The Other Worlds Shrine

Your place for discussion about RPGs, gaming, music, movies, anime, computers, sports, and any other stuff we care to talk about... 

  • AI and skill is pointless

  • Because playing them is not enough, we have to bitch about them daily, too. We had a Gameplay forum, but it got replaced by GameFAQs.
Because playing them is not enough, we have to bitch about them daily, too. We had a Gameplay forum, but it got replaced by GameFAQs.
 #87276  by Don
 Wed May 18, 2005 7:33 pm
This mostly applies to MMORPG, but can be applied to most slow paced games as well (strategy, RPG, etc). I can remember as early as EQ where AI is supposed to be a big point offering better challenges. Skill can be traced to at least as early as the days of Starcraft where Protoss is billed as the race of skill (if you only had the skills, you'd be able to beat the Zerg). Clearly there's the implication that SC was a good game because as you master this game you'd be able to do what no newbie Zerg player can do, though of course this is hardly the beginning of a game trying to use skill as a selling point. The 50 different adaptation of the classic RPG systems, to me, suggests that developer obviously wanted to come up with a system where skill sort of matters (or they could just be really bored). Almost every MMORPG after EQ came out has made AI and 'your skill matters' as their selling point. So, companies clearly care about that, or at least care enough to advertise those aspects. But are they really important to your gaming experience? I think not.

First let's start with AI. Almost every RPG uses a completely random targetting mechanism. That is when the bosses uses his super powerful hit one target attack he pretty much picks one guy at random (though all the super powerful attacks tend to hit everyone so it doesn't even have to think about that). In a MMORPG, the enemies work off a deterministic hate list where each action has an associated hate value, and the target selection is solely determined by who is on top. This is not challenging, the players complained. We should be fighting smarter guys! Then, I propose the cleric-killer AI, writeable in a day certainly, that does the following:

1. Determine every person who can heal.
2. Sort heal effectiveness in descending order (guy who can heal the most at top) by class (i.e. all clerics are considered the same).
3. Kill a random person with highest healing effectiveness. (this is assumed to always be possible. If not the boss is not capable of winning against the players to begin with, and any AI is futile).
4. Repeat until no more peole can heal, then add all the people who can resurrect (in case some people can resurrect but cannot heal) and repeat 1-3. After that the AI is free to do whatever it wants.

This AI would certainly be challenging. In fact, it is almost certainly unbeatable. It doesn't really do any 'cheating' (any sentinent beings with intelligence ought to at least be able to figure out which classes have the best healing capabilities). Yet you don't see this AI anywhere in a MMORPG! Because it'd utterly spank any players. No, players want something like say something with the AI of a World of Warcraft NPC, that will do something that appears smart before ultimately getting killed. Or, to quote something from Furor, a famous EQ guildleader, "The enemies need to be so hard that only the best (my guild) can beat it." Except, of course, that Furor's guild wasn't the best guild in EQ so when EQ does throw such an encounter at Furor's guild, he started complaining how the game is broken.

I'll write more on this when I get home!
Last edited by Don on Thu May 19, 2005 12:35 am, edited 1 time in total.

 #87277  by Tortolia
 Wed May 18, 2005 7:40 pm
I remember reading an article in PC Gamer way, way back, where they were talking to the guys behind Mechwarrior 2. Basically, they said the same thing - writing an AI that can kick a player's ass every time is easy. Writing one that is a fair challenge is truly difficult.

(Incidentally, Furor's worked at Blizzard on WoW for about a year now.)

 #87278  by Don
 Wed May 18, 2005 8:45 pm
Well hiring Furor is most likely to just silence a guy who sort of has some influence in the MMORPG community. As far as I know he is a quest designer which means he gets to write bad Warcraft fanfiction and put them in a quest, which I assume is still a lot better than whatever he used to do for a living.

Continuing on AI, what AI turns into is basically just tricks that equates to just adding extra HP or damage to an enemy, but cleverly disguised. The problem is that it is not hard to break any special tricks into deterministic action. Regardless of how complicated an action is, you can boil the result to whether you can counter it, or if you can't counter it. If the price you pay for not able to counter is too severe, then you'll just end up learning how to counter it everytime (and if it can't be countered, it's not winnable and I recommend using the clerc-killer AI I listed and save the AI development costs). Or, most of the time, clever AI tricks just get treated as additional damage/HP and dealt with the traditional way, i.e. more damage, more healing, more tanking.

Now to skill. I think people are attracted to 'skills' or 'challenge' because they want to believe they're that special someone, that they're the only guild that can defeat that chosen encounter. It is interesting to know that such chosens have been repeatedly spanked by EQ in the face of a truly difficult encounter that demands the best of the best. Fire of Heaven, Afterlife, Conquest, and any other famous guild you might have heard of are not the first guild to beat the cutting edge EQ or WoW encounters, but they sure make a lot of noise. I think Fire of Heaven is still blaming the fact they played on Horde as to why they're not the best guild in WoW (which says a lot about the decision to choose Horde if they think it's at such a disadvantage in PvE, seeing FoH is a PvE guild).

Recently EQ has an encounter known as the Overlord Mata Muram encounter where it really does test the skill of every single member of your team. There are so many intricies that it'd take pages to explain just what each person must do. And guess what, the players get utterly spanked by this encounter because it demanded a guild where everyone has to be good to win. Of course players are quick to declare what they can't beat as broken. After all if I can't beat it, it can't be that I suck, it must be something else.

During the Gates of Discord expansion when a lot of big EQ guild quit due to 'incomplete content', they were not the ones who got to the impossible Tqiv twins in Uqua that some no name guild did. No these guys died to the first 3 trash in entrance and declared the zone broken. To be fair, pre tuned Uqua was about five times as hard as it should have been, but if everyone in your guild played perfectly you'd be able to get further than the first 3 guys at the entrance at least. You probably still can't beat the pre-tuned Tqiv twins, but at least you'd actually get to a point that is actually impossible.

My conclusion is that if anything requires skill, it's destined to fail, because if you're in a game with hundreds of thousands people playing, then chances are more than likely you or your guild is not the chosen one. If something is genuinely difficult, it should just wipe the floor with you so hard that you should never bother attempting it. Of course people don't like to be told they suck.

 #87283  by Flip
 Wed May 18, 2005 10:51 pm
Is it really skill, though, to just fight a guy repeatedly until you figure out the patterns and counters? Sounds like a lame SNES adventure game where if you just keep trying you will eventually find the path the developers intended you take to win...

The way i figure it, the computer games that require skill are mainly RTS's and FPS's, notice MMO isnt on the list. Granted, a 'good' FPS'er may just be exploiting a map, but when you get watsed by someone in open ground you know they are doing something you arent, not just finding a pattern.

RTS's also has the no skill 'good' players who exploit maps and items, but when you play someone who recons, prevents you from doing the same, builds the counter, traps you someplace on the map, then wipes your town, you know they have skill.

I dont think MMO's require skill and i think that aspect turns me off from them. You can be 'good', but 99% of the time it is due to something i just mentioned.

 #87284  by Don
 Thu May 19, 2005 12:30 am
RTS games are a lot about patterns on the macro level. Certain things just counters other things so well that no amount of fast clicking can make up for it. Anyway if you want something that requires a lot of quick reflexes, EQ has this raid call Trial of Foresight which basically turns it into a FPS game in terms of avoiding damage (all damage in this trial solely comes from your ability to move around the room at the right time to avoid projectiles). On the 'impossible' tuned version I managed to survive about 2 minutes with an entire raid's (54 people) worth of projectiles coming my way. On the normal version I can avoid the entire guild's worth of projectiles almost indefinitely by myself which means I can in theory beat what something that needs 54 people (well wouldn't possibly have enough firepower to kill the targets but you can just have everyone but the person with great reflexes hiding until the projectiles run out, then kill boss with no resistance whatsoever). But anyone who failed to do that won't admit they suck. They'll say it was lag or the projectile are bugged and become unavoidable. I don't know of any other person who have avoided an entire guild's worth of projectile on this trial so it's safe to say I'm really skilled at this part but does that really mean anything? I guess it's good to know you can be counted on where 40 others can't, but like I said if you raise the bar such that I can't avoid them then the community will just deem it as impossible, and it probably is impossible unless your guild has that few talented guys able to dodge a projectile every other second.

Eventually the Trial of Foresight worked out in a way if you can dodge at least 1 out of 3 projectiles you can win, and actually if you're lazy you can probably just ignore them completely and concentrate on the boss and rely on healing to cover their damage. You make an encounter that requires skill, either people will find a way so it does not (the eventual version of this trial) or it becomes impossible except for very few skilled (most likely not you), and the latter is no way to make a successful MMORPG because people don't like to be told they can't win because they suck.

 #87307  by SineSwiper
 Thu May 19, 2005 8:51 am
Actually, certain AIs on fighting games are pretty advanced. VF still has that learning ability into the AI, where it can learn some of the advanced combos/juggles that you do and do it to you. (Kage learns the Ten-Foot Toss of Doom, and rings you out, for example.)

Depending on the complexities, it would be easy to adjust the level of AIs, though it's also a matter of how intelligent it would look. For example, on SF3, it would be pretty easy to make an impossible AI that parries everything, or something that parries about 25% of the time. Same with blocking, and punishment of blocked/whiffed moves. However, having an AI that just sits there and takes damage looks dumb, and doesn't even look human. (At least the mall scrubs would be hitting buttons and blocking most of the stuff when he stands still.)

Obviously, there are still plenty of fighting AIs that rely on bad patterns, mainly because they are hard-coded into their engine. The more the AI engine is based around past experiences, the better it gets. I was just noticing something the other day on that. I was training in SF3, and I turned on the computer mode for my dummy. (IOW, an infinately long match.) At first, I was doing pretty good, but I think the AI was starting to learn my patterns and eventually started kicking my ass. Funny that this was with Alex, who is known for some of those bad patterns. It still had those patterns, but it was more aggressive, and favored better stuff to try on me.

 #87324  by Don
 Thu May 19, 2005 10:24 am
Didn't they have this computer learn your moves things since Street Fighter 2? Or at least peopel said it did anyway. I'm not sure what's the reason of a computer AI learning anything when it already knows how to do any combo in the game.

 #87343  by kent
 Thu May 19, 2005 11:39 am
i'm fully immersed in FFXI endgame so i have quite a bit of experience in this area.

MMORPG skill is not the same kind of skill as you'd apply to other things.

MMORPG skill has more to do with knowledge and the application of knowledge in a timely matter than other kinds of skill. skill is also placed in being able to communicate and work with other people.

to say an MMORPG doesn't require skill is ridiculous, if it requires no skill at all, than any group given the right number of players with proper jobs can overcome any encounter and quite simply that is not the case.

other "guilds" have wiped to certain monsters with twice as many players as what it takes my "guild" to kill them. the only way to explain that is some "guilds" are more skilled than others.

being able to formulate a strategy is a skill. being able to pass on that information to other players is a skill. and simply being able to follow orders given to you by someone else is a skill.
Don Wang wrote:1. Determine every person who can heal.
2. Sort heal effectiveness in descending order (guy who can heal the most at top) by class (i.e. all clerics are considered the same).
3. Kill a random person with highest healing effectiveness. (this is assumed to always be possible. If not the boss is not capable of winning against the players to begin with, and any AI is futile).
4. Repeat until no more peole can heal, then add all the people who can resurrect (in case some people can resurrect but cannot heal) and repeat 1-3. After that the AI is free to do whatever it wants.

Don Wang wrote:It doesn't really do any 'cheating' (any sentinent beings with intelligence ought to at least be able to figure out which classes have the best healing capabilities).
but it does cheat, players don't have their jobs stamped on their foreheads. so for "any sentinent being" to automatically figure out what job each class is in itself is a cheat.

if enemies were allowed to use this AI, than players need to be given abilities to hide their jobs.

how does an opponent recognize who the healers are? by looking at them? if so, than players need to have the ability to equip everyone to look the same way.

if the logic behind it is simply that players eminate a magic aura that monsters can detect and players have no spells or abilities to hide/dimish this aura, than you might as well make a game where all the enemies have rocket launchers and give players slingshots.

i don't think there's anything truly wrong with the hate system, having most hate means the monster considers that player is the biggest threat, the AI to attack and kill the player that poses the biggest threat is logical.

"the cleric-killer AI" assumes that healing is the be all end all player ability. it assumes that a monster can ignore all variables and charge directly towards healers even though they are being stabbed by spears, thunder bolts are being thrown at them.

for there to be some kind of competition, monsters can not be immune to pain or other annoyances, when they are being stabbed it has to react in some way.

Hit Points in itself is already a flawed system, a monster or player at 20% HP should not be functioning with similar effectiveness as it did at 100% HP, in most cases monsters get stronger at lower HP which doesn't always make sense as it is.

part of the hate AI has to take into consideration, that monsters don't like taking damage and that actions that cause them damage is perceived as a threat.
Tortolia wrote:writing an AI that can kick a player's ass every time is easy. Writing one that is a fair challenge is truly difficult.


there really isn't a better summary than that imo.

 #87346  by Kupek
 Thu May 19, 2005 11:57 am
Don Wang wrote:Didn't they have this computer learn your moves things since Street Fighter 2? Or at least peopel said it did anyway. I'm not sure what's the reason of a computer AI learning anything when it already knows how to do any combo in the game.
For the reason Tort gave. If the goal of the opponent is to kick your ass, then yeah, they don't need a sophisticated AI to do it. But if the goal is to offer you a challenge and force you to learn new things, then a learning AI makes sense.

 #87347  by Don
 Thu May 19, 2005 12:08 pm
There is no plausible explanation why any game can't employ the cleric-killer AI.

All the stuff you mentioned would matter if somehow the enemy's raw power is weaker than the collected sum of your party. But it is not. In invariably all cases the enemy's raw power is most certainly greater than that of your party combined which is why there is supposed to be a challenge in overcoming the enemy. Last time I checked it is usually universally agreed that caster classes don't wear plate mail unless you're a cleric. Even in that case, you can discern the difference by the fact that the cleric is most likely not the person running up to melee range swinging a mace. If there is supposed to be a penalty for ignoring others and make a run at the healing classes, then there is an even bigger penalty to make a run at only the tanking classes. Whatever you're doing you can still do if the enemy is pounding on a tanking class except he's also much harder to kill. There is absolutely no cheating involved. As a player you certainly know that you should take out the weaker support classes first. Your enemies are generally stuff either with way more raw power or way more intelligent, usually both.

As for having skill from one guild or another, that is such meaningless statement without knowing the composition or the gear status of your guild. I'll say this again: everyone who thought they had skills as a guild in EQ ultimately turned out to suck when the game throws them a truly difficult encounter. If you think your guild is special, chances are very good you're wrong.
Last edited by Don on Thu May 19, 2005 12:11 pm, edited 1 time in total.

 #87348  by Don
 Thu May 19, 2005 12:10 pm
Kupek wrote: For the reason Tort gave. If the goal of the opponent is to kick your ass, then yeah, they don't need a sophisticated AI to do it. But if the goal is to offer you a challenge and force you to learn new things, then a learning AI makes sense.
Doesn't exactly take a complicated AI to record down the kind of combos you used and just start blocking them 100% if it sees them used more than X times in a period of Y. Actually that's not AI at all.

 #87349  by Kupek
 Thu May 19, 2005 12:30 pm
Don Wang wrote:Doesn't exactly take a complicated AI to record down the kind of combos you used and just start blocking them 100% if it sees them used more than X times in a period of Y. Actually that's not AI at all.
One, doing anything 100% isn't a realistic opponent, and two, it's still using past behavior to predict future behavior.

 #87352  by Don
 Thu May 19, 2005 12:55 pm
It's not really predicting because the computer is more than capable of blocking/countering anything as it sees them without any past history due to its reaction time. It's more like if I haven't seen this attack used in the last X period of time then I will pretend I don't know how to block/counter it.

 #87383  by SineSwiper
 Fri May 20, 2005 2:04 am
Honestly, I think the hate-based system is stupid because it's so easy to control and predict. Consider the problem with pets fighting in PvP. Obviously, the pet is not going to "aggro" the human player, so all pet classes lose in PvP because they will attack the player and the player will be weak against a direct attack. That sort of difference between players and enemy is the reason why the AI doesn't act like a real "Artifical Inteliigence".

MMOs are not about skill at all. Maybe that's why I don't like them. Eventually, looking at the pretty cities and pressing the same keys over and over again gets fucking boring.

 #87394  by Don
 Fri May 20, 2005 10:31 am
At one point in EQ stuff was programmed to automatically ignore a pet but then the classes with pets said it wasn't fair, so now pets get ignored at melee range but not from range. If you've any kind of random targetting it either becomes too hard (randomly target a person who have no chance to survive) or too easy (if even your most vulnerable classes can survive from being targetted).

The enemies in a MMORPG can't ever be very smart because the cleric-killer AI will win against the players every single time unless it's a trivial encounter and it's very minimal intelligence that requires no cheating whatsoever, unlike say a RTS's computer that already has the equivalent of map hack. At any rate players aren't exactly any more interesting to fight than NPCs unless it's a very big battle.

 #87415  by kent
 Fri May 20, 2005 3:29 pm
I'll say this again: everyone who thought they had skills as a guild in EQ ultimately turned out to suck when the game throws them a truly difficult encounter. If you think your guild is special, chances are very good you're wrong.
i never claimed my guild to be "special". we just happen to be superior to the other NA guilds on my server. we know that we are actually inferior to a few of the Japanese guilds for various reasons which include as you mentioned "gear".

and there are "guilds" on other servers that are superior as well. there are still 3 enemies in FFXI that my guild has not been able to defeat that others have.
My conclusion is that if anything requires skill, it's destined to fail, because if you're in a game with hundreds of thousands people playing, then chances are more than likely you or your guild is not the chosen one.
your conclusion is broken, plain and simple. for something to require skill doesn't imply the necessity for perfection.

a skilled basketball team still misses shots. by your definition any team that misses any shots has no skill.

skill is in masking the deficiency of the players so that you can still accomplish your goal.

if skill is truly pointless, than explain what makes the difference between some groups being able to kill something in 1/10th the time as other groups with less players.

here's a tip, gear/equipment is not the end all be all in all MMOs.

also, depending on implementation, your cleric-killer AI seems very very easy to kill enemies by simply kiting them with clerics.

 #87420  by the Gray
 Fri May 20, 2005 4:32 pm
Too bad you won't give GuildWars a chance Don. I've found that every single time we take Monks (healers) in our group, they are targeted by the biggest damage dealers first. It has really forced me to changed MY game, as I now have to have 2 or 3 Hex breakers on me so they will live long enough to heal themselves. Then I learned that as soon as I broke their hexes, they started turning on me! Not fun, as my Mesmer is pretty weak.

Now the game's AI isn't perfect, but it's pretty damn good. Coupled with the fact that you can't level up to infinity in order to get better, it keeps me interested.

To keep going you have to get better at playing the game. What an original idea...

 #87427  by Don
 Fri May 20, 2005 6:11 pm
Obviously the enemy has to be faster than you can run or have some way of preventing you from running, otherwise there is no AI that can possibly win under any circumstance if whoever is targetted can avoid damage by just running away.

Equipment determines what you can do in a MMORPG. If it does not it's usually the sign of a poorly tuned encounter with a gimmick and that encounter gets solved by everyone immediately as soon as the gimmick is figured out.

There is one encounter in EQ that was viewed to be next to impossible because there are these guys that mysteriously hits 3 times+ harder with infinite melee range (which is to say, enough to kill anyone instantly) whenever they're not hitting a Warrior, Paladin, or Shadowknight and the number of those guys is determined by the number of aforementioned classes you bring. 'Skilled' guild which usually represent people who have inside info figured out to bring a total of 3 War/SK/Pal combined which gives you the minimum number of super hitting guys (6) and beat it no problem (the only challenge in the fight is dealing with the special guys), while less skilled guys tried to compensate the damage by bringing more people who can tank, which actually made it even harder! My guild does that encounter with 54 because the leadership doesn't know this so we bring a lot of tanks to deal with the super hitting guys, which in terms gets us the maximum number possible (24) as opposed to 6 strong guys and 18 weak guys. Lesser guilds have beaten it with 30 fighting the 6 strong 18 weak guys version.

Every instance of people claiming to do with something less there is some inside info or some outside factor they did not tell you about. If you do it 1/10th the effort/people the most probable explanation is that you have 10 times the gear to make up for having 1/10th the people. The very best 6 players in my guild is certainly 10 times more powerful than the average group. That'd also mean that suppose the average 30 people can beat encounter X, you can replace those 30 guys with 6 of our best people, and it'd take less skill to beat the same encounter because these 6 guys actually have the raw capabilities of 60 average players.
Last edited by Don on Fri May 20, 2005 6:23 pm, edited 1 time in total.

 #87428  by Don
 Fri May 20, 2005 6:23 pm
I watched my brother playing Guild Wars and it looks like the moment anyone who is not a tank does anything, everything just makes a beeline for that person. This isn't necessarily 'smart' but it certainly makes the game more challenging. Generally speaking the less time commitment in terms of preparation a game has, the more it can get away with doing totally arbitarily 'cheesy' things. If it takes a total of 100 man-hours to get to a boss you probably won't think it's very fair to get spanked by the cleric-killer AI. But if it's just 5 guys @ 30 minutes for 2.5 man-hours, it might be okay.

 #87429  by the Gray
 Fri May 20, 2005 7:02 pm
Don Wang wrote: If it takes a total of 100 man-hours to get to a boss you probably won't think it's very fair to get spanked by the cleric-killer AI. But if it's just 5 guys @ 30 minutes for 2.5 man-hours, it might be okay.
That's pretty much been my experience with GW so far. It doesn't take forever to get to a boss during a mission, so if you get spanked you aren't that pissed off about a major time investment.

Funny thing about the Tanks, ie Warrior/Monks. In PvP matches, I just absolutely demolish them with my Mesmer/Necromancer. Not even fair. Then again, I hate those frickin Rangers. Monks I avoid as well. Stalemate there.

 #87437  by Derithian
 Sat May 21, 2005 5:34 am
I love my ranger warrior. I do great damage with a bow and every attack I whip out is specifically designed for pure damage dealing as they all cause bleeding (A slow loss of life over a scertain number of seconds. I may not do a ton of damage every hit but the 20 seconds of bleeding sure as hell does......that and I can throw my douchebag at people to really mess them up......the funny thing is with the warrior sub when people try to get in close to attack me I can switch to my sword and actually do more damage over time while taking less myself. pretty good combo if you don't mind the inability to heal yourself and very good for pvp with all the bleeding....

 #87460  by the Gray
 Sat May 21, 2005 11:43 pm
You're a Ranger! Why the Hell don't you have Troll Ungent equiped!

That skill is awesome.