Tobold's Blog
Friday, May 19, 2006
Stop Playing Memory

Yesterday's post on the repetitiveness of video games made me think a bit more on the subject. I think there are two major game components in video games, but they aren't really easy to keep apart. Let's do a thought experiment, by designing 2 very simple games:

Game 1 I call the reaction game. You and me sit face to face, and we both have a red card in the right hand, and a green card in the left hand. Now I raise one of my hands at random, and your job is to raise your hand holding the same color. We do that 20 times and your "score" is based on your average reaction time, the faster you react, the higher the score. If ever you raise the wrong hand with the wrong color, you're out, and have to restart.

Game 2 works exactly like game 1. Only that instead of raising my hands at random, I always use exactly the same sequence of red and green. I call this the memory game.

It is obvious that your score in the memory game will be equal or better than that of the reaction game. Because if you weren't using your memory at all, you could still simply react to what you see me doing, and the game becomes a reaction game. But the more often you play game 2, the more your score is going to improve over the reaction game score, because you *know*, from your memory, which color card I am going to raise next. Of course playing memory is more repetitive than playing the reaction game, and while your score is higher, the game is less interesting.

Video games often have both elements. The first time you encounter a situation in a game, you will need to react to it. But if you fail and die / wipe, you can reload / rez and do exactly the same situation again. Thus you know what is going to happen, and your memory helps you to overcome the encounter and win. That is true for many games, from first-person-shooter console games to World of Warcraft.

Take for example the Onyxia encounter in World of Warcraft. There are guides for this encounter out there, telling you exactly about the three phases of the combat, what Onyxia will be doing, and how to avoid getting killed in each of this phases. An Onyxia raid is successful if all the participants know this information, and play in accordance with a prearranged plan, based on *knowing* what will happen.

Now imagine what would happen if Onyxia wouldn't react always the same way, but would have some sort of artificial intelligence. You stand there with your fire resistance gear, having quaffed a greater fire protection potion, and Onyxia breathes acid on you, dealing nature damage. Or instead of using deep breath in the second phase, Onyxia sees that everybody is widely scattered, and that the most annoying characters are those cloth wearing, dot casting types, and does a short dive attack, ripping that cloth wearer into pieces.

Why are people so proud to be the first guild to have killed a new raid boss? Because as they were the first, they didn't have the knowledge on the first try, painstakingly acquired the knowledge over several wipes, and then played and won the memory game. The guilds coming after that can worked with borrowed memories, making the encounter a lot easier.

Now the memory game has one distinctive disadvantage. Everybody, including the game developers, knows that the memory makes the encounter a lot easier. So the difficulty level of the encounter is planned in a way that for a person or group *with perfect knowledge of what will happen* the encounter is still challenging. Which usually means that if the player or group does not have the knowledge, but relies solely on reacting to observations during the encounter, the encounter is impossibly hard.

What I would like to see, and what I think could be the future of video games, is better artificial intelligence, and encounters that are different every time you play them. These encounters would be easier, because memory wouldn't help you much with them. Instead of relying on people knowing what will happen, which is by necessity repetitive and becomes boring fast, in each encounter players would need to react correctly to what they see. Ideally the monster would likewise react on what it sees the player(s) doing, so doing always the same strategy would actually make the encounter harder, not easier.

Think of it: Every single guild killing Onyxia would be the first guild to beat that particular encounter, because every Onyxia encounter would be different. Instead of players learning how to beat Onyxia, Onyxia would have a neural network type of artifical intelligence, learning how to beat players that always do the same. The encounter would remain challenging and interesting forever, because it never repeats.
movies are made since late 19th century.
book have made since era's.
games have been made since 1970's.

what's the common goal they have: to entertain people.
what's the common problem they have: they are repetitive.

if you say that only mmo's are repetitive...just read
a thriller, a fantasy/sf book, watch a movie.
they all have a common plot with once in a while a funny, interesting angle, but it's written in solid-rock. nothing new under the sun there.
i think the best comparison with mmo's is the have a subscription on a library or a dvd-store. u go in, rent a book/movie, read/watch it a couple of times and then return it.
i know we have the technology that is called AI, but do the mass want that? i doubt it, why in the hell made Titanic such a profit then? not a very "interesting and mysterious" plot ;)

i love difficult, weird strange movies, but heck, i enjoy myself watching Pirates of the Carribean. fighting against the system and or have a dream about mmo's is very good and ok, but don't forget to let yourself go once in a while and enjoy the entertainment plain and simple.
hehe, onyxia uses AI. She realizes the priests won't let her kill those damn warriors. Onyxia ignores the warriors, kills all the priests, and oops, WoW's pve game is irrevokably ruined.

What you suggest is awesome, but your not suggesting smarter monsters, your suggesting a total revolution of the entire fantasy pve MMO genre.

better AI could only imho accompany a game that had overall more realism. Why can't we break onyxia's wings to prevent her from flying? Why can we bring in several large shields to form a barrier against her fire attack? If she did take a swooping dive against a cloth wearer and they blinked out of range, would she hit the floor and break her neck? Could our archers focus fire on her wings when she flies? Could we have our mightiest warrior cut off her tail to prevent her from tail slapping us around the room?

Your idea is great, but I wonder if such a realistic game could be made?

Personally as a fairly hardcore raider, I think the scripted encounters in wow, while obviously part of the memory game, are exciting, because they add so much variety to each fight. Games could go along way, but will they? Only time will tell.
Even in a relatively scripted encounter like Onyxia, there are still random elements that keep it from being an entirely repetitive, memory driven event. A few priests don't quite escape a deep breath, a tail swipe sends someone into the whelp cave, the MT can't pick up aggro at the start of phase 3, or you are playing for the first time as your mage alt, rather than your rogue. All of these require improvisation and add excitement.

In my guild, most Ony raids either include enough players who have just gotten keyed and really don't know what to do (which can be too interesting), or are over in 30 minutes and its off to practice in BWL.
Instead of players learning how to beat Onyxia, Onyxia would have a neural network type of artifical intelligence, learning how to beat players that always do the same.

Hmm sounds to Terminator-ish
"The Skynet(Onyxia) funding bill is passed. The system goes on-line August 4th, 1997. Human decisions are removed from strategic defense. Skynet(Onyxia) begins to learn, at a geometric rate. It becomes self-aware at 2:14 a.m. eastern time, August 29. In a panic, they try to pull the plug."


You sound like an awesome person to play Dungeons and Dragons with, what an imagination :)


I think the Developers are thinking about this one, I mean in the Nefarion battle and Chromaggus fight in Blackwing Lair, they have different color breaths and different color Drakonids that are random when the instance is created. But to have an Onyxia you describe would be too much too soon.

But something really random would be interesting, but all boss mobs can't be that way.

Ooh- well, the standard reply would be "O god not learning AIs!" And this is entirely correct.

The problem is, while having Onyxia try out new strategies and use what works in these fights sounds like a good idea, in practice it comes out terrible, for a few reasons.

1. The early stages of learning AIs are utterly sad. Stand in a corner and drool sad. In order to get a decent result, you'd need to run several iterations before she'd even be any threat, culling the failures.

2. The midpoint could be okay, but can be gamed pretty easily- spot a slightly dumb Ony a few wipes, and she'll have gone stupid again. You'd have to put in almost constant monitoring of encounters and cut out any versions that looked to be slipping.

3. The endpoint will be utterly unassailable. Imagine deep breaths targeted on your tank's healers every time, or getting herded into the cave deliberately by tail swipes. For another encounter, imagine Kazaak going into his time limit mode the instant you aggro him.
In MMORPGS PVP introduces the intelligence of real humans - but even there successful players rely as much on lot on good preparation (gear and teasmspeak) as well as memorised tactical knowledge.

In my opinion first person shooters on the PC are more promising route for non formulaic gaming. The WASD and mouse control system but does require skilll to master but once mastered it allows a level of fluid freeform control that is unmatched in gaming. Certainly a lot better than the auto-aiming substitutes used on consoles. A multiplayer PC FPS match against human opponents is rarely formulaic and is a genuine test of skill. Some recent games are also pushing the boundaries in terms of machine AI. One notable example is the case of npcs in FEAR who will navigate through a building to sneak up behind you.
It always bothered me that tanking works...

Imagine if healers had to kite to stay alive :)

Tobold, I think you're describing one possible evolutionary path for gaming. We're about reaching the practical limits (certainly interest-wise if not theory-wise) of current technology and programming - it's getting time for the next jump. It could be to smarter enemies, more advanced AI. I think that would be cool to experiment with and see what kind of fun encounters could be designed.

As an above-poster notes, there is an intermediate version - randomization. Chromaggus is a good example - where he has 2 random breaths/attacks out of a possible 4-5. Granted, with such a limited number there really isn't all that much advancement as the reaction becomes "Okay, he has no. 1 and 4 this time." Still, it's a step in the right direction. Imagine a WoW boss who always spawns as a tank with a huge flaming sword and glowing shield BUT then chooses 2 special melee abilities and 3 special spell abilities where the special abilities correlate to in-game class abilities. So he could spawn with Cleave & Seal of Command one day and Mortal Strike & Gouge the next. That sort of thing. Open the potential mob abilities up large enough and that will provide more reaction-based encounters.

There are other examples of this in WoW. Some BWL trash have random color/ability assignments (out of a possible 5 choices I believe). Same BWL trash has a specific magic vulnerability too, to one of 5 magical types (ice, fire, shadow, nature, arcane - just like Chromaggus shifting his spell vulnerability). They also are highly resistant to the non-vuln types. Very weak examples with so few choices but again, baby steps.
I think when we reach the true Virtual Reality era, when we log into a game via our temple data jacks, then we'll see "games" where the bosses, and even the monsters, react randomly.

"You walk down the hall and hear the guard approaching, but you knew he'd be coming because this is the fifth time you've had to replay this level.

The guard passes your door and you give him a few seconds before slipping out to backstab him. Too late your brain recognizes the odor detected by your nostrils.

The guard is leaning up against the wall smoking a just lit cigarette.

WTF? He didn't do that the first four times!

You already have your dagger out in preparation for the backstab so you lunge at him and he backs away, but not quite fast enough. Your blade enters his throat, but you're also too slow and he lets out a brief but loud yell before he dies; "Intrudaaahhhhh!!!!"

Running footsteps from behind you warn you of the fast approaching guards and you dive back into the room. You know it's but seconds until the guards arrive and find their dead mate.

Now what do you do?"
There's a limit to how varied you can make an encounter before the memorization process fails. There comes a point when there's too much to remember (e.g. "which response do I choose for pink polka-dotted mobs?" when there are 299 different color variations). Memory failure equals encounter failure which results in player frustration. Sure, the raid leader could use a second computer and database as an aid to memory for this 234th variation of 299 boss types. Type in "pink polka-dotted [boss name]", spend ten minutes studying, and get:

"Uh ... Priests. The pink polka-dotted version will attack you on sight unless you behave like a faerie. No Jonbon, NOT a "fairy". That was for for the maroon green-striped version of this boss. We're talking about the little winged smiling pixie kind - Tinkerbells - for this boss color. So once in sight of him you all must, in EXACT unison or he wipes the raid, curtsey in his direction, giggle, slowly twirl 360 degrees ten times, then sit down and don't do a thing until everyone else is done with their opening moves. KEEP SMILING! Do this right and the boss will ignore you for the rest of the fight no matter how much you heal. You have the easiest part in this boss color, Priests, so -50 DKP if you screw it up!

"Now Mages, your part is harder. The pink polka-dotted boss likes frogs. You need to hop up and down and type "ribbit!". A typo will wipe the raid, so put it in a macro! So you hop and "ribbit!" together while in in a conga line, tallest race first, smallest last -- that means ALL humans in front and ALL gnomes at the end for the morons in this raid -- and alternate male and female. The line needs to thread in a certain pattern around the boss that, seen from above, looks like the hand-written word "FUROR". You MUST complete the word at the same time the Priests are sitting down. This will mesmerize the boss and bind him in place."

"Rogues -- No, DarkShadowStabStab, I have no idea how someone figured these out. Many Bothans died, no doubt. Listen up rogues! ..."

Adding more and more scripted variations of behavior to mobs, randomly or not, doesn't lessen the need to memorize. It increases it. It's still the memory game.


What's a non-memory game? Tobold covers it pretty well. I'll summarize and paraphrase the things I'd like to see. I'm sticking with the traditional fantasy RPG kind of game so we're talking apples and apples.

* No Groundhog Day. You never repeat an encounter (and never grind XP/money). You might fight the same "boss" again if you failed to kill him the first time, like if he ran away before death, but other than the boss himself the encounter will be new. It could very well be in a different location with different minions, traps, etc. At the very least, the boss will remember your last attempt on his life just as well as you do and (re)act differently because of that.

* Encounters can be successfully done on the first attempt. There's never a second attempt, so this is a must. They should provide a challenge scaled to the particular group of characters, but never be impossible. A tight group of very experienced players with perfect class makeup will be challenged as much as a newbie group and/or a group with a non-optimal mix of classes.

* Many ways to "win" an encounter, including cleverly bypassing it entirely if it's not important to the goal. Ideally, the NPC(s) will comprehend any creative on-the-fly attempt to get through or around an encounter, weigh the validity of the attempt, then allow or deny it with the goal being increasing the fun of the players, not necessarily beating them. Fast talking, reasoning, and bribery should be among the possibilities if such things fit with the type of NPC creature (i.e. you might reason with a mad wizard, but not with an enraged slime monster).

... and more, but that's enough to show that computer AI and the content creation ability of development studies aren't up to it. But what if you replace computer AI with a human being? And what if you let anyone who desires, the human in control or others, to create new content?

I'm sure some have guessed where I'm going with this. It's PnP Dungeons and Dragons with a live, human Dungeon Master. "Wait! We're talking about computer games." So am I. There's already a computer game wherein you can have everything I described: Neverwinter Nights with a live, human DM using the DM client. There's even a player/DM matching service at where anyone can sign up and join such a game.

But that's just one computer game and it's six years old. Sure, NWN 2 is coming out, but that's an update/replacement rather than a second game on the market. Why, if people really would like this kind of non-memory gameplay, aren't there more? That's too big a question to be dealt with in this (already overlong) post. ;)
Follow up. I wanted to wait until after experiencing a DM-run game to complete my thoughts.

I played a NWN multiplayer game with four other players and a DM yesterday. The game required level 5 characters. The DM simply gave the appropriate XP and gold to our new characters and provided a store. From the start I was already experiencing a huge departure from any other multiplay options out there. No need to "earn" our levels because levels were merely a facilitator for the real gameplay.

In brief, the session confirmed what I wrote in my previous comment. It was what's called a "one-shot", a single-session game with beginning, middle, and end contained within ~3 hours of play. As such, it wasn't as deep as (I've heard) the multi-session campaigns can get. Nor was it as open-ended. Even so, there was no Groundhog Day gameplay -- no repeats of anything at all, and there were many different options for successful progression towards the end. For an example, we encountered zombies that we couldn't hurt (no magic weapons). But unlike in the movies where the zombie slowly shambles after the running girl, and every time she looks back it's right behind her, we could and did simply run away from the zombies towards the goal. Left them in the dust and the DM didn't ban us for bypassing content! ;)

Okay, so it's good stuff and it's a kind of online multiplayer gameplay that can't be found anywhere else. Totally unique. Why hasn't some other developer included the option in their multiplayer game? Because it's niche. More niche than WoW raiding, if you can believe that. But unlike WoW's raiding content, NWN multiplay with a DM isn't a prestige activity that can be promoted as incentive to buy NWN. So few play DMed session-based NWN games that Obsidian, the makers of NWN2, didn't even bother to mention that audience when deciding whether or not to release the DM client as a post-release patch for NWN2. They centered their comments on PW (Persistent World - NWN's version of a MUD) players instead because, although they're very niche too, they're about ten times less niche.

Why it's so niche is a question best answered by going to and reading the general forum for topics related to declining players. In summary: this kind of game requires a ton of effort on the part of the DMs, leading to a limited supply of games, leading to players not finding games to play and leaving, leading to DMs not finding players for their games and reducing the introductory-level games scheduled, leading to more players not finding games in their available time slots ... a vicious cycle. I question whether it would have gotten off the ground at all if there hadn't been an immense glut of players in the beginning of the cycle. With NWN, the cycle may begin anew with an influx of new players.

In conclusion, even though there's a non-memory game already out there, and it's tons of fun, such games won't become popular until the DM can be replaced by computer AI. There just aren't enough humans willing and able to fill the role. And human DMs can't do their thing on demand, whenever a few players want to play a game.
Edit: should be "With NWN2, the cycle may begin anew with an influx of new players."
Post a Comment

<< Home
Newer›  ‹Older

  Powered by Blogger   Free Page Rank Tool