even the RTW AI. :worried:
I think the game has tried to be too clever and evolved past the current level of gaming AI. They'd need many months of scripting time I think to make the AI as intelligent as it should be, which they aren't going to spend on a game that's already been released and has been critically acclaimed.

[Mikeus Caesar]

Critically acclaimed for having the intelligence of a two year-old.....

[Red Harvest]

I haven't seen AI in any game that could properly handle RTW. Even the RTW AI. :worried:
I think the game has tried to be too clever and evolved past the current level of gaming AI. They'd need many months of scripting time I think to make the AI as intelligent as it should be, which they aren't going to spend on a game that's already been released and has been critically acclaimed.

Sid Meier's Gettysburg/Antietam/South Mountain AI was pretty decent and that is OLD. It had to handle formation changes and charges. It had ranged attack, melee, hold commands, fallback, wheel, column, line, skirmish, dismount, etc. It had cavalry, artillery, and infantry. It had various terrain effects: slope, woods, water/marsh, bridges, towns. It had substantial morale effects with increased morale/effectiveness when supported on the flanks, in cover, backed by a friendly, etc. It had unit rally. It even had command structure that exceeds our RTW grouping by two orders of magnitude. Now that I think about it, its AI was quite a bit better. Adding to that, it also had a more useful control interface. However, it was simulating a much smaller battle scale and with much longer battles than RTW.

[Ldvs]

It always seemed to me that CA got this backwards. Reinforcing stacks with generals should always be controllable by the human player, stacks without should be AI controlled.

I'd rather they do not even give the stacks with no general in it to the AI. After all, this stupid feature wasn't in MTW, so why here?

[Spino]

Sid Meier's Gettysburg/Antietam/South Mountain AI was pretty decent and that is OLD. It had to handle formation changes and charges. It had ranged attack, melee, hold commands, fallback, wheel, column, line, skirmish, dismount, etc. It had cavalry, artillery, and infantry. It had various terrain effects: slope, woods, water/marsh, bridges, towns. It had substantial morale effects with increased morale/effectiveness when supported on the flanks, in cover, backed by a friendly, etc. It had unit rally. It even had command structure that exceeds our RTW grouping by two orders of magnitude. Now that I think about it, its AI was quite a bit better. Adding to that, it also had a more useful control interface. However, it was simulating a much smaller battle scale and with much longer battles than RTW.

Amen! I was going to mention those games! Sid Meier's Gettysburg was the first war/strategy game that made the hairs on the back of my neck stand up when facing the AI for the first time. It was the first time in my life that I felt an AI opponent was acting sensibly, let alone 'intelligently'. I can recall playing the demo and chose the scenario where you play the Union and have to defend a treeline north of the town. I set up my forces all along the treeline expecting the usual bullheadedness from the AI to smash and break itself against my fortified position but no! The AI marched onto the field of battle, deployed in line and then to my shock and horror, paused and then redeployed a brigade to its far left flank in order to overwhelm my far right flank that I foolishly thought would be ignored. I got my Yankee ass handed to me on a plate. I was shocked and thrilled at the same time.

Not that SMG was perfect, far from it. But to date only the fully patched version of MTW:VI comes close and/or surpasses the brilliant code written by Sid Meier for his Civil War games. However, I firmly believe think CA could learn a thing or two from Sid's AI design for those games.

[Mikeus Caesar]

Oh dear god, the ultimate humiliation for RTW.....i have just been playing Age of Empires, and Pacman, and then i played RTW. It wasn't too bad, until i ended up looking past the eye candy and seeing the AI's hideous tactics. So here, i quote something i was shouting last night:

'DON'T JUST STAND THERE YOU AI FOOLS!! Why are you standing there!! Can't you see that my archers are raining fiery death on you? Or are you retarded!?'

End quote. So i immidiately went back on Pacman and AoE, and had much fun, getting my ass wupped by superior AI.

[Arrowhead]

You got youre ass wupped on AoE? :laugh4:

Shame on you, man.

[Mikeus Caesar]

I put it on my personal super duper hard mod. It's hard. Very very hard.

[Ar7]

I can remember how the MTW AI manevoured and tried to hit my flanks or my rear. The RTW AI will just deploy and wait until it is hit or if it is attacking then it just takes a charge at the spear/legion wall.

Yesterday I had a great example. The Carthage army had the greatest defensive position one could imagine. The whole map was flat with a huge hill in the middle. Imagine the highest hill you have ever seen in RTW, well it was bigger. Anyway, they started on the top and I was below facing the steepest side of the hill. They deployed and stood on the top. Alright, I just walk my army to the other side of the hill and begin to approach them from the more "flat" side, so the height advantage wouldn't be so great. The AI just watched me, doing nothing in return, it doesn't try to bother me with it's skirmisher cavalry or the 3 chervon slingers, they just stand there and watch their godlike position go poof. But isntead, what the AI did was truly "brilliant"....it just walked off the field ~:confused: They lost half of their army due to my cavalry catching them and also because of the withdrawing.....why?!?! Basically this was their main force on Sicily and I won the island without a fight :dizzy2:

[Chelifer]

I reckon the thing is the real 3d battle was the one that made RTW a bestseller. Unfortunately ~:)
I would be happy if I am wrong.

[Crazed Rabbit]

Sid Meier's games all seem to have great AI (SMG for the tactics and Civ's for the strategy and diplomacy). And if he could do it and get a game released on time, why can't CA? Too busy making the head hurlers look right, perhaps?

I hope the patch fixes the:

Horrendous AI

Terrible Pathfinding in cities

Crappyness of Phalanx units (Ooo, the enemy general is a riding our way! Let us give him a cold reception by lifting our spears and turning our backs on him! We shall greatly hurt his feelings!)

Generals riding straight through hoplites and phalanxs

Super speed and super kill rate

Smallish maps

Catastrophically constircted modding abilities (make it so you don't need to steal or blow $3k to model units perhaps? And custom battle maps? OR BE ABLE TO ADD MORE FACTIONS?!?!?! HOW COULD THEY NOT SEE THAT PEOPLE WOULD WANT MORE FACTIONS?!?!?!?)

Many, many bugs

Harder to assualt cities (pushing the ladders off the wall, anyone?)

Broken Diplomacy (Why do they always act as if selling me their map info is like desecrating their parents graves?)

I'd also like it to make the barbarian nations unique, instead of having 4 factions like gual (gual, gual-with-chariots, gual-with-axemen, & gual-in-spain) and 2 factions with falxmen (Dacia, & Dacia-with-spearmen). I mean, they said they made egypt look stupid so there would be variety, then they make half the factions of the double click on your enemies and hope for the best variety! Not to mention that Rome gets the widest selection of cav in the game! And the best cav! And super infantry! ARGGHHH!

Ok, I'm done ranting. I'll be happy if they fix the AI. Eventually, I will have EB and BL and METW and CTW (Both chivalry and citadel TW) and Sengoko Judo (horribly misspelled, I know) and won't have to worry about the blandness of RTW.

Crazed Rabbit

[Mikeus Caesar]

Rome gets the widest selection of cav in the game!

In fact, when selecting a faction to play, if you hold the mouse over any of the roman factions, it even says 'limited cavalry', yet they have the biggest selection of cavalry than any other faction. Ridiculous. And as for your comments on the barbarian factions just being copies of gaul or dacia, too true!!! One thing you didn't mention was the f*cked up battle timer. 12 minutes too get in and capture a city guarded by a stone wall!? wtf is up with that?

[Old Celt]

Ok, I don't mean to be oversensitive, but I'm a programmer by trade, and think that rants against the AI are understandable, but I also believe, you give far too little credit for the level of complexity involved in designing brilliant AI. How many years did it take to build a chess program which only has to deal with max 32 pieces and 64 fixed positions which was capable of defeating a human chess master? You really are talking rocket science levels of programming expertise, and you aren't going to see that kind of talent wasted on games. That kind of talent writes AI code for the United States military simulations.

To design great AI, you would have to have a real live brilliant military tactician, schooled in the ways of ancient warfare to "brain suck". Then you would need thousands of hours of playtesting to see how the performance measures up to human competition. Oh, and along the way, it would be nice to make some money of the project to pay the bills. You cannot program common sense into all AI decision making: there are simply too many variables.

You can expect the patch to fix things like spears going through walls, and who gets to control reinforcements. You cannot expect a transference of Julius Caesar's military genius onto your CPU. Principles like sacrifice, interlocking fields of fire, and feints, are just too much for us to expect at this stage.

The game is an awesome achievement in its overall excellence. If you were willing to pay thousands of dollars for a copy of the game, instead of a lousy $50, you could demand cutting edge development. But let's be realistic. You get what you pay for in the end. Did you ever stop to think that what you pay for a game doesn't even pay for 20 minutes of an expert programmer's time? RTW is playable with its flaws, and quite entertaining for the majority of its target audience. I commend the programmers for what they have done on what I'm sure was an extremely ambitious project for the money that was available to execute it.

[Khorak]

I don't want a great AI, I want an AI that can do the absolute simple fundamentals right. And RTW can't. Tell a unit to form up after unit cohesion has been lost, and watch as the AI mashes all the men into a clump of idiocy, then sprays them out to where they're meant to be standing in the formation, then has them shuffling about, trying to push past each other and then, if it's in a real good mood, suddenly shattering unit cohesion once again because it's so damn thick its initial plan for putting all the guys in the right place FAILED. So now they're running about like headless chickens in their own formation trying to inexplicably try to place themselves in different part of the formation.
I want my phalanx to march in formation. I don't want them to lift up their spears and start wandering towards the target now with their spears bouncing up and down like dancing showgirls legs. I want my cavalry to charge every God damn time, as opposed to sometimes just not bothering in the slightest, instead running leisurely into the enemy unit and getting slaughtered wholesale. Attacking a plaza? Quick men, make for the invisible entrance to the invisible wall around the plaza! Are we in a phalanx formation? Then lift spears and completely change formation!

My objection to the RTW AI isn't how stupid it is tactically or strategically, it's the basic fundamentals of movement and control. I cannot properly enjoy a game in which I must combat shoddy work on simply commanding your units before I even have to worry about murdering the enemy.

[Alexander the Pretty Good]

Sid Meyer's Gettysburg! Woohoo!

At one time I could beat it on "medium."

I think RTW spoiled me.

RTW AI General:

"Seeing as we have a terrain advantage and the enemy will have to march long and hard to get to us in a timely fashion where we can slaughter them with a downhill charge...

"All right, men! Move out double time! We'll make sure that darn human can roll our flank up or my name isn't Generic General!"

*proceeds to charge his whole army into a literal wall of spears, confusing the would-be attackers to no end*

:charge:

[HarunTaiwan]

Okay, the one poster is talking about command and control of units, not really AI. That should be easy to do, right?

Now, for AI, I think we'd be happy if enemy Cavalry didn't charge phalanxes head on.

That's just silly.

[Fridge]

The thing is, just changing that one suicidal general bug could change the whole game for the better.

At the moment, my main problem is the battles are too easy. Why? Because my troops are always led by a general, giving combat and morale bonuses. The enemy's far larger force is led by a captain, and therefore barely survives the initial contact before running away en masse. Cue 2500 casualties to, what, 150? I was wondering where all the enemy generals are (I must have fought 30 or 40 battles against the Julii without once seeing a family member.), then I realised... they're all dead. Probably lost their life attacking some band of peasants or a single unit of German spearmen. If the AI can just keep their generals alive long enough to get some stars, the game would be much better. It's a small thing, but it could have sweeping effects... Bit late to be talking about it now, but if CA did pay attention - as they say they did - to the bugs reported on forums, that one has always been top (along with friendly fire!) so hopefully it will have been fixed...

[Old Celt]

I'd like to respond to some of the comments made since yesterday. The initial complaint was over how the AI performs. This is a very different issue from fixing known bugs such as odd phalanx behaviors and pathfinding in city battles. It does sometimes amaze me how people will continue to do something when they know it's bugged and somehow expect it to work! Case in point: grouping is bugged to a crippling degree. Workaround: DON'T USE IT! Is it really that hard to select units individually or drag a box around them? It's not a showstopper, it's a bug, and I'm sure it will be addressed.

Now the case of the AI choosing to charge its general at amazingly stupid times IS an AI issue. Although it seems simple to see you shouldn't charge into spears to human eyes, the AI has no concept of fear or sense of self preservation. Many variables in conditions would need to be evaluated to make a proper assessment of risk vs reward, and the actual combat tactics the particular civ typically employs. Were there generals and warlords killed by impaling themselves in a dumb cavalry charge? Yes, of course. Can it be fixed? Yes, but this depends upon budget constraints, sales of the product, and resource availability in development and testing.

I don't work for CA, but taking a guess, I would say that they probably have very competent programmers and QA, but we developers are really good at testing the LITERAL aspects of things, like: "Does the enemy general lead his troops? Yes. Does he charge his cavalry after infantry. Yes. Okay, it works." People in QA typically don't know much about advanced military tactics, so they are looking for bugs, not poor decisions on the AI's part. From a QA standpoint, the AI works, it just seems to be a brain dump from Moe, Larry, and Curly sometimes.

So, I guess I was peeved that people were trashing the whole of the project when really it is a relatively small number of AI weaknesses and outright bugs. Sure I think it's retarded when the AI stands in the plaza and gets pelted with missiles until dead.

Here are some ways I've been able to maximize my enjoyment of the game:
1. I know battles inside the city have plenty of bug/AI issues. I'm more or less of a "true to life" type player. So, I don't direct assault most cities because, just as in real life, a direct assault causes horrendous casualties. I lay siege and wait. Usually the enemy will sally, and a much more enjoyable field battle can be had. There are times when a direct assault must be made, and I try to bring overwhelming force on those occasions and absolutely devastate enemy defenses with artillery prior to going in.

2. I don't use phalanx warfare. Again, because I like to play truer to history than most, my view is that phalanxes were totally obsolete by the age in history this game encapsulates. Phalanxes are easily defeated, so why bother with them? If the enemy uses them, I let them deploy and get all lined up, but it's no contest when the combat begins. The phalanx is the equivalent of a crippled Tiger tank. It has alot of potential to do damage, but not enough manueverability to matter.

3. Since I dislike the bugs just as much as you do, when it comes to defense, I try hard to avoid being beseiged. If I see the enemy coming, I'll go out and meet them as far from my valuable real estate as possible.

I know lots of people won't agree with me, but I think the game is good enough to be worth working around most of its flaws. I truly do hope you get a patch that satisfys.

[Chelifer]

Case in point: grouping is bugged to a crippling degree. Workaround: DON'T USE IT! Is it really that hard to select units individually or drag a box around them? It's not a showstopper, it's a bug, and I'm sure it will be addressed.


What's wrong with grouping by the way? I did notice that I loose more often when using it. I just could not work out what exactly was wrong.

[fret]

People expect too much from video game AI. I agree the AI is pants, but what do people expect?

It took many of the worlds most brilliant minds over 20 years to write a chess AI capable of beating the best humans.

A chess board has...

-64 locations (a fixed number)
-64 locations... all the same (fixed attributes)
-32 Pieces in total (a fixed number)
-only 6 differant kinds of pieces (a fixed number)

...and very well documented stratagies and very specific and rigid rules of engagement. All of these things are always exactly the same for each player at the start. And finally there are only 2 players.



A typical PC stratagy game, in this case RTW, will have...

-10,000 squares (this number comes from CA claiming there are this number of unique battlefield locations)
-10,000 squares....... all differant attributes
-??? units in total. I dont know how many, buts is a lot more than 32.
-Probably an infinate number of possible combinations of units and number of units - on both sides.
- a variable number of players
- plus an unimaginable number of other variables, that all have a part to play in determining what the AI should do next.


I dont expect to see good AI in a complex stratagy game for many years to come.

[Duke John]

Pathfinding is probably the most difficult, and IMO the AI does that pretty well all things considered.

However it fails horribly at things that aren't that difficult to code;
- stick together
- add reserves, and keep them reserve until unit within A metres is wavering. At the moment AI units attack the same unit stacking upon each other.
- no crisscrossing of units. The AI now frequently has one unit going left and the other going right, thus intermingling and loosing cohesion.

Proper coding for assigning army formations. The current code only has a list of army formations and no tactical code behind it. The result is the oh so common single line of units regardless of being outnumbered or not.
There is code for traits where the size and composition of opposing armies is compared. Why didn't CA use that code to assign army formations? If outnumbered, make a crescent shaped formation. If outnumbering make an hollow crescent.

The quickest way to defeat an army is to cut it up and attack flanks. The current AI should not be written to do that, but to prevent it. I think players will have a much harder time when the enemy army sticks together regardless of decoys, defends its flanks and keeps some reserves. Fairly easy to code since it doesn't need much complicated movements, just keeping cohesion and a closed frontage (just like in history).

Couple that with a smarter AI that outbuilds the player (why not, the strategic part is mostly a matter of economic calcuations). If the AI also sticks together on the campmap with 3 stacks close together then the player will have it very hard.

[fret]

Yes, pathfinding is bad as well, however, I dont think its applicable to lump the pathfinding algorithms into the AI debate, they should be (probably are) mostly seperate to the AI logic.

The big issue with AI, and the evidence that it is still in the dark ages of development (accross all games, not just RTW) is that when the user selects a more difficult AI setting, the AI logic probably wont change at all, it will instead weight the level of oppisition by means of production bonus's , the strategic element will not change one bit.

When we have an AI that uses increasingly complex logic to increase in difficulty, rather than production or performance bonus' to increase the difficulty, we will have good AI. Until that time, we just have to live with what weve got - one AI logic and 'wieghting' to make it less or more difficult.

[Duke John]

Not entirely true I believe that a developer once said that at easy difficulty the M:TW AI has lesser algorithms to calculate its strategy.

[fret]

Im an guessing now but what they probably meant by that was the 'depth' the algorithm went to - the distance into the future it looks at the possible outcomes of its actions, and then bases its decision on this information. That isnt a logically greater or lesser AI. the AI that only looks 1000 loops of the program ahead will be worse than the AI that looks 10,000 loops of the program ahead. Its jut a trade-off of computational power against speed that makes the method either effective or not. A strong AI will assess the possiblities further into the future, the lesser AI will just look at the absolute-immediate situation.

To highlight this method apply it to a strictly turn based setup: this is usually the sole way in which a chess AI will increase its difficulty, the number of turns ahead it assess's.

The reason I think thats what they meant is because they used the example of the lower AI using lesser algorothm's, if they had said the stronger AI has stronger algorithms, then it would be impressive, unless it was just an increased depth of thought and it was run on a supercomputer.

[fret]

EDIT - the second possilbity, again because of their use of the lesser AI ather than stronger AI, is that they just remove chunks of the code for certain aspects of the decision making process.

{very simple} example:

Lesser AI will assess the location of opponent units

Stong AI will assess the location of opponant units AND the terrain



For a truly stronger AI, it would look like this...

Lesser AI will assess the location of opponent units AND the terrain

Stong AI will assess the location of opponant units AND the terrain but make a better judgement and decision

[Old Celt]

I read somewhere in all my forum hopping on RTW from a credible source that the only differences you get when you change difficulties in battle mode are the morale settings of your troops vs the enemy. On Easy, your troops have uber morale, while the enemy have poor morale. On Very Hard, the enemy has good morale, and you have lower morale. That's it. no difference in tactics.

The campaign mode is similar. Changing difficulty just gives the other side more resources, but it doesn't change the logic of how those resources are used.

[Red Harvest]

It's a lot more than morale. On the battle map the MTW/VI received morale increases plus +4 attack and +4 defense on expert. Hard was +2/+2 IIRC.

In RTW the AI gets some morale bonuses and +7 attack on VH. Apparently there is no defense stat change (although morale will improve defense.)

[RJV]

Not entirely true I believe that a developer once said that at easy difficulty the M:TW AI has lesser algorithms to calculate its strategy.

Yes there was a fairly detailed post that described how the different levels of AI compared. At Easy it just massed it's troops and that was that. Move up and it would check to see whether it was caught in enfilade fire. That sort of thing, only a lot more descriptive. Sure there were combat modifiers in there as well, but it would at least try a number of different things.

The difficulty is of course, the problems it has (unit cohesion et al) exist at all difficulty levels and once you take into account the excessive combat modifiers at VH, you can easily end up with situations where you carve the AI to pieces, only to have your elite generals carved up by a lone guy wearing a pair of ragged of trousers and carrying a particularly vicious comb.


To the programmer earlier who was talking about the CA devs being competent programmers rather than military enthusiasts - agreed, and this is a common failing of software development in all areas, not just games. The company I work for has bought a major financial package from a 3rd party, which as a piece of software is OK, but doesn't handle the business intricacies particularly well. When discussing change requests and upgrades they focus on the programming side but have a relatively poor comprehension of the environment those changes will be used in. To get a serious improvement in the future games companies need to have people on their staff who not just 'dabble' in the relevant game area, but are enthusiasts, even experts in it.

Cheers,

Rob.

[fret]

At Easy it just massed it's troops and that was that. Move up and it would check to see whether it was caught in enfilade fire. That sort of thing, only a lot more descriptive. Sure there were combat modifiers in there as well, but it would at least try a number of different things.

Still, that isnt a logically stronger or weaker AI, the strong AI will still come to the same conlusion as the weak AI given that they both have the same information available to them in a particular assesent/calculation.

All they are doing is restricting the data available to the lesser strength AI, therefore giving the impression that the stronger AI is stronger.

For a stronger AI to be stronger in the case of decisive inteligence, rather than net outcome of a situation, it has to make a better decision than a weaker AI based on both AI's having exactly the same starting data.

[Fridge]

Just a layman's question, but how much of the AI is actually AI, and not scripted If... then statements? Is there in fact a qualitiative difference, or is it merely quantitive - ie, is what people refer to as 'AI' just a sufficiently large and complex list of scripts, or is AI actually something else?

On another point, RJV: nice sig ~D

[fret]

The AI that is used on a PC game will be limited by the logic of the 80x86 platform. This hasnt changed in over 20 years.

Its nothing special, just a whole bunch if's and boolean logic evaluating a stack of variables and making decisions based on their values.

For it to be any differant, you would need a computer with a processor that operated differantly, like the much talked about quantum processors that are curretly being theorized about a lot at the moment.