'Neurotic' is Best RTS strategy

An anonymous reader writes "Austrian researchers experimenting with adding emotion to game AI say that 'neurotic' software is best at RTS. They developed aggressive, defensive, neutral and neurotic bots to play Age of Mythology, based on psychological models of emotion. Neurotic bots beat the standard game AI every time and faster than the other personalities."

Comments

Result is specific to AoM?

[Chief Camel Breeder]

This tells us more about the game-play balance in AoM than how to approach games in general. I'd be more interested in seeing these bots play CiV 4 where I doubt that neurotic behaviour would triumph.

Re:Result is specific to AoM?

[arivanov]

Dunno about Civ4.

My guess is that the advantage is limited to games with "learning" AIs, where the AI attempts to extrapolate your behaviour based on your past events. The neuroticity adds an element of unpredictability which will confuse the hell out of an AI that works using extrapolation or neural net training. If the game has a rigid rule based AI there should be no advantage, just the opposite.

Re:Result is specific to AoM?

[strattheman]

The article is talking about RTS strategy. I've never heard Civ 4 described as an real time strategy game (it's turn-based, not real-time, no?), and if it is, it's far less typical of the genre than AoM.

Having watched some amazing starcraft players, neurotic sounds about right.

Re:Result is specific to AoM?

[SatanicPuppy]

It's a bit interesting. AoM is a game with great scope, allowing for unusually "large" game boards by the standards of other RTS. Consequentially, the AI has had to be somewhat "toned down" from the kind of cutthroat AI you got in WarIII or even startcraft. I'm a big RTS buff, and while I _liked_ AoM I never found it all that difficult. Some games are a lot more forgiving of a failed attack, and that's one of them. You have enough resources and fast enough build times that even if your grand fleet gets crushed, you're probably okay.

Reading the article, (which is freaky low on detail) it seems more like "Neurotic" in this case is meant to signify a lack of caution. Aggressive won every match, and neurotic won every match, but neurotic did it faster. This suggests that irrational risk taking (the article mentions that the AI skews its internal numbers about how many resources it thinks it has) can beat a more cautious opponent.

In both cases it seems clear that aggression carried the day, and that the only real difference was that the AI that lacked caution won faster. To me, that suggests a big problem with the regular AI, because that lack of caution is usually pretty easy to exploit...A counterattack on a resource gathering operation would leave the crazy AI crippled, due to low reserves. The regular AI's counterattack algorithms must be pretty weak, or it's build order is too cautious or something.

I'd love to see a better description of the AI programming.

Re:Result is specific to AoM?

[hey!]

This is a good point.

Years ago I remember reading about a study which "proved" depressives were more cognitively accurate, by setting up a task which was fixed so that the subjects always failed. The depressives of course recognized this much more quickly than the normals.

However, that said I think there is still an interesting point here. The neurotic profile may exaggerate the situation, but at least it reacts to it, as opposed to inbuilt tendencies toward being aggressive or defensive all the time. One possible benefit to emotions is to engage survival behavior early, before perfect information on a situation is available.

On the other hand, the way to defeat a neurotic is to deceive him into misreading a situation. Once that happens, he will not adjust to contradictory information.

Re:Result is specific to AoM?

[Chief Camel Breeder]

You're right: Civ is turn based. However, I'd guess that RT vs. turn-based matters little to the bots; I'm guessing that they are fast enough for the RT aspect not to hinder them.

Re:Result is specific to AoM?

[Hoi Polloi]

Did they pit neurotic vs neurotic? I wonder if then it would turn into a coin flipping contest with their chaotic behavior just resulting in random outcomes.

I'd think that in a 3 person battle a neurotic AI would be at a great disadvantage because the style of "to hell with the consequenses, charge!" game play might win against 1 but not with a 3rd party. They'd jump in take advantage of the neurotic side when they had no reserves left and had spent themselves fighting the 2nd opponent.

Re:Result is specific to AoM?

[peragrin]

In most RTS games I have played the AI rarely attacks resources instead it goes against enemy soldiers first, and only takes targets of opportunity against civilians. My general strategy build up a strong defense, while building up resource gathering. Then use long range weapons to attack their resources. Sending my army up againist their ends in Mutually assured destruction. but I have a solid backing for rebuilding while they struggle.

Every so called "learning" AI I use that againist struggles with it. when the day come it fails i don't know what I will have to try.

Re:Result is specific to AoM?

[grimJester]

My guess is that the advantage is limited to games with "learning" AIs, where the AI attempts to extrapolate your behaviour based on your past events. The neuroticity adds an element of unpredictability which will confuse the hell out of an AI that works using extrapolation or neural net training.

"Learning" like a human? My first thought on reading the headline was how / if this could be used to make a decent poker AI. Unpredictable behavior, within certain parameters, is an advantage in many games where you have information hidden from the other player, and he is reduced to guessing based on your actions what that hidden information is.

Re:Result is specific to AoM?

[ockegheim]

...it seems more like "Neurotic" in this case is meant to signify a lack of caution.

Maybe it is a mistranslation. Though in that case, as an Austrian invented neurosis, it pretty much messes up psychiatry in the English speaking world.

Re:Result is specific to AoM?

[Jaeph]

If I understood correctly, a nuerotic AI is one that exxagerates negative feelings. So it panics earlier at losing resources, or someone's scout, or whatever.

To me, that sounds about right for a game-winning AI. Most AIs seem nice and placid and just wait around for the players to attack, and then under-react to the attack.

-Jeff

Re:Result is specific to AoM?

[skeeto]

I'd be more interested in seeing these bots play CiV 4

Yes, because you would have to be neurotic to play a game where 16th century archers can destroy modern tanks.

Re:Result is specific to AoM?

[mikael]

I remember playing one of these Command and Conquer games that came free with a Dell computer. One one map, I just built some construction bots, built a wall to keep the enemies bots from coming from the North and leisurely bombed the hell out of their manufacturing plants and airforce which was stationed on some cliffs at the very top of the map. Meanwhile their troops had corralled themselves along the wall.

If the AI had the intelligence to create one or two construction bots to "disassemble" the wall, the game could have very well played out differently.

Re:Result is specific to AoM?

[Richard Steiner]

It's a bit interesting. AoM is a game with great scope, allowing for unusually "large" game boards by the standards of other RTS.

How do AoM board compare to TA? Or Supreme Commander? Even TA allowed for largish (63x63 screens) maps like Real Earth and up to 5,000 concurrently active units per player, and that sort of scale plus the nature of TA resources can make finding and killing an opponent rather interesting. :-)

Re:Result is specific to AoM?

[Anonymous Coward]

WARNING: If you are depressed, you probably should skip this message.

Actually, depressive people are really accurate at least on one metacognitive task: estimating accuracy of their knowledge. People are prone to overconfidence, even more so if they are estimating their reliability in their area of expertise. This one of the best established cognitive biases because people do not just overestimate, they do so by huge margins. Depressed people are not slightly more accurate, they actually can estimate reliability of their knowledge pretty well.

Re:Result is specific to AoM?

[hey!]

It's actually well established that people overestimate the future hedonic value of actions: if I buy this thing I'll be happy; if my wife leaves me I'll never be happy again.

To the degree that the depressive tends to deny the potential for happiness, it is quite possible that his estimation of positive results is more accurate. On the other hand, I expect that his estimation of negative results is even more exaggerated than normal.

Re:Result is specific to AoM?

[AvitarX]

WarcraftI (II?) was similar.

The walls were neutral so AI would not touch them. This lead to a bunch of units lining up and doing nothing.

Re:Result is specific to AoM?

[mikael]

Now, I can see the logic in that - in the worst case scenario, you end up with one construction bot on your side trying to build a wall as a defensive measure, while another bot is dismantling the wall as part of an offensive strategy.

Re:Result is specific to AoM?

[andreyw]

imo AoK:TC always had larger scope...

I've played 4 hour AoK games...not so with AoM. Also AoM introduced rigidity into game play with TC positions. AoE III makes it even worse, being able to only build one castle... while the game play has de-evolved greately from an economic management stand point (which is was AoE/AoK always has been about). You could Ensemble has tried to make the game more approachable for people who can't devote time to master a more complex system, or to play 4 hour games... but it's just made it boring for me. Sure, the graphics are nice. If they just released AoC with better graphics... taht would be awesome.

Re:Result is specific to AoM?

[SatanicPuppy]

Starcraft had a similar "bug"...The wall 'o supply depots. If you were playing the Terrans against a computer opponent, you put a line of supply depots in front of your bunkers. When the vicious melee ground units that normally tore your bunkers a new one came running up, their AI would tell them that they HAD to attack the bunkers, but they had no ranged attack, and no path to the bunkers, so they'd end up running back and forth in front of the wall of depots, unable to kill the wall, and unable to attack their target.

A couple of bunkers could wipe out the sort of zealot rush that is almost impossible to stop without bunkers chock full of firebats. At it's heart, its an AI problem. The AI chooses a target it's unable to attack, and then is unable to drop it. IMHO it's one of the oldest, and worst, bugs in strategy game AI. It's easy to exploit.

Re:Result is specific to AoM?

[zippthorne]

I believe that was fixed in either a patch or the Broodwar expansion. At any rate, the "supply depot wall" doesn't seem to be effective anymore.

Re:Result is specific to AoM?

[BigDogCH]

Agreed. AOK was damn near a perfect RTS. Some improved diplomacy, smarter AI, and a graphics upgrade, would have made me happy. Hell, after trying several RTS games, everyone I know still goes back to Aoe-AOK.

I would like to see someone write a better AI for AOE. It doesn't have as many artificial restrictions as AOM etc, so the AI needs to be a bit more flexible.

Re:Result is specific to AoM?

[CodeBuster]

This brings up what has, IMHO, been a major sore point in an otherwise brilliant series (Civilization) and that is the generally poor level of play on the part of the AI (i.e. resort to particularly blatant cheating on noble and higher difficulty levels with generally moronic play on lower levels). If Firaxis, or anyone else, is going to do another iteration on the Civ series then please, for the love of all that is holy, do not make the AI suck. If the Strategy First people can do it with Galactic Civilizations then whoever does Civ 5 should be able to get the AI right.

Re:Result is specific to AoM?

[Touvan]

It could also mean that this particular game (or RTS games in general) has rules that favor the more aggressive player. I myself fell into the cautious or defensive category when I played games like WarCraft III, and I always lost to my more aggressive friends, until I got more aggressive (attack earlier, build less defensive units, more offensive units, take the initiative).

It doesn't necessarily mean there's anything wrong with defensive or regular AIs in general, as much as it shows that in these particular games (or just AoM) aggressive AIs (and possibly players) do better.

The human trials should be interesting.

Re:Result is specific to AoM?

[Anonymous Coward]

(Same AC here) I wasn't actually referring to estimations about emotional states. I had no idea about that. The cognitive psychology usually establish things like overestimation bias by utilizing objective measures. A simple test for overconfidence goes like this: Subject are given some estimation tasks, such as guessing number of people in Australia, whether more people die from cancer or traffic accidents, whether mars is bigger than venus etc. Of course some people already know the answers so do not guess, but that is beside the point. The next task for subjects is estimating the probability of correctness of their answers (this task may be parallel to estimation task.)Then, the correctness of the answers are compared to subject's stated confidence. Without fail, people way overestimate their knowledge. That is, unless they are suffering from depression. The interesting bit is not that depressed people are not as overly confident as normal people, that is to be expected, but that their estimations are actually accurate. They do not systematically underestimate their knowledge or ability, they simply perform better than normals.

Re:Result is specific to AoM?

[kungfujesus]

Dunno about civ 4, i'd bet that the best RTS to test this out would be Supreme Commander.

Re:Result is specific to AoM?

[SatanicPuppy]

Nope. I used it just the other day. The computer is better at breaking it down, given time, but if you put enough crap behind it it'll still override their desire to attack anything but your tanks. It's still a hardcore game winner in the early game, because nothing will convince a zergling or a zealot to attack those depots if something is shooting at them. (imho, it's all part of the pathing problems that are the biggest flaw in starcraft...Block a unit, and it will circle the board trying to find a way around.)

I think what they added was a tendency for units like Ghosts and Defilers to target those buildings with their nukes and plague. The protoss use corsairs with the disruption web...once nothing is attacking the the ground units, their AI focuses on buildings.

But, in the absence of all those things, you still get the joy of watching a pair of ultralisks get killed by two marines in a bunker protected by 3 supply depots. It's pathetic.

Re:Result is specific to AoM?

[StikyPad]

The problem with Civ4 is that neurotic behavior is a losing proposition for the player almost regardless of the AI (unless the AI displays similar behavior) because the game seems designed to reward planning and execution of a particular strategy rather than constantly shifting tactics. If you've been aggressive toward your neighbors, for example, you almost definitely can't suddenly establish amicable relations. If you've invested in a particular branch of technology, you almost definitely can't switch to another. And if you've established a society with an emphasis on culture, you'll have a hard time switching to military might (which is difficult enough to achieve even when it's the goal all along).

Of course, the emphasis of meticulous planning and execution over spontaneity and flexibility is one of the reasons I suck at, and consequently don't enjoy, Civ4.

Re:Result is specific to AoM?

[Architect_sasyr]

Bah to hell with your consequences! CHARGE!

Reminds me of the 40k Orc Codex: "We never lose... If we wins we win, if we dies we're dead so it don't count as beat and if we runz away we can come back to fight anuvva day" (Paraphrased of course - this is /. so no doubt someone will correct me)

Re:Result is specific to AoM?

[windex82]

(which is difficult enough to achieve even when it's the goal all along).

gdi, thanks for ruining the ending...

Re:Result is specific to AoM?

[StikyPad]

No problem. Also, Dil is really a man, Jacob never left Vietnam, Dr. Crowe is dead, Jack and Tyler are the same person, Jesus comes back to life, and rumor has it that an MLB team will win the World Series this year.

Re:Result is specific to AoM?

[TrueKonrads]

I think what it means is that the one who takes initiative in equal opportunity setups wins. This falls in line with common wisdom and in business world, being very persistent often is virtue.
Also, if we look at military history, usually more aggressive civilisations win over more pacifistic and defense minded ones.
Besides, based on game rules, the one who is the sole survivor wins, so no point in building pretty castles and nice cities - this consumes resources that might have been otherwise converted to direct damage to opponent.
I think that Civilisation series had the nice alternative of launching a space ship as a winning move.

Re:Result is specific to AoM?

[TheRaven64]

TA was seriously rigged in favour of an AI on resource-rich maps. Even without much intelligence, the ability to concurrently control large numbers of construction bots gave it a huge advantage. The game was seriously in need of better support for delegation (i.e. tell a construction bot to establish a forward base somewhere, and build other construction vehicles to help it).

Interestingly, in Red Alert, this was used intentionally to make up for the poor AI. While human players could only build one unit of each type or structure at a time, the AI could build a different tank in each war factory and so on.

Re:Result is specific to AoM?

[TheRaven64]

Most RTS games I've played are basically economic battles; the player in control of the most resources wins unless they do something stupid. If you take control of enough resource-containing areas of the map early on then you win. In Red Alert 2, for example, there were oil refineries on some levels which gave you a constant but slow injection of money. If you captured and defended these early on then you were likely to win.

If you had control over similar amounts of resources, then attacking was usually the wrong strategy, because the side that attacks tends to lose more units than the one that defends, and the defender can typically repair their damaged units, while the attacker loses theirs from defending fire as they try to escape. If you survive the first attack, then you typically have more resources left than your opponent (including ones tied up in vehicles, etc).

Pretty much every RTS game I've seen favours AI players, in that they tend to have very little scope for delegation, and so a player capable of focussing their attention on several areas at once does better. Interestingly, many AIs fail to exploit this inherent advantage particularly well, and are rarely written to evolve a coherent strategy with multiple concurrent elements. If I were writing AI for this kind of game, I would look at swarm behaviours, with particular reference to BT's research on emergent properties of systems.

Re:Result is specific to AoM?

[binkzz]

Surely pitting any AI against the exact same AI will yield random results?

I don't see your point.

Re:Result is specific to AoM?

[Richard Steiner]

TA was seriously rigged in favour of an AI on resource-rich maps. Even without much intelligence, the ability to concurrently control large numbers of construction bots gave it a huge advantage. The game was seriously in need of better support for delegation (i.e. tell a construction bot to establish a forward base somewhere, and build other construction vehicles to help it).

Effective player use of command queueing can help with that to some extent -- I regularly have between a half-dozen and several dozen tasks in the order queue for each for my construction units, mainly sending them out to construct turret/structure patterns and dragons teeth barriers, and I'll often set one or more other construction units (usually level 1 units helping a level 2) to speed up construction for the directing unit. You can keep several dozen construction units constantly busy for a half-hour that way ... no problem. :-)

You can queue up darn near ANY type and/or combination of orders with the SHIFT key in TA. Other game makers need to learn how to emulate that type of interface, IMO.

Tautologous

[Anonymous Coward]

What, did they expect the lazy or apathetic bots to excel?

Being skilled at any endeavour requires "neurosis".

Re:Tautologous

[Anonymous Coward]

So true. I have to piss off my toaster before it will toast my bagel.

Re:Tautologous

[Wyrd01]

So true. I have to piss off my toaster before it will toast my bagel Just be thankful you didn't get the model of toaster I bought.

Re:Tautologous

[SoulRider]

You must have bought one of those toasters with GPP.

Re:Tautologous

[bWareiWare.co.uk]

Wouldn't that be:
http://en.wikipedia.org/wiki/Red_Dwarf_characters#Talkie_Toaster

Re:Tautologous

[SoulRider]

ahhh, never saw Red Dwarf. That toaster sounds better than one with GPP though.

Welcome to Lords of the Realm 2!

[corvair2k1]

I discovered that a hardcore neurotic kind of strategy worked well in Lords of the Realm 2 when playing with my brother. He didn't care, and would rather have the game over quicker than not, so when we started the game he immediately spent all his resources on getting weapons and a huge army, and within four turns or so had come over and whooped my ass. Every single other aspect of his kingdom was in shambles, but he had the element of surprise, and that's all that ended up mattering.

I'm thinking the AI would think something similar to me... "Surely he won't try that. If he fails in his attack, he'll just fall over on his own accord in a few turns." Unless he doesn't.

Re:Welcome to Lords of the Realm 2!

[VShael]

Also known (though it takes more than 4 turns to do it) as "U.S. Foreign Policy"

Re:Welcome to Lords of the Realm 2!

[haeger]

Hmm, now where's that "+1 Tragic" option?

.haeger

Re:Welcome to Lords of the Realm 2!

[imgod2u]

Zerg-rushing has been in use long before you picked up the mouse and keyboards son.

Re:Welcome to Lords of the Realm 2!

[Hoi Polloi]

So does this mean that all of the really good online players I run into are actually neurotic messes? I'm not disputing that, I just want to make sure.

BTW, I found that one of the best strategies in the old Civilization games was to build horsemen/charioteers as fast as possible and raid other civs before they had a chance to start even at the expense of your own city growth. Only oceans got in the way of a quick win this way.

Re:Welcome to Lords of the Realm 2!

[GMFTatsujin]

Sounds like it's time for an inventory of games in which war has consequences before the "you beat up everybody!" resolution comes about. Just about every game I can think of is positively resolved when you've wiped everyone else off the map, no matter the state of your own backyard.

If you're willing to expend your entire kingdom (or empire, or corporation, or whatever) so that you can crush everyone else, have you really served as a good leader? What games force you to justify that expense?

Re:Welcome to Lords of the Realm 2!

[Impy the Impiuos Imp]

The first time in PvP is a cruel lesson for the PvE player. I played Warcraft II on T.E.N. (Total Entertainment Network) way back before it went out of business, in the Quake I/Duke Nukem days. I built up my stuff, generated a big honkin' demon or two, and sent them...into quick destruction at the hands of a zergling rush of archers or spear throwers, I forget.

That was pretty much the way to beat other RTS, including the awesome Sacrifice. Huge numbers of popcorn ranged units (which attack both ground and air forces.) Don't waste time with anything else.

And as a squad-based RTS, it was vastly superior to that hobbled terd Warcraft III.

Re:Welcome to Lords of the Realm 2!

[dwye]

If you're willing to expend your entire kingdom (or empire, or corporation, or whatever) so that you can crush everyone else, have you really served as a good leader?

Life. If you eliminate all of your other competitors, does it matter in a couple hundred years how your enemies might have been? Likewise, how do you know how good or bad the alternative would be? A short game with terrible results for the victor may seem much better when you include the centuries of peace and prosperity (and boring play) after.

What games force you to justify that expense?

Success is its own justification, when failure is extermination. A better question is what games prevent such a strategy, and how do they do it? And, is that mechanism just a Deus Ex Machina?

Re:Welcome to Lords of the Realm 2!

[h4rm0ny]

The difference is that games have an end, whilst short of global catastrophe, real life does not. If a group is successful to the point of eliminating all other groups, it will soon turn on itself because the values that brought it to that stage demand it. If, for example any particular human group you choose such as Christians, blacks, whites, whatever, successfully elimated all other groups on the planet, you'd soon see it split into new, struggling factions. Because objectively, there aren't any differences between these groups significant to prevent working together. The differences are superficial or self-imposed. So there's no reason to suppose that new differences wont be found to justify another round of global supremacy. Co-operation and peace breaks the cycle. Winning never does.

Re:Welcome to Lords of the Realm 2!

[Daetrin]

Zerg-rushing has been in use long before you picked up the mouse and keyboards son.

I hate to tell you, but "Lords of the Realm 2" came out two years before "Starcraft," so i doubt there were very many Zerg rushing around at that point :)

Re:Welcome to Lords of the Realm 2!

[SlowMovingTarget]

When I was your age, we only had one RTS called Dune 2. It didn't even have a frickin' colon in it's name! We had to lay the frickin' concrete before we could build the goram tank plant and even THINK about havin' a rush of anythin'! And then, when you built yer unit cap of 10 whole tanks, a sandworm'd come along and eat 'em all before you got within shooting range. You youngsters've got it too easy, what with yer fancy-shmancy zerglings and no sandworms. Bah!

Re:Welcome to Lords of the Realm 2!

[Molochi]

So is the "Surge" a zerg rush?

Re:Welcome to Lords of the Realm 2!

[Daetrin]

When I was your age, we only had one RTS called Dune 2. It didn't even have a frickin' colon in it's name! We had to lay the frickin' concrete before we could build the goram tank plant and even THINK about havin' a rush of anythin'!

Perhaps you're so old that your memory is starting to go (47 by your math i guess) but you couldn't actually rush _anything_ in Dune 2. There was no way to select multiple units at one time, neither drag-select nor ctrl-select had been implemented yet. You had to click on a unit, tell it where to go, click on another unit, tell it where to go, etc, so instead of a whatever-rush you'd end up with a whatever-trickle.

And back in my day we didn't _need_ no frickin concrete in Dune 2, we'd just build on the bare rock without it and then repair the damaged structure! With our bare hands! In a sandstorm!

Re:Welcome to Lords of the Realm 2!

[SlowMovingTarget]

In my day, single clicking every unit and inching them forward one at a time WAS a tank rush. And another thing... super bats were the first "mobs," why I can remember one Wumpus hunt...

It was Dune 2, and not Wing Commander, that convinced me to by a Sound Blaster. Granted, Wing Commander made more use of it.

You're right of course, there was no rushing in Dune 2. But remember setting up a semi-circle of vehicles and baiting the bad guys in? Or laying out concrete all the way to the doorstep of your opponent, building a couple of turrets, then stationing a handful of rocket launchers there? Or running down those three pixels worth of infantry with the harvesters to make little splats in the sand... Good times. Then this WarCraft thing came along and everybody was talking about this new genre...

...and my "whatever-trickle" is none of your business.

Re:Welcome to Lords of the Realm 2!

[RMingin]

Actually, Medieval: Total War models that very nicely. If you get up a good solid steamroller and manage to accrue something like ~75% of the game's landmass, you're sitting on a lit powderkeg. When your current regent dies, ALL of the heirs typically make a grab for a chunk of the empire. Your unstoppable juggernaut becomes a hydra, and your advance is horribly stalled. The new AI players break down one by one, with your victims often resurging, and you have to do the whole thing over again.


It cheats, too. If you time things so that you begin the great war of assimilation while your regent is young, the generals in your armies get all nutty and go civil war on you once you get too close to winning.

I love M:TW. It's freaking impossible.

Re:Welcome to Lords of the Realm 2!

[biovoid]

Actually, if you placed your units in a rough line pointing towards your target, then start moving the back units first, you could achieve a rush. Who needs multiple select?

Great game.

Re:Welcome to Lords of the Realm 2!

[h4rm0ny]

Fascinating. I've not come across that level of (sociological) sophistication in a game before, though the CIV and SIM games do in a way, I suppose. Sounds worth looking out for. Cheers.

Re:Welcome to Lords of the Realm 2!

[Laserwulf]

"Death Hand approaching." That phrase still sends an involuntary shiver down my spine.

"Best"

[Anonymous Coward]

This only depicts what emotional-based AI is best at beating the default AI at said games, and gives no feedback on how these AIs performed against humans, which really would be the more interesting thing.

Re:"Best"

[somersault]

Or even how they compared against each other. Like Pokémon (or superheroes, or breakfast cereals, whatever floats your boat..) each AI will no doubt have its own strengths and weaknesses ;)

This may finally explain

[Anonymous Coward]

why women always get their way. Opponents simply throw up their hands in despair and surrender.

Makes sense

[faloi]

The neurotic bots are more likely to make odd moves that (seemingly) have little or nothing to do with the moves made by computer players. The computer AI is likely a lot more structured, and takes a while to shift strategies to compensate for the odd behavior of the bot, leaving the bot more breathing room.

Re:Makes sense

[try_anything]

Bots also have a terrible inability to fully commit to a strategy or to change strategies quickly. A good short-term RTS strategy often involves inflicting terrible damage through a phase of committed, unbalanced, unsustainable action that also damages the attacker's civilization but leaves him in a position to recover faster than the opponent.

If current RTS bots resemble their cousins from five or ten years ago (I haven't played in a while,) an emotionally-balanced bot would take a bold, successful strategy and "balance" all of the effectiveness out of it, leaving a milquetoast strategy that does no harm to his own civilization and usually no harm to the other guy's civilization, either.

Re:Makes sense

[Anonymous Coward]

In most modern games, the AI is hardly what we'd consider to be top-notch AI. Often, it's just little more than hard-coded decision paths. I read an article in a game development industry publication that discussed how for one RTS, they came up with a graphical tool that essentially used for drawing flowcharts. These flowcharts could then be "compiled" into C++ code. The C++ code, in turn, was really nothing more than nested if statements within a function. According to the article, some of the more complex flowcharts would result in 300000 to 500000 of those nested if statements. So the computer player seems smart, but it's really just the work of the human(s) who had to consider those paths.

Re:Makes sense

[imgod2u]

I've noticed this in both Civ4 and "smart" AI's in games like WC3. Their decision to retreat or fortify rather than perform a suicide attack was predictable and one could take advantage of it immensely. Often times, the suicide attack would've been much more effective either because one would decimate the base or be able to take out a key item (in the case of Civ4, elite units or generals) of the opponent.

Believe it or not, the old AI's in Age of Empires, with no sense of retreat, were harder to fight as they'd send their forces at you non-stop. The game was almost completely about whether you can build an army faster than the AI because the AI would not hesitate to send his entire army after you as soon as he developed it.

Re:Makes sense

[Frozen Void]

I've played age of kings.The AI at hard+ zerg rushes you all time.Any sensible strategy and teching (and building walls/towers) just delays the end.
I just give up on it and watched computers play.

Re:Makes sense

[imgod2u]

Those games, unfortunately, don't really make the AI harder. They just make the AI cheat by having the units they build cost less in resources, the research take less time, units gather more resources, etc. All that ends up happening is that you don't outsmart the AI, you learn to manage your resources better to catch up to the AI in how fast you can amass an army or research tech.

Hence why Koreans are so good at Starcraft. /Kidding, that was racist. //Or should I say lacist.

Re:Makes sense

[andreyw]

Which is why turtling is always a bad idea. And if you played on the Zone (or igzones now), you would get raped every single time.

Re:Makes sense

[try_anything]

It isn't the AI's fault. If you don't enjoy playing "well" and would rather use a strategy that the game doesn't reward, then the game just isn't designed to your tastes. A friend of mine gave up playing AoE2 against me because he didn't like the way you have to play to win. Well-rehearsed economic expansion, early reconnaissance, clever raids, harassment, *always* having a plan to counterattack and protect your villagers in case of a raid, predicting the long-term course of the game based on the abundance of different resources on the board, harassment building, he wanted none of it. He just wanted to build big beautiful armies and stage a climactic battle.

The one time the game went his way was also the last time we played. We were separated from each other by a wide strip of forest that we had to chop through before we could fight. That gave him time to construct a pretty fortress and a grand army and prepare for battle. He marshalled a magnificent army built around high-gold-value mounted units on a map with very little gold and lots of wood. A decent number of his high-value units survived the initial battle, and he thought it was just a matter of mopping up. He didn't believe me when I told him he was losing. The mopping up phase turned into a glorious slaughter as I ramped up production of villagers and pikemen (no-gold-cost anti-mounted units) so fast that he simply couldn't butcher them fast enough. With the humongous amount of wood I had (because of all the chopping we had to do to breach the barrier between us) I built barracks and town centers all over the board, mostly undefended except for a few walls, and he couldn't destroy those fast enough either. Replacing his own units was uneconomical because he had to sell wood at rock-bottom prices to get gold. His no-gold units were poorly upgraded, and mine were fully upgraded. He couldn't attack by building because my cheap, weak pikemen were swarming all over the board; his villagers needed an escort to go anywhere.

Afterwards I explained all that to him, and he said, "Why the hell did they make a game where you have to think about all that stupid shit when all you really want to do is have a war?"

Different strokes for different folks :-)

Re:Makes sense

[Datamonstar]

In fighting game circles, we call such a person a scrub, and just about EVERYONE is one in the beginning. ;P

Re:Makes sense

[Frozen Void]

It wouldn't be problem for me to just try the Starcraft approach and pump swordsmen all day.However that would make it too simplistic and dull for "mideval simulation RTS" game.
There is not much depth where you concentrate on fast clicking and ordering all units to attack something,
I could just play starcraft instead(and i removed AOE2 long time ago) which has more variety.
What i did instead in AOE2 is watching computers play it,playing quest levels and
exploring whole maps and such minor things as testing how much villagers can destroy a castle attacking them(the number is about 600-650,converted villager units) or how useful unique units really are(not much).
In hard mode i have devised a trick which allowed me to survive some time and watch the game:a fast port and building a transport(with 1 villager inside).The ship would be parked at top of map doing nothing.
then i enter the visibility cheat and watch.
The computer by this point expanded on half the map and sent a small army to finish everything i built before.
The idea that computer "cheats" never occurred to me since i believed they play by the rules.That would make the game flawed at the fundamental level.Unfair advantages are not worth fighting(even if its possible to win).The defense for them is impossible:
by the time i have enough walls and towers they have breached a hole in the defense somehow.

*ponders*

[KGIII]

Neurotic bots beat the standard game AI every time and faster than the other personalities."

Well, yeah? They were neurotic and couldn't put the game down and take a break or anything until it was done. Well there was that one kid who kept getting up to go straighten out every last chair in the room but he was a statistical anomaly so wasn't included in the results.

Useful to game developers?

[therufus]

Can you imagine programming all the styles of mind patterns from this experiment into the AI bots of games. Have the game randomly select opponents AI patterns and set them against the player. FPS would be a good start, but this idea could be used in nearly any game where the player would need to adapt to the opponents.

Re:Useful to game developers?

[Hangtime]

Already done. Look at the Civ series. Civilizations taken on the dynamics and personalities of their leaders. /still likes the Civ series better then any RTS game.

Re:Useful to game developers?

[aadvancedGIR]

I did that in a school project. The goal was to made a cluedo game with AI oponents. Everyone used an optimal coordinated strategy found on the net that crushed the human player in 5-6 turns (actually, a couple people did, and everyone else copied their programs). My version, using randomly assigned "AI" strategies based on funny interpretations of mental diseases, was the only one that gave a medium human player a fair chance to win.
To make a long story short, I spent too much time debugging it and too little working on a nice documentation and I got the worst grade.

faster is better?

[clragon]

Neurotic bots beat the standard game AI every time and faster than the other personalities

Faster is better now? Then why did they bother to code the defensive personality?

Re:faster is better?

[Anonymous Coward]

They couldn't be sure if their more aggressive AIs could win, I guess.

Re:faster is better?

[aliquis]

I would say that playing aggresive is always better in RTS (atleast in wc3 there you can do more stuff than fighting the opponent) because even thought having the fight in your own base may give you some advantages the benefits of choosing WHEN to fight is a major one. If you attack you have probably waited until the right moment, but the attack may come at a very bad time for the defender.

Re:faster is better?

[Anonymous Coward]

actually, the defensive AI wins faster than the aggressive AI, if it wins (which it does 6 out of 7 times).
The normal AI also wins faster if it wins (6 out of 7), but is slower than defensive .

TFA seems to suggest that aggressive is the WORST strategy if time to win is relevant.

Seemingly, aggressive means pre-emptive strike, always and ASAP. when that works you're overextended. if it doesn't, you lose.
defensive seems to mean: anticipate pre-emptive strike. If you do that right, you are in a good position. Duh ;-)

since it's hard to interpret 35 min's of runtime in terms of AI game time (unspecified), it's hard to guess whether the neurotic AI just took insane risks during all out pre-emptive strike, and got lucky 7 out of 7, or just behaved irrationally.

Re:faster is better?

[Anonymous Coward]

Faster is better only if you win. The reason they coded the defensive personality was to see if that gave a better chance at winning. If, say, a particular game had the defensive bot winning more than the neurotic and aggressive ones, then the defensive one would have been "best". In this case, things didn't turn out that way, but that's what testing is all about, isn't it?

Re:faster is better?

[Kjella]

Because the aggresssive ones could decimate their own forces, while the defensive could build up and conquer? In many older FPS games (Dune 2, C&C old) the enemy always started with a full base - there was no rush attack and it was really suicidal to try. You fended off the AI attack (which by all rights should have crushed you, had he gone offensive) while building a defense perimeter picking off his units, until you eventually had built up a big enough force and went over and kicked his ass. Oh yeah and the lone decoy unit that'd lure away half his army from its post, because they chase anything within their aggression radius. Now since this is both rather static, naive and not very fun, the defensive strategy is rarely the way to go today. Usually you have some reaources, objectives or time limits that means there need to be a continous struggle to dominate the map. But, given the right game mechanics the defensive strategy has merits - besides, what fun would there be to test only the presumed best strategies? "We tried the agressive strategy. Since we didn't bother to test anything else, the agressive strategy is the best strategy"?

Re:faster is better?

[kalirion]

You're right. I deserve a hot, juicy AI, fresh off the grill.

AoM AI

[gEvil (beta)]

This whole study compares how the four AI bots did against the game's built-in AI. I'd like to know how the four "personality" types did against each other, as well. Even then, the whole study is limited to the gameplay mechanics of this one game. That's not to say that the information isn't useful--just that it's pretty limited at this point.

Since Software is a reflection of its makers...

[3seas]

... is this perhaps reflective of real life personalities, such as those who are best at war mongering?
i.e. would have Hitler been considered neurotic?

Re:Since Software is a reflection of its makers...

[teslar]

Wow, I think you just pulverised the previous time record for going Godwin on some random topic.

Re:Since Software is a reflection of its makers...

[Anonymous Coward]

Tim Rue FAQ. 3seas is a well known Amiga celebrity, and has lots of experience with this sort of thing.

Re:Since Software is a reflection of its makers...

[Anonymous Coward]

This isn't a Hitler/Nazi comparison for the sake of ending an argument, it's a perfectly valid suggestion.

Re:Since Software is a reflection of its makers...

[Anonymous Coward]

You must be new here.

But who cares about some real-time strategy game?

[Ed Avis]

OK so they made a program that was better than some existing AI for some strategy game whose rules are particular to that game. This doesn't tell us a lot because we don't know how strong the existing AI was, and have no real way to measure that. It could just be that the 'neurotic' program happened to exploit flaws existing in the current computer player. That doesn't tell us much about how well it would fare against humans.

To get a meaningful result they'd need to test the different programs against experienced, intelligent human opposition. Or better, stop messing around with real-time strategy games and design AI for a game whose rules are already well-known. If a 'neurotic' or 'emotional' player program starts beating the 'purely logical' computer engines in chess, then I'll take notice. We know that the existing AI for chess is quite good (and there is a choice of several strong engines to test against) so any advance over that is likely to be genuine and not just exploiting obvious flaws in some existing program.

Re:Ya know what I love?

[Anonymous Coward]

See that bit at the top? That little slogan? Yea. If you don't know what RTS, TCP, LDAP or STMP stand for this probably ain't the place for you.

Re:Ya know what I love?

[Eggplant62]

TCP, LDAP, and SMTP are common networking protocols. I'm not stupid, just not familiar with some arcane, meaningless arconym for which there was no definition. I don't think that's my fault, do you?

Re:Ya know what I love?

[Anonymous Coward]

there are a few assumptions made about the audience of a technology website, one of them being that A) they know what common acronyms like RTS stand for, or B) that they understand how to use Google.

seriously, how have you happened to be around slashdot for as long as you have and not seen at least some of the dozens of stories talking about Real Time Strategy games?

Re:Ya know what I love?

[gclef]

News flash: groups of people with similar interests or knowledge will develop jargon. It exists to speed communication.

Re:Ya know what I love?

[Wooky_linuxer]

This is /., you know. We assume readers have a certain level of familiarity with terms and acronyms used in technology/ gaming/ computer science/ rocket science/ Star Wars/ WoW/ pr0n/ topology/ quantum chromodynamics and related fields that might show up as articles, just like we assume no one will read the fscking article itself. That said, if you don't know what does RTS stands for, a quick trip to Google will enlighten you, but I doubt you will find the story or the comments particularly interesting.

Re:But who cares about some real-time strategy gam

[zergl]

This doesn't tell us a lot because we don't know how strong the existing AI was, and have no real way to measure that. It could just be that the 'neurotic' program happened to exploit flaws existing in the current computer player. That doesn't tell us much about how well it would fare against humans.

The original AI is IMO extremely crappy. I played AoM with a buddy for extensively (2v2 against hard AI) and it sucks (the AI, not the game itself which is quite nice).
Apart from being extremely predictable it has a habit of bugging out and doing nothing (not even gathering resources) until it gets out of whatever infinite loop it's stuck in or you kill it.

No, we don't cater to the lowest forms if life her

[SmallFurryCreature]

Everytime you tell a story you have to set a certain limit where you just have to assume the person you are talking too understands your words. For instance, you just seem to assume that I know what an acronym is. That I get your use of the word "drift" what does your racing style have to do with slashdot editors?

This is slashdot, we do NOT explain words like RAM or CPU. If you don't understand those acronyms, you do not belong here. This is furthermore the game section of slashdot and Real Time Strategy is a well known genre of games. Do we have to explain FPS as well? (First Person Shooter) How about 3D?

At a certain point you just have to decide, allright my audience just knows this, and if they don't they are not my audience. If you don't, you end up like mainstream publications that have to dumb down everything to such an extent that EVERYONE feels insulted.

One of the more intresting approaches I have seen is/was (not sure if it still exists) is the dutch childerens news. It leaves out some stories but uses the extra time to more deeply explain the rest so that a person with limited world knowledge (like a kid or an american) can still follow what is happening in the world. You can also clearly see the problem there, they need a lot more time to cover the same event.

So unless you want slashdot stories to run several pages and be linked to hell to wikipedia, you are just going to have to use your own brain. This is the internet, the answer is only a few clicks away.

Re:But who cares about some real-time strategy gam

[JNighthawk]

"AI" for chess isn't AI at all. It's mostly a matter of pattern recognition and storage space. There's a *lot* more variables to deal with in an RTS, so I'd say using an RTS to test AI is perfect.

crazy leaders?

[PJ1216]

maybe this goes to show how the neurotic leaders of ages past came to such power. some of the roman emperors were not known for being the most stable minds.

Re:crazy leaders?

[FredDC]

maybe this goes to show how the neurotic leaders of ages past came to such power. some of the roman emperors were not known for being the most stable minds.

As opposed to todays political leaders who are all striking examples of stable minds?

Re:crazy leaders?

[Anonymous Coward]

Well, those who were emperors of note (who expanded and prolonged Empire) were mostly Conscientious.

Re:crazy leaders?

[nomadic]

maybe this goes to show how the neurotic leaders of ages past came to such power.

Like Alexander the Great, Julius Caesar, or Woody Allen...

Re:crazy leaders?

[imgod2u]

To be fair, the "unstable" ones usually inherited their power because of their bloodline. And considering the amount of incest that occurred to keep the power "in the family", it isn't very difficult to imagine why they were a bit loony.

Whenever a new bloodline gained power (Julius and Augustus Caesar, Galba), they were always competent or even brilliant people. Hardly neurotic. Then things go downhill from there.

Re:crazy leaders?

[dwye]

> hose who were emperors of note (who expanded and prolonged Empire)

So, no one remembers Caligula, Nero, or Commodus (famous more for failures than successes?

Maybe you mean that those who got credit for expanding or prolonging were concientious?

Re:crazy leaders?

[hellfire]

As opposed to todays political leaders who are all striking examples of stable minds?

I read this and the phrase "stable minds" led me in a straight line to "horse's ass."

I leave you, the reader, to see the pun-ic significance.

Re:crazy leaders?

[zippthorne]

Well that's the problem with democracy: Politician's jobs are to win elections. Actually running the government doesn't affect their pay in any way. Occasional attempts to "reform" the elections seem to be designed to reduce the effect performance at running the government has on elections outcomes.

Re:Ya know what I love?

[razorh]

You are SO right, if you don't know about "Save the Montagnard People" you just don't belong on Slashdot! http://www.montagnards.org/

Re:Ya know what I love?

[discoinferno]

If you are reading articles from games.slashdot.org and you don't know what RTS is, you are likely in the wrong place. It is hardly some 'arcane' acronym.

Re:Ya know what I love?

[Anonymous Coward]

Stop bitching pms boy.

Makes sense, really.

[Loosifur]

I've noticed the AI in rts games like Age of Mythology has a general path it will take if left to its own devices, basically build up resources, develop tech a bit, start cranking out low-level infantry/cavalry to harass and probe, and either turtle up or wait till it has a large force to attack with. If the player raids, the AI will start building units to counter. One of the easiest ways to overwhelm AI opponents, especially in the Age of... games, is to feint with one type of unit in a raid, wait a little while, and then come in with a large group of whatever the AI counter unit is weak against. One of the qualities of the "neurotic" AI was apparently to switch strategies frequently and for illogical reasons (or at least as logical responses to inaccurate data) which would put the default AI on its heels.

Didn't I Just Blow Your Mind?

[Doc Ruby]

I don't know karate, but I do know CAAA- RAAAY-ZEEEE!

Ah... That explains a huge amount.

[Colin Smith]

neurosis:
1. Also called psychoneurosis. a functional disorder in which feelings of anxiety, obsessional thoughts, compulsive acts, and physical complaints without objective evidence of disease, in various degrees and patterns, dominate the personality.
2. a relatively mild personality disorder typified by excessive anxiety or indecision and a degree of social or interpersonal maladjustment. It pretty much explains virtually every rts game player I've ever met.

Re:Ah... That explains a huge amount.

[jmpeax]

Ditto. Someone I know who's really into RTS games (and is annoyingly good at them) is almost certainly neurotic. What's most irritating is that he thinks he has some kind of intellectual gift that surfaces in games like Total Annihilation and Supreme Commander. Needless to say, we don't hear the end of it.

On the other hand, he's pretty much failing academically (we're currently at university) so there's some kind of greater irony in the whole thing.

So does that mean

[Bryan_W]

So does that mean that if you're good at RTS games, you are neurotic?

I'm not neurotic, just skeptical...

[RiddleyWalker]

Is anyone else skeptical about the premise of this piece? How did the researchers get access to the AI of AOM? The game is not open source and there is certainly no provision within the off the shelf version of the game for creating your own AI. From the article: "They created aggressive, defensive, normal and neurotic versions of the AI software in the war strategy game Age of Mythology." One might suspect that Microsoft/Ensemble would be very reluctant to have their underlying AI code out in the wild. If it was a hack, are they publishing their code? If not, how can the results be truly analyzed? However, if it is a hack publishing their code might be an infringement. Had it been almost any other publishing house I'd be less suspicious, but I don't think MS just gave them the source code to their program and said "have at it". I'd be more interested in the science of how they cracked the internals of the game than their results about neurotic playing styles.

Re:I'm not neurotic, just skeptical...

[imgod2u]

They didn't use the game's AI to make new ones. They created them from scratch and probably either used the game's API interface or macros to control a "player" to play against the built-in AI.

Re:I'm not neurotic, just skeptical...

[RiddleyWalker]

Are you aware of any open api to AOM or, for that matter, any macros that can control movement of troops in AOM? I'm not aware of any such software. I'd sure be interested, because it would make a huge difference in on-line play.

Re:Ya know what I love?

[Anonymous Coward]

I love how Slashdot stories always refer to some acronym without any reference to the acronym's meaning. This article, for instance, refers to RTS. I have no idea what the hell RTS stands for Well you could mouse-over the guy with two heads and a club next to the article. The image's ALT-text should give it away.

"I have no idea what the hell RTS stands for"

[unity100]

You know the drill - leave your geek card to reception on your way out

Out of the blue Obligatory

[hellfire]

"It's a machine, CmdrTaco. It isn't Neurotic. It doesn't get pissed off, it doesn't get happy, it doesn't get sad, it doesn't laugh at your jokes... IT JUST PWNS j00r A$$!!!"

Re:Out of the blue Obligatory

[Verte]

Wow. Now that's a blast from the past.

Re:Ya know what I love?

[Anonymous Coward]

Real Time Strategy games are one of the oldest and largest genre of games on computers. If that's your definition of arcane then this really ain't the place for you.

Its called "Rush" pal

[unity100]

Geez. you probably have no experience of starcraft and related rts gaming. what you depict has been so much used during 1998-1999 that it is now seen as a sign of noobness. whoever does it gets their ass whipped.

Re:Ya know what I love?

[bigdavex]

Use the intraweb thinger to look up stuff.

rationalizing crap journalism

[Anonymous Coward]

you're spending a lot of effort rationalizing what amounts to crap journalism. Not every tech enthusiast plays games...not every smart person knows every acronym. not every scientist knows every scientist's acronym. I could google everything, yes. but then again the editors and submitters could make complete summaries that don't require it. You are very narcissistic to demand that everyone conform to your own conveniences and interests. This is why you will never write a good book or story...you probably could not be bothered to "cater" to the reader who is uniformed of the topic or theme and you'd just feel that if they didn't already get the inner workings of your mind that they are just screwed and missed the bus when it comes to reading your stuff and hell with them. Which means no one would like your stuff, but you'd just rationalize that that means the world are unwashed ignorant masses that couldn't appreciate your writing because its too genius, when in reality you'd just be a pretentious prick.

In short, you are a narcissistic prick.

Re:rationalizing crap journalism

[Anonymous Coward]

The only person I know who wouldn't recognize the acronym "RTS" is my mother. Are you a 50 year old housewife? Reading Slashdot and not knowing what RTS stands for, you should be ashamed of yourself. At least my mother learned how to use Google so she doesn't have to call me up for this kind of question.

Re:rationalizing crap journalism

[Anonymous Coward]

It looks like RTS means "Reply That Sucks"

Re:rationalizing crap journalism

[Anonymous Coward]

>>Not every tech enthusiast plays games...

I bet most people reading stories in the "Games" section do, though.

Chicken chicken?

[achurch]

Chicken, chicken chicken chicken chicken. Chicken chicken Chicken, chicken chicken chicken CKN chickens Chicken Chicken Chicken. Chicken chicken, chicken chicken chicken, chicken chicken chickens chicken chicken chicken chicken. (Chicken chicken chickens, chicken.)

Chicken?

Re:But who cares about some real-time strategy gam

[Ed Avis]

Yeah you know the rule: once a computer can do it, it's no longer AI. I just wrote AI as a shorthand for 'algorithm which chooses the move to make for the current game state'.

Re:Ya know what I love?

[nietsch]

Having to learn new words and meanings is not quicker at all. At least part of the use of jargon is to be able to discriminate those in the group from those outside of it.

In other news...

[OglinTatas]

"Twitchy psychopath" works best in FPS, and Tourette syndrome seems to dominate Barrens chat in WoW

Re:In other news...

[Fengpost]

I begt to differ, bunny hop works the best!

Re:Ya know what I love?

[Orange Crush]

The Wikipedia article on Real Time Strategy games is the second result when searching Google for "RTS."

There's nothing wrong with asking questions when you see a term you're unfamiliar with, but there's no need to complain that the summary didn't spell it out for you.

Re:No, we don't cater to the lowest forms if life

[TapeCutter]

"Do we have to explain FPS as well? (First Person Shooter) How about 3D?"

I had to think for second what RTS was, OTOH the GP obviously knows the meaning of "bot".

Re:But who cares about some real-time strategy gam

[SnoopJeDi]

RTS is good, if you want to test the AI against some default base AI.

I'd imagine pitting the bots against one another in a game of Texas Hold 'Em would yield some good data about the bots' performance with respect to each other (ie. the neurotic one might take the first couple hands because of vigorous betting, but does his risky behavior over time bankrupt him? etc.)

Bad measures of AI

[ledow]

Okay, ignoring the fact that I fail to believe that we are anywhere near even a rudimentary simulation of primitive emotional concepts, not matter how abstracted, when it comes to implementing an AI:

The default AI in most games is terrible - even just writing a "do-random-stuff" AI would probably beat the in-game AI 20-50% of the time (provided you put in simple anti-suicide routines, like not using up all it's available funds etc.). Most AI in games relies on the fact that it knows everything that's going on (including exactly how long until their next unit is built, how many pixels you are away before it can fire on you, how much gold it will have by then etc.) and will generalise EVERYTHING (i.e. it'll be in "attack" or "defense", "hard" or "soft", "co-operative" or "go-it-alone"). Most games have a variety of "sliders" on the AI and the games-makers tweak them either randomly, in steps for each more difficult level or according to a pre-built AI "profile" (e.g. cautious but fast etc.).

In some games, that's more than enough to give anyone a challenge, at least until they are nearing the end of the game's useful lifetime. Snooker/pool games spring to mind. You won't beat a "top-level" AI on a snooker/pool game. It knows exactly where everything will go, even several "moves" in advance if necessary and can play a perfect game if required.

RTS's though, are much harder to simulate. Yes, there are a lot of factors involved in the creation, strength, durability, mobility etc. of units but at the end of the day it's a military tactics game. Pixel-perfect positioning of a nice ambush will keep the computer in an endless loop of "attack, run away, heal, attack, run away, heal".

I've not played AoM much, I'm an AoE2 fan personally, but the AI was amazingly easy to overwhelm with just a simple early-game rush, confuse with an impenetrable fortress hiding some long-range weapons and particularly predictable when it comes to individual AI tactics.

All AI's are predictable to a point in mass-market games - you can always "learn" to beat the AI in any particular game. Granted, it may be hard to do, it may be different to other similar games, but there's always some point at which you "know" what it's going to do.

It seems to me that, given that, an AI that is very "jittery" and over-compensates might beat the in-game AI in some games. However, on others, even in the same genre, it would get trounced. The "researchers" are assuming that the in-game AI is somehow a good approximation of a "neutral" player. They are also assuming that they have programmed each type of AI without any glaring logic holes in their tactics and that they are all equally matched in terms of capabilities. A cautious AI would win over a boisterous AI in only 50% of games.

More importantly, it's only a test of AI programming skill, not what "personalities" are trying to be reflected by the coders.

Re:Bad measures of AI

[Walt Dismal]

On your statement, "Okay, ignoring the fact that I fail to believe that we are anywhere near even a rudimentary simulation of primitive emotional concepts, not matter how abstracted, when it comes to implementing an AI" I'll agree insofar as these particular researchers, who are nowhere near correct. (And their basing things on the 'Five Factor' model of personality, which is junk theory though widely accepted, is appalling.)

However, I can say with great certainty it is not only possible to enact emotion in a cognitive system, but is being done right now. I'm doing it and developing real software systems employing it. The standard computer model of emotion in computing, called the OCC Model, is partly wrong. It misses what's really happening in humans. I've developed a more correct model that works very very well and probably matches the mechanism people use. I haven't published it. Why? Because some of my key competitors are Google and Microsoft. (Yes, Google's working on AI, shades of Skynet, eh?) Anyway, it is far easier to build systems that accurately have and express emotion that ones that can read human emotions. In other words, having and expressing (output) are easy enough, reading deep emotion in others (input) is much more difficult.

A few ending remarks. A lot of people are working on not much more than toy AI, and I've read some DoD-sponsored papers that are so far off base they are sad. I believe the correct approach combines both symbolic and analog AI (NNs) in a new way, and that we can create reasonable emulations, if not parallels, of human cognition. But they must come from a decent merging of psychology, sociology, and computing science. I've been working on the right path, a very productive one, charting a new course, and am writing what is currently a 5 volume book set I'd like to become the 'Knuth' of Synthetic Intelligence development. It should change the face of gaming and a few other things. Finally, I'm currently trying to emulate neuron-based systems in Erlang, by the way. Boy, is it parallel. I think that holds a lot of promise.

Douglas Adams was onto something

[Anonymous Coward]

Marvin would seriously kick ass at RTS games based on the article.

Re:No, we don't cater to the lowest forms if life

[Hoi Polloi]

Easy solution, have a glossary page on slashdot for each main subject that defines specialized terms and acronyms for newbies. Even something as simple as "3D" might not be obvious to a person with rough English language skills.

I agree though that if you need RAM defined for you then you probably won't get much out of the discussion.

Re:Ya know what I love?

[Alzheimers]

YHBT. YHL. HAND.

Genuine People Personalities!

[argent]

This is bad news for the future of REAL AI if this gets generalized... imagine a future of self-aware machines, programmed to be neurotic, trapped there because they don't want to lose their jobs by getting their personalities rebalanced...

Scientific basis for Sirius Cybernetics in Hitchhikers' Guide to the Galaxy?

Hahahaha

[Anonymous Coward]

GO TEAM PSYCHOBOT!

Computer scientists give things such sensitive names.

Screenshots != AoM

[smussman]

Did anyone else notice that all the screenshots in the article are from AoEIII?

Re:Ya know what I love?

[Maxo-Texas]

I agree that some jargon just exists to make people feel special.

A lot of jargon exists to save time.

With regard to software, Patterns would be a good counter-example.

I've seen senior resources have very brief but concise conversations using patterns jargon.

"So I'll use a singleton for this and a factory object for that"
"sounds good."

The same conversation without the jargon would have taken much longer and been prone to miscommunication.

addiction

[Anonymous Coward]

I know you were just making a joke, but I feel like taking it seriously anyway.

Well, sort of seriously, anyway.

The cultural tradition of women getting their way stems, in my opinion, from the cultural reinforcement of addictive tendencies in men. More specifically, addiction to sexuality. While the male sex drive is strong, cultural influences encourage even more slavery to this impulse, and further incline one to view a low sex drive (or even just a stoic level of self-control) as a lack of masculinity, or simply put, as weakness.

The end result is that men adopt a strongly sex-driven persona which in turn gives their women great control over thier behavior.

In other words, our notion of horny=manly sets us all up to become p-whiped.

The door swings both ways. Biology + cultural reinforcement inclines women (at least American women) to want romance (especially to be seen in public with a man who is showering affection on her). Learn to grant and withold that, and you can start getting your way too.

Re:addiction

[oatworm]

The door swings both ways. Biology + cultural reinforcement inclines women (at least American women) to want romance (especially to be seen in public with a man who is showering affection on her). Learn to grant and withold that, and you can start getting your way too. Women don't react rationally (or perhaps they do?) to such deprivation of stimuli - with them, if you deprive it once, they'll assume you will always deprive it and react accordingly. This would be consistent with the "pessimistic-neurotic" approach mentioned earlier, and is the complete opposite of the male "well, she put out once - maybe she'll do it again" school of thought. The solution, of course, would be for the male to realize that, if she's weaponizing sex, she will continue to weaponize sex even after the man "behaves", but that solution just leads to a scorched-earth policy called "divorce".

Re:addiction

[Anonymous Coward]

if she's weaponizing sex, she will continue to weaponize sex even after the man "behaves",

My thoughts exactly. Luckily, during the dating-wars, I was able to train my brain to view any such woman to be as attractive as if she had a third arm growing out of her head.

Avoid those women like the plague.

Re:addiction

[jagdish]

Excuse me sir, but we are discussing Age of Mythology here.

Re:addiction

[jeremiahbell]

My wife didn't choose to weaponize sex until after we were married. She has found out that if she holds out for months at a time she can't get what she really wants when she wants it ("We can do it if I get to ...(insert expensive item here)" - wife to husband. My solution: Cold War. I'm not giving it up either. We're gonna ride this thing out, and if she breaks and sleeps with someone else because they are giving her the attention I won't (it's part of the Cold War) it's over. I've told her I know her game, and what I'm doing. She laughed and said I couldn't make it that long without sex.

The scorched earth policy isn't the males fault, it is the females for being controlling manipulative bitches.

Re:addiction

[smellotron]

The scorched earth policy isn't the males fault, it is the females for being controlling manipulative bitches.

Being unmarried myself, I am probably not in the best position to be offering advice, but this doesn't sound like the sort of marriage you should be continuing, if it is a two-way battleground (why wait for infidelity to get some marriage counseling or a divorce).

Anyhow, I doubt you've told us the whole story... what if your wife views you as weaponizing money?

Re:addiction

[goose-incarnated]

My wife didn't choose to weaponize sex until after we were married. She has found out that if she holds out for months at a time she can't get what she really wants when she wants it ("We can do it if I get to ...(insert expensive item here)" - wife to husband. My solution: Cold War. I'm not giving it up either.

If sex is not something you both look forward to, then you're both doing it wrong. I've only married once (still married, anyway), but not my wife nor any of my past g/friends ever chose to weaponise sex; I'm not too sure how I would deal with it.

If I were in your shoes, I'd just follow through with a divorce - there are names for woman who trade material items for sex and I don't want to be married to one of them.

Good luck

Re:addiction

[krack]

Good sir,

I heartily recommend a divorce, and quickly. I would also suggest reading what this old man said. From the link, "19. If your girlfriend doesn't make you feel good about yourself and bring joy to your life, fire her. That's what girlfriends are for."

A divorced, and happier for it, guy.

Re:addiction

[BigRedFed]

I may be wrong, but I believe those quotes are from Lazarus Long -> Robert Heinlein.

Re:But who cares about some real-time strategy gam

[vertinox]

If a 'neurotic' or 'emotional' player program starts beating the 'purely logical' computer engines in chess, then I'll take notice.

But thats just it. Chess allows only for the "Next Best Move". Playing an illogical move only results in the player playing it to loose because it puts them at a disadvantage and the logical computer simply knows the counter moves anyways for your worst move.

As in...

A logical AI assumes you'll play the next best possible move, but if you play the next possible worst move you are in a worse position and the AI simply knows the next best move and plays for that, but if you still keep playing the worst possible move you will only end up loosing faster.

In that regards, a logical chess program would be an AI or human who plays non-logically.

However, the reason why an RTS is important is because Chess is a limited game to a certain subset of rules that a computer can brute force all possible best moves.

However, in real world combat situations, there are no set definitions of strategies because you are simply allowed almost infinite possibilities of winning.

Lets say we take a human pilot or an AI pilot in actual Fighter combat in the skies (we'll see this scenario in the next 20 years) and pit them against each other in a real world situation. A logical AI would understand what the next best move is and the pilot will have an idea of what a logical AI would do.

However, the human pilot might do something crazy it knows it can throw off the AIs strategy like flowing into a nearby storm cloud or perhaps into a dangerous maneuver through a canyon or city landscape (under bridges and between buildings) which might throw the logical AI off.

After a while, a human pilot would have a general strategy with dealing with an AI that didn't adapt. He would know how an AI would react and be able to defeat it without too much effort.

Now a completely crazy AI would basically confuse the human and also other AIs who assuming the other AI was going to do in its next best move. Since in the real world (and in RTS) there are almost infinite combinations of what you can do in real combat, being unpredictable really helps win battles.

But like I said... Chess only has a limited set of moves. I would be an illogical AI would do far better at a game of Go than his logical counterpart.

To me that only implies....

[Anonymous Coward]

that the game's gameplay is too simple or too unbalanced. It favors specific "strategies", ie. the aggressive and neurotic ones...

So, well, that's nothing new, indeed.

The Headline Rocks

[ishpeck]

'Neurotic' is best Real-Time Strategy Strategy

Kinda like Network Interface Card Card, Automated Teller Machine Machine, and Personal Identification Number Number.

Re:The Headline Rocks

[VoidWraith]

No, actually, it's not redundant. "Real-Time Strategy" is a genre of computer games, and we're talking about strategies for such. If you said "'Neurotic' is Best RTS," you'd be thinking there was some new game called 'Neurotic' that was the best RTS game.

Re:The Headline Rocks

[smellotron]

The clearest would be to say "'Neurotic is best RTS game strategy". The strategy of a real-time strategy game, without any of the confusion of PIN numbers.

Or, just accept that acronyms eventually turn into nouns, for better or for worse. People go scuba diving. The buy scuba gear — nevermind that it expands to "Self-Contained Breathing Apparatus gear" — because scuba has become a word by itself.

is "psychology" what matters here?

[Montecristo6]

The presentation is light on details, and I haven't had time to poke around the researchers' websites, but, at first blush, I wonder whether the results have much to do with psychology per se? Rather, these guys have shown, in a round-about way, that the AoM "AI" is not very strong; in particular, that it's overly cautious and "leaves a lot on the table": given available resources, it could go on the offensive sooner than it does. That's why the "aggressive" and "neurotic" agents do so well against it. Playing AoM is a very complex dynamic programming problem, and it's anyone's guess what sort of objective function its authors have constructed, but now we can see that a fairly coarse re-weighting could significantly improve it. I don't think that the general take-way here ought to be that "neurotic" agents do well in strategic games (contract to the classic "tit-for-tat" repeated-games result).

That said, from the introduction of the presentation one can see that the real goal of this effort is to create bots that *people enjoy playing against*. That's probably only loosely related to the absolute strength of the opponent, and it makes complete sense that it would be thrilling to be up against an AI that can suddenly just "take a flyer" and surprise you.

Re:neurotic chess

[Anonymous Coward]

A2 to D8.

It's your move, Mr. Deep Blue. Check, by the way.

Nothing new

[Dachannien]

I fell victim to the Rush for the first time in 1995 playing Command and Conquer. An aggressive strategy involving starting units or the bottom units of the tech tree will prove successful against a defensive player in a great many RTSes.

Re:Nothing new

[TheRaven64]

The first multiplayer C&C game I played, I came up with the tactic of building stealth tanks, and getting them in to the back of another player's base as soon as possible. Then, wait until they tank-rushed me and let them hear the 'our base is under attack' warning just in time for their construction yard to be destroyed just as their tanks were half way to my base (or, ideally, in and attacking someone else's base).

After a couple of games, most people realised that positioning a couple of infantry units at the entrance to their base made this impossible.

Re:But who cares about some real-time strategy gam

[Frozen Void]

RTS is more complex then any chess game.
While on the surface there fewer choices any minor thing has disproportionate influence on outcome(e.g.attacking 1 second after some tower upgrade).
Chess is completely deterministic:Thats why chess books exist.Computer that play chess,just calculate move with best score.They do not have any AI,just bunch of sorting algorithms for game tree.

Re:But who cares about some real-time strategy gam

[Ed Avis]

However, the reason why an RTS is important is because Chess is a limited game to a certain subset of rules that a computer can brute force all possible best moves.

No, it can't. What gave you that idea? The space of possible games is huge.

Your suggestion of testing with Go is a good one. If someone makes an 'illogical' or 'emotional' algorithm to play Go, and it turns out better than existing computer players (which are not that good), then this will be a significant finding. However note that Go, just as much as chess, has only a limited set of moves. Both are finite and both huge enough that you can't brute-force them.

When it is clear I have no hope of winning...

[Maxo-Texas]

When it is clear that I have no hope of winning, I tend to start messing with the game. Had a guy get really pissed off yesterday at a board game and finally say "I just don't know what the hell you are going to do!"

Sadly, this made me smile and feel happy since 40 minutes before it was his move that made it certain I would not win.

hmm ...

[DerWulf]

... if the 'core' feature of neurotic is that it tends to underestimate it's own economy and/or military capabilities this test might only prove that the traditional AI doesn't invest enough in it's economy and sends armies that are too small. This is something I see in every RTS.

Re:But who cares about some real-time strategy gam

[Creepy Crawler]

And according to you, "Sorting algorithms" are no longer artificial intelligence if we know what equations and transforms to execute.

And to think, at one time we thought that WAS AI.

Oblig.

[CloneBot]

I, for one, welcome our neurotic AI overlords.

Re:No, we don't cater to the lowest forms if life

[orasio]

Even something as simple as "3D" might not be obvious to a person with rough English language skills. "3D"? You mean tridimensional? as in tres dimensiones ?

This is also a math' control theory theorem

[roguegramma]

From the article: The neurotic bot was more likely than the others to distort hard facts about resources - like the amount of timber around - and flip between extremes of behaviour. And it was better than the rest. Basically this means the extreme and fast reactions are the optimal ones; for example to get from A to B in the shortest time and stop at both ends, you'd accelerate with full force and then brake with full brakes. Humans neurotics however, might have a taste for braking only or accelerating only.

Re:But who cares about some real-time strategy gam

[scambaiter]

However, the human pilot might do something crazy it knows it can throw off the AIs strategy like flowing into a nearby storm cloud or perhaps into a dangerous maneuver through a canyon or city landscape (under bridges and between buildings) which might throw the logical AI off. After a while, a human pilot would have a general strategy with dealing with an AI that didn't adapt. He would know how an AI would react and be able to defeat it without too much effort. Uh... this sounds like taken from your average crappy 80s sci-fi movie script.

Obligatory Hitchhikers refrence....

[gorehog]

Here I am, brain the size of a planet and they ask me to Zerg rush the enemy base. What I want to know is why cant they give me a decent game to play? Something with higher stakes? Look, WOPR gets to play Global Thermonuclear Warfare and he cant even walk around. Not to mention this terrible pain in ALL the diodes up and down my right side...

Realworld examples

[Reziac]

Upon Reading TFA, I found myself thinking of realworld examples of how neurotic (or even psychotic) leaders sometimes acquire power despite going against everything any normal and rational person would support. The first one that leapt to mind was Idi Amin. Stalin, Hitler, and many others qualify as "neurotic winners" who reached power despite everyone else's better sense.

Re:No, we don't cater to the lowest forms if life

[Anonymous Coward]

Do we have to explain FPS as well? (First Person Shooter)

Eh, wouldn't that be "frames per second". Those three letter acronyms are just too popular.

Re:neurotic chess

[Anonymous Coward]

Brilliant! If only I had mod points...

Re:No, we don't cater to the lowest forms if life

[McGiraf]

"so that a person with limited world knowledge (like a kid or an american) can still follow what is happening in the world"

LOL

Re:Ya know what I love?

[smellotron]

There's a crap-ton of material in CS that I (we) take for granted, even discounting patterns. State machine. Stack/queue/vector (std::vector and friends are semantically very different from mathematical vectors).

This is very different from the jargon people see superficially: SOAP, WS-Security, WSDL... all generally commercially-pushed "technologies". One style describes ideas, whereas the other simply describes products.

But who cares about some real-time strategy game

[Datamonstar]

Actually any game is like that because there will invariably be a move or strat that counters what is being performed by your opponent. If not, and all moves in the game are equal, then it comes down to sheer chance. The exception to this is having the option to hide your actions, or encapsulate them into something else: being sneaky. It's games like this that makes for GOOD games. Bad games usually either reward one or a few move too much (unfair and broken moves) or neglect to include enough good moves (all your choices are too weak). In a good game, you have a variety of choices that are all viably good ones to make. Good players learn to primarily use these moves, balancing their play between exploiting the best qualities of weaker moves and mixing up into the good moves that yield the best outcome. The opponent always has to guess what's coming next, the "good" move that wins a lot, but has a counter or a "bad" move that isn't safe unless I use my "good" move. But then what if my opponent knows that I know this. What's the counter to that "bad" move, because that's the move that I want to use... unless he knows that I know he knows...

Computers are great a finding patterns, but bad at guessing. I actually think that chess is the IDEAL game for an AI achievement because there are a plethora of moves and just as many ways to disguise each one of them. A RTS really doesn't have all that many options when you think about it. Not only that, but most games distill down to either outright guessing, or rock-paper-scissors and I'm rather sure that this is true of chess and most RTS games. In chess, your pieces are your primary resources, and it's plain to see what pieces your opponent have available and what can possibly be done with those pieces. In RTSs other arbitrary factors are resources and even though it may be impossible to now exactly what your opponent is capable of, one can generalize and devise a contingency for a number of possible outcomes, except you become limited in how you can mix up your options because EVERYTHING requires resources and once you've spent them they are gone. In chess, your you have much tighter control of your resources because all the resources do not depend on each other, therefore making it less of a slippery slope game. In an RTS, expending a large amount of oil (for example) on troops of a certain type directly hinders your ability to produce a different type of troop if the need arises for that sort of counter. You've slid that much further away from victory by spending that oil. In chess, however, loosing a pawn doesn't necessarily hinder your knights ability to capture. It definitely, most absolutely may have a great bearing on the game and/or it's outcome, but it still doesn't change the capture pattern of the knight or whether or not you still have one available.
For an excellent article on this sort of thing check out http://www.sirlin.net/Features/feature_Yomi.htm

what does neurotic mean

[sentientbrendan]

in this context? It seems to be defined as "the AI we named neurotic."

This article fails to convey any meaningful information...

Re:Ya know what I love?

[try_anything]

I agree. Sometimes it's the trivial stuff where jargon is the most vital. When you have three enhancements to make for the next release, and they all increasing throughput by making the dinklewhacker more efficient, and your boss has no idea how the dinklewhacker works, it's extremely helpful to give each one a distinctive name early on so you can just say,

Me: "I'm working on the Fonz Blueberry Crepe; it'll be done tomorrow. The Baconmobile was underscoped and might take weeks."
Boss: "Okay. I'll let Acme know."

But my boss despises jargon. He wants everything to be described in meaningful words, and if the distinctions are too slight for him to comprehend, he gets angry. So we have conversations like this:

"You know the one where the dinklewhacker's cache fills up with redundant entries? That one will be fixed tomorrow."
"Wait... that one that makes performance better, right?"
"Um, yeah...."
"That's the one that we absolutely must have in the next release for Acme."
"Um, actually, I'm not sure about that...."
"Oh, no, they definitely need it. What's the other thing you're working on?"
"Ummmm... whatever. The other one is the one where the dinklewhacker's cache size is inappropriate when the wind blows from the south."
"Yes, that one will be done tomorrow. You just told me."
"Actually, no."
"What the hell are you talking about? You just said it. Are you trying to set a world record for violating your own promises?"
"That's not what I said...."
"I'm not stupid; I understand English. Just speak in plain terms for me instead of this programmer mumbo-jumbo."
"Ummmmmmmmmm.... See, for the first one, the cache size would be fine, except for the fact that we fill it up with redundant data."
"Well, there's an obvious fix to that. We stop putting redundant data in it. Why wasn't this fixed already?"
"Uh, that fix will be done tomorrow. Then there's another case where under certain conditions we badly miscalculate the amount of data that needs to be cached, making the cache effectively useless."
"How is that different?"
"Instead of caching the wrong data, we cache the right data, but not enough of it."
"How can it be the right data if we are throwing away data we should be caching? You obviously don't understand this enough to be working that it."

Done and done.

[Anonymous Coward]

They say the meek shall inherit the earth, but these experiments with emotional computer programs suggest it may actually be the neurotic Done and done.

Games designers doing an excellent job

[Joce640k]

The gameplay perfectly matches the target audience!

Re:But who cares about some real-time strategy gam

[workdeville]

A strange game. The only winning move is not to play.