I've had the opportunity to fool around with some XNA Studio 2.0 stuff at work recently -- mostly using it as a quick 2D prototyping engine for showing people some system dynamic or another in motion.  I have to say, it is pretty slick.  Combined with the joyousness of C# programming, I've been quite pleasantly surprised how quickly I can turn something around from a discussion of a concept to a running app illustrating it.  Of course some of the cool things it can do (like all the Xbox 360 stuff and the potentially-disruptive distribution model announced at GDC) are not really relevant, but they do add to the package for most people.

When I have a little spare time I plan on making a little weekend project of getting a game up and running at home and on my 360, so I can start to talk to my kids about what programming is.  Heck, maybe I'll make a little educational game out of while I'm at it...

Another highlight of the show for me.  Both the speakers were funny and charming, much in the way Portal itself was...

 

---

 

Integrating Narrative and Design:  A Portal Post-Mortem

Kim Swift & Erik Wolpaw, Valve

 

Warning, Spoilers!!!

 

Why should you care about Portal?  Small team, no more than 10 people at once.  Been a commercial and critical success for Valve.  After all is said and done, no regrets.  Unlike the guy with the portal tattoos…

 

By itself, story wouldn’t make much of a novel.  Gameplay on its own would be dry.  Tight integration of story and gameplay resonated.  Team size imposed constraints on our design choices.  Creatively sidestep constraints – Glados as disembodies voice might not have occurred to them if they had the resources to do otherwise.

 

Games tell two stories:  story story, gameplay story.  Lowering the delta between the two will make your story more satisfying.  Games with a high story delta:  Clive Barker’s Undying.  Good shooter, talented writer who cares about story.  Shooting lots of monsters but the game stops you to play one of these cutscenes where you calmly interrogate a member of the staff.  At no point do you freak out and tell them about the monsters.  The two stories make no sense when you smush them together.  Good game being undermined by a disconnect between story story and gameplay story.

 

“story” story must never intrude on the gameplay story.  Less is more.  Ruthless about trimming the narrative fat.  Lean storytelling machine.

 

Process:

 

Playtesting – really important!  To both narrative and gameplay.  Most important thing ever done on Portal.  Watch people play your game.  Find out what your players actually want, adjust gameplay to what players look like they need.  Adjust story to enhance what players are already feeling.  Playtest keeps you objective – expose what isn’t working.  If a player can’t recall the story, it isn’t working.  Fix was almost always to cut more exposition.  Playtest early and often!  First room iterated on from the very first week.  Forcefield that playtesters just didn’t know what it was, traced the edge of the cage and got confused.  Helped us refine our art direction – a lot cleaner and simpler.

 

Advice:  writing a funny game?  God help you.  Tough guy dialog is endlessly macho.  Funny dialog is funny once.  Maybe.  Trust your instincts, don’t despair.  Playtest!  Remember initial reactions.  More playtesting = more hearing people react to your dialog for the very first time.

 

Embed exposition in the environment.  Unless it is in emails or voice recorders.  Reading email = not actually that fun.  Be creative!  Easy to say… Apply a rule set, and be ruthless about the rules.  Choice between embed or cut, makes you really creative!  In portal == wall scribbling behind the scenes.  Put yourself in the position of your characters.  “The Ratman”, a fellow escaped test subject.  Originally wanted you to meet him but had two artists, so not feasible…

 

Evolve narrative out of gameplay.  Write to enhance what playtesters are feeling.  Keep the story “wet”.  Don’t get too attached to anything.  Could disappear at any time.  Don’t get too attached, you might need to cut it and that is probably for the best! 

 

Weighted Companion Cube.  Box Marathon level – long level with the box, in the end put the box on a button.  Take one:  moving list obstacle over a good pit.  Players would destroy the box and had to go back, frustrating and annoying.  Back to the drawing board – remove the lifts and goo pits, added more where having the box was necessary to solve the puzzle.  Cube as stairs, cube as weight.  Take Two:  Gameplay events so you can always see the button.  Still needed something else.  Erik to the rescue!  Hint using the environment – lighting, geometry.  When all else fails, great dialogue is an excellent hint.  Was reading declassified govt manuals, isolation causes association with inanimate objects.  Maybe if Glados needles you a little bit and it worked.  Sometimes goofy ideas tend out to be really good ones.  Incineration station = boss battle training.  Perfect training location, more satisfying level ending.  Players learn better when not stressed (no ticking clock).  Forcing you to euthanize it was great!  Revenge, incinerate Glados the way she makes you incinerate your best friend.  Gameplay > story > gameplay.

 

Sometimes gameplay isn’t enough.  Original ending of portal.  What does a boss battle look like?  Complex puzzle?  Results:  pain.  Attempt one: James Bond lasers that followed the player around.  Lasers = bad.  Boring to dodge, difficult to aim.  Hard to tell if you’re hit.  Abandoned in favor of rockets.  Attempt two:  Portal Kombat.  Lots of testers were hardcore shooter gamers.  Didn’t work out very well.  High intensity = bad.  No one paid attention to Glados, and alienated players who like the slower-paced cerebral nature of Portal.  Attempt three: chase sequence.  Pacing was horrible, players didn’t see where she would go and then only slowly realize how to follow her.  Chase = action packed, instead players were confused.  Failed in many ways.  Complex boss battle = nope.  More complexity = slower pacing, not a good match for the climactic sequence.  So what now?  We’re screwed.  Playtesting to the rescue!  One thing worked well for players – the fire pit. 

 

What made it climactic?  Time pressure.  Makes people think something is more complicated than it is.  Good visual impact – pit of fire!  High drama, for the first time Glados is openly trying to kill you and you first assert control over your environment to escape.  Those things combined made an easy puzzle like a much more complicated puzzle.  Holding on to this idea of a complex puzzle at the end, simply wasn’t true.  Applied all of these things to the time pressure.  Just had a countdown!  Neurotoxin = green particles cheap!  Drama meant we had to write six minutes of decent dialog, previous ones required infinity worth of dialog = sad.  Easy puzzle important.  Wanted people to see / hear the end!  Wanted players to be genuinely happy and leave with a smile on their face.

 

What makes people happy?  Catchy song…  Ultimately a lot of this all came out of constraints.  Couldn’t make a big FMV.  Cost to happiness ratio of a great song was really high!   

 

Embrace your constraints!  Have faith in yourself and in the skills of your team.  Playtest, playtest, playtest! 

 

Q:  Why are audio recordings / emails a bad choice?  Just from playtesting?  None of these are prescriptive.  Just a personal choice they made.  Don’t like to read in games that much.   How often do you use a voice recorder to leave secret messages for your friends and family? 

 

Q: Why does Glados have an incinterator in her room?  Kind of fun… backstory that they weren’t sure if she was going to go out of control or not.  So made this room so that they could incinerate her and get rid of her.  Red phone in there, supposed to be his job to call somebody if the AI goes rogue… didn’t work too well.

 

Q:  Playtesting critical, not outsourcing it.  How do you go about getting new players through that process?  Easy to go to the nearest Gamestop “we’ve got this game, wanna play it”?  Friends and family, Gabe’s kids…  lots of people working at Valve on their own other projects.  Q: People come back for additional tests?  Usually playtest only once but a few at Valve would repeat play.

 

Q:  In the dev commentary, more portal games?  How do you overcome the challenge of not having as many constraints?  Dunno, just want to bask in the moment now without people bugging him about the next one!

 

Q:  Portals shooting through portals?  Not a tech restriction but design one – didn’t want to bypass training, shooting ourselves in the foot.

 

Q:  Why change the look and color of Portals?  Just iteration.  We liked Orange better than Red…

 

Really like the way turrets screamed…

 

Q:  Suggestions for students breaking in?  Try and make a lot of games and fail.  So many times failed at Digipen…one worked out but the others didn’t.  Practice makes perfect.  Shows commitment to the project.  Make games!

 

Q:  Fire pit –shock and confusion of getting past what you thought was the game’s end.  Defying of game narrative intentional?  Some intentionality there, one of the things that never changed in the story.   Vault > Fire Pit > Glados was the most basic structure.  Worked better than they thought.

 

Q:  Why the humor?  Something inherently funny about portals…  In Erik’s comfort zone.

 

Q:  Why kill off the Companion Cube?  I think we explained that… thank the United States Secret Service.

 

Q:  Lots of trial and error in the game, did that have to be smoothed over with management?  Management at Valve is different than other companies, no producers really.  People on the game make those decisions.  Makes for a more satisfied group of people making games.  Commitment to playtesting is the Valve philosophy.

 

Q: Dev background, how did you score your spot working with Valve?  Students at Digipen, made Narbacular Drop, predecessor to Portal.  Expo for graduating seniors, saw it and took them to Valve to show to Gabe.  Asked them what they were doing after graduating… offered them a job on the spot.  Erik did a lot of writing, was hired at Double Fine, freelanced, and then got email from Gabe one day. 

 

Q: Why not submit to IGF?  Narbacular Drop was in the student showcase.  

 

Q: Where did the cake come from?  Was something in there from the beginning… comedy tool to obsess over something unexpected.  Just having fun.

 

Q:  All characters (main character + glados) female?  Working to define that, used to be a balding dude (citizen model from HL2).  Gabe suggested making it a girl.  Just sort of what if… Post feminist, didn’t occur to us not to make it a girl… For Glados we knew we needed a great actress / actor who was good and wouldn’t be offended when a lot of the stuff was just recording text-to-speech stuff.

 

Q:  Is Chell a cyborg?  No, another failure of writing.  Got these crazy cyber future shoes…

 

Q:  Portal is a short game with a lot of success, might make a movement back towards shorter games on the market?  Common complaint is that people don’t finish games anymore, don’t have time…. Plus 40 hours a week playing WoW.  Practical constraints but really wanted to make a game that everyone who plays it can finish it!  Most games just peter out, never finish them.

 

Q:  Dialog cut from the game, any of it fairly final?  If so, going to release it?  Not a lot of final stuff.  Everything went through text to speech as temp audio, mostly got cut at that point.  Other stuff wouldn’t want to give to anyone, or if it’s good keep it for later…

 

Q:  Considerations for making it multiplayer or co-op?  Thought about it… technically possible, just didn’t have the time to think through the gameplay.  Tried it a little, was less fun than you think it is…

 

Q:  Experience balancing keeping it wet and needing to get it done, or having momentum?  Playtesting!  You’ll know right away whether your idea is good or bad.  If you can’t see how to fix it, cut it!

 

Q:  Dev process as a style of painting, what would it be?  Cubist…

 

Q:  Chell and Gordon gonna hook up?  Dunno…

 

Q:  Can make exposition and narrative fun for playtesters?  Less is more… just don’t like it that much in games.  As concise as possible.  Game writer, but games aren’t necessarily the best medium for stories.  Like providing a film score.  Films aren’t about music but they are a useful tool for enhancing, same for writing in games.  Also a lot to be said for leaving mystery in games, people are smart, give them credit!

 

Q:  Black Mesa references in Portal… when did you decide to be in the HL universe?  Middle of development… when Erik joined just one of the constraints for the game because they weren’t sure of positioning and were reusing a lot of assets.  Eased up at the end, didn’t need the G-Man to show up or anything…

 

Q: Best games teach us something, what do we have to learn from Portal?  “The cake is a lie.”  Explore the idea of manipulating space and thinking about 3d space in a different way, look at situations in a different way.

 

Q.  Companion cube kind of silent, any dialog in the sequel?  What is this “Sequel”…

 

Q:  When did you figure out how to market and sell Portal?  At first was supposed to be a 15 minute tech demo.  Further on constraints changed… middle of development started talking about doing the Orange Box and seemed like a reasonable conclusion to package it with these other games that have a lot of clout.  Honored to be packaged with TF2 and Ep2. 

 

Q:  Playtesting as method to test whether players got storyline.  Compartmentalized testing, how did you get feedback if only playing 1-2 levels?  Portal was short through most of its lifetime, never 8 hours… so if someone didn’t finish weren’t super worried.

 

Q:  Storyline, less is more… applies to other genres?  Don’t have a surefire solution for every game but in Erik’s opinion, in general, less is more. 

 

Q:  Portal and HL2 have very different tone and themes.  If they tie together more, how will you resolve that?  Honestly not a big problem, Aperture tone is mostly just from Glados.  Experiencing the world through Glados’ eyes, so if Chell is in HL universe nothing is in contradiction there.  Inside the Aperture Funhouse different tonal rules apply.  Same physics and art style already apply.

 

Q:  Chance of an artbook?  Dunno… talking about doing an Orange Box artbook.  Most of our concept was for the Glados battle.  Compared to TF or HL not a lot of concepts there.  Thumbnails of stuff… but not really great pictures.

 

Q:  Why portals ovals?  They look better than rectangles…  Wanted to fit the bounds of where you’d actually pass through, so filled an oval through the rectangle of where it technically is.

 

Q; Game alludes to a backstory (test subject), how much actually written vs just use your imagination?  A fair amount, 15-page document about Aperture science, some of which is hidden on the Aperture website.   Sent it to Jonathan which helped for the song.

 

Q:  Portal was a triumph of creativity, awards and acclaim and so different.  How does that affect the community of games as a whole?  Approach to making games as a creative medium?  Too early to tell… I hope more people make games like this since we like them!  Independent games are really getting awesome, higher profile and production quality.  Maybe people take more risks on more oddball games…

 

Q:  From a film background… as writers, what are your influences?  Kim Swift:  Miyamoto, Nintendo games… knew she wanted to be a game designer from an early age.  Katamari Damacy.  Wolpaw:  Direct inspiration was a book, “Destination Void”, clones that go into space and tasked to build the AI, failsafe mechanisms to deal with AIs that go rogue. Doesn’t read a lot of science fiction now… Roger Scheckley, funny sci-fi. 

 

Q:  Portal design process was an animal, what would it be?  Joke question…

 

Q:  What the hell is on those slides!?  Friendship…. And fire…. Just pictures we couldn’t fit into the talk in any meaningful ways. 

 

Q: Dev team, female lead is unusual.  How many girls on the Portal team?  2, Kim and one of the artists.  Q:  Does that help in the creativity process?  Dunno… honestly, no matter what your gender you make a game you want to enjoy yourself.  So maybe by extension but should be fun for everybody… Erik:  makes the room smell better….

 

Duelin’ Firemen!!!!

GDC Rant Session 2008

 

Goal:  to make these underground conversations public.  Do these rants direct the work that we do?  Keep a finger on the vibe of issues that are concerning developers.    Last year’s rant helped IGDA create a speaker’s bureau to help our presence in the community.

 

This year: game designers ranting.  2008 is the year of game design.  Great lineup at the awards: Portal winning GOTY.  Indy games being recognized was strange but now it has come home.  Not easy being a game designer.  Often they are listed below the people making the code and graphics.  Other disciplines come from other places, art, business, code.  But game designers are ONLY in our industry.  They are the heart of what we are doing so they have a particular animus about being kicked around and wanting to make our industry better.

 

Clint Hocking:

 

First idea, ranting about creative stagnation.  But such a tired and generic topic, and not even sure there really is one.   Pound for pound the most creative industry in the world.  “Dude, it’s code… we can do anything”.  Being creative is fairly easy.  Having the courage to create something that challenges people – that’s hard.  Why don’t we make games that challenge people, that matters?  Art and books do this all the time.

 

Here’s a game that made me cry, heres a game that means something (Marriage).  Effective budget zero.  Meanwhile, we made a bunch of versions of Duck Hunt in the mainstream.  Why don’t we take those techniques and use them in our blockbusters.  What if the things you did in the game supported those relationships in a systemic way, not a cutscene?  What if the goal was to comfort my spouse and the monster was just a metaphor?  Why can’t CoD actually be about Duty?  Or MoH about Honor?  What would it be worth to you if you could put honor in a box and sell it?!  Not Tom Cruise on the front, but package the experience of knowing what it is to be honorable?  Most people in the world have never felt that but they would love to – and they would have to practice it a ton to get a 5-star honor rating.  What if you could get the Medal of Honor in a game that demanded you be honorable to get it.  How terrible it would be to have people who are honorable because of our games?

 

Meanwhile Passage and Marriage are free.  Sad that these two guys have advanced the industry more than all of us making AAA games.  It’s great that we have little games but I like making big games with 200 people in it.  Not talking about taking action from Halo, but using proven techniques for making emotional investment so that Halo will reach a real audience.  Halo 3 reaching 1/10 the audience of LOTR movies.  The things that move people in LOTR are the dagger that glows and a +5 rope?  People care about the fact that Frodo has to trust Sam.  Mechanics of trust are not harder to model and simulate than the mechanics of rope!  Yet our games are full of rings and ropes and armor and we have this object fetish.  So the most meaningful relationship developed this year is with a CUBE! 

 

Maybe it’s nothing and creative.  Not creativity that is holding us back from being the dominant media.  We lack the courage to make games to challenge something in us besides reflexes.  Lack the courage to be seen crying in the movie theater when he thanks Sam.  That’s the only thing separating us from the mature medium we will become and the juvenile media we are today.  We have the money and the creativity, and the demand.  Dude – it’s code.  We can do anything.

 

 

 

John Mack(?), Indy game designer (Everyday Shooter?)

 

(Everyone stands up and bounces balloons around)

 

 

Another Person, Kim ??:

 

Some sort of rant?  Nothing to be angry about now.  

 

Thus ends John Mack’s rant.

 

 

 

Jane McGonigal

Avant Game

 

We’ve invented a medium that kicks every other medium’s ass.  We occupy more brain cycles and make more content than anyone else in the world.  People are waking up and pouring effort into these games.  Games have already won.  

 

Industry spent last 30 years optimizing human experience.  In the past, our brains for playing games and recently our bodies for playing games.  We know our hearts can play games too.  Our experiences better designed for engaging humans than anyone else.  We basically rule the world.

 

The bad news:  we rule the virtual world.  Reality is too messy and broken, and we don’t want to fix reality, we want alternatives to reality.  For now all of our games work better than the reality.  All our effort into optimizing virtual experience.  We need to change that.

 

Reality is broken.  Our responsibility as the smartest people on the planet to fix it.  How to make people superheroes in real life?  Make the real life more like a game.  Graffiti – “I’m not good at life.”  I don’t wake up every day with missions and purpose.  Can’t be good at reality the same way.  In the games I have all the information I need and skills and people to count on.  Just better than real life.  New field of “positive psychology” – what gives us a good experience?

 

Satisfying work to do

 

Being good at something

 

Time spent with people we like

 

Chance to be a part of something bigger.

 

We are making the ultimate happiness engine.  “Soap Kills Germs”.  Figured that out in 1932.  20% of people in hospitals killed by germs on doctors hands.  Maybe people not on their deathbed but their lives suck – people playing and buying our games, games are their highest quality of life.

 

“Why should we care about games?”  Because life is crap, makes it worth living is art and play.

 

“Games kill boredom”.  “games kill alienation”  “games kill anxiety”  “games kill depression”.

 

Games are the ultimate happiness engine.  Like we invented the written word and decided only to write books, not exit signs.

 

5 things to fix today

 

Running, being on a plane, playing fetch.  Running == leveling up?  Flying = going to die.  Virgin has chat system, why isn’t there a game to play on the plane so that you win when you land?  Social network for dogs?  Play games with our dogs, fetch for an hour every day.  If I could play an MMO for my dog.  Trackstick – records where you go and see on googlemaps.  Virtual environments on commute.  Neurosky thing, knows if you are relaxed.  Zap people who are lighting up my brain.

 

Can make games that interface with reality.  Yes we should fix it. Will we?  I have no idea.

 

Alter your reality.

 

 

Chris Hecker:

 

“take a civil tone next time you speak, and even if you make the same points maybe more people will listen and find that they agree”

 

“Maybe the art that games could be isn’t some sick mishmash of what has gone before?”

 

Computation power is not orthogonal to game design – last year’s rant.

 

Still into ranting… complacency = death.  Games could be the preeminent art form of the 21^st century.  But on our current trajectory, will we get there? 

 

Take our boat and sail it between Scylla of thinking we are all that, and the Charybdis of being too scared to talk to the human condition.  Sextant for this journey is harsh and honest criticism.

 

Ranting has to be followed by acting.  Constructive criticism is the best thing.  But just criticism by itself is valuable.

 

Speak your mind and tell the truth.

 

 

Genova Chen(?)

 

Job is not to start fires, but to make fun stuff.  2008 hard to rant about, great year of games and innovations.  Current talk about input device is just no longer true.  Games are innovative, are art, can engage gamers with emotions.  Digital Distribution, barrier to entrance quite low.  Lots of indy games moving to the console!

 

How to come up with better game designs?  Not just a designer to rant.  But can rant as a gamer, just a guy playing games.  Less games in the past 18 years.  Interest dropping.   Excited to play first racing game , but no feeling at all in new racing game.  Excited to kill nazi’s, but now not exciting. 

 

Learn.  Brain engineered for sucking information.  Games can teach.  But more than toys?  Never say too old for you to read books or play sports, why too old to play games?  Intellectual, emotional, social. 

 

Grown-up gamers?  Don’t want to be abandon game because too old.  Need more mature content for people who still want to play games.  Not talking about manhunt or DOA.  That’s really for 18 yr olds.  Average game age is 33.  What do you need – intellectual, emotion, social content.  WoW == social doing OK.  Last year, feeling.  This year == intellectual.  Can games make you learn something?  Make you think?  We don’t need Three Little Pigs, we need Little Prince.  Make you think about your life.

 

What knowledge do we have beyond making weapons or level design.  Other great art comes from author’s own experience in their lives. Passage is awesome == gameplay makes you think.  Think of gameplay, not story.

 

What do you want to share with the mature gamers?

 

 

 

Daniel James, Three Rings

 

GDC is inspirational.  That’s nice.

 

Played with Lego as a kid.  Now they are making Lego MMO on the internet.  Brilliant!  After Lego, played Elite.  Used to dream of playing that with other people – now we have Eve.  Fantastic!  After that played a Mud.  “go north”.  Chat with people – now there is 10 million people playing WoW!  WTF Amazing!  “The Tomorrow People” – we are in the future.

 

20 million people using this Facebook app!  Totally awesome!  People given money by investor types – that’s incredible.   Sea change in how we bring games to the world.  If US econ fails, hey there is always China!

 

Grandma saying now she plays PP with her grandkids instead of watching TV alone.  Kill TV is a mission we can all appreciate.  Virtual environments set off the same mind patterns as those things to in the real.  Line between the two is very thin.  Puts us in an incredible place.  People do things and then justify reasons afterwards.  How does that apply to you?  Observer is the pivotal point in shaping the real.

 

 

Q:  game design touching real world already in a bad way – frequent flier miles.  Now everyone has it and it’s a leech on the system, you get the money back that you never should have paid in the first place.  What if I just want to run and enjoy the simplicity of things?  People who want to run can run anyways – but a lot of people would be happier if it were less painful or embarrassing.  Making things difficult for people better.  Example of flier miles is great – gamelike isn’t just points or rewards.  Promotion people aren’t game designers.  Beauty of play, those who understand it should be teaching others how to be more engaging.

 

Q: Reviews of Flow give bad score because not enough boss battles.  Failure of discourse?  Role of game criticism?  Game review community needs to grow as fast as the developers.  As designers we are a bit head.  Majority of review can be bought?  How do teach game reviews?  Movies have classes, degrees for this but reviewing games has no education.  Needs to catch up.

 

Q:  Growing rift between hardcore gamers and everyone else?  What happens as we age?  Hardcore gamers are fighting a losing battle.  Want to play games with increasing convolution and that’s cool, but that isn’t where growth is.  Don’t have time to play those games anymore.  Only going to reward XP for time, rather play Passage.  If maker of Far Cry 2 is saying the same thing as the guy who makes Flow, there is something there.  Retaining people when they get to be 35, you need to make it something they care about.  Can get XP in a hundred games, can’t get honor in any of them.  Want to explore that in games, whether we would really stand up in a tough situation.

 

Q: What is stopping us?  What’s the barrier?  Biggest one is what we think about as a “sexy” game, how big and how much graphics it has to have.  Think of other things as being sexy that would be a good start.  Barrier is thinking that there is a barrier.  That there is something you have to do, to do it right.  Figure out what makes you happy and do that.  Tweaking numbers, or complex rulesets – just do it.  Don’t worry about this rift, who cares?  This is your art, your expression – own it!  Make it your own!

 

Taming the Mob: Creating Believable Crowds in Assassin’s Creed

James Therien & Sylvain Bernard, Ubisoft

 

AC is an animation intensive game so crowds deal a lot with animation. 

 

Measuring our expectations: Looking at other games, getting reference from real life footage.  GTA.  Focus more on crowds.  Consoles of this gen weren’t even out yet so hard to know their capacities.  Interesting trailer:  Dead Rising, crowds of zombies.  Showed what the console can achieve in numbers of characters on screen. 

 

To create a realistic crowd:  move in tight areas, realistic contact, proper reaction to the environment.  Incorporate the crowd into the gameplay. 

 

Rich NPCs: every NPC is an individual.  Emergent crowd, not explicit in code except in spawning.  Make those groups of NPCs believable and the crowd emerges from that.

 

Level design and art direction was a real life scale.  Didn’t want Prince of Persia – realistic capabilities.  Did need to cheat sometime – survive a 66 ft drop (20m). 

 

Animation style: keyframe editing, mocap where possible.  Same group as did Sands of Time, brought over animation like walling, but with more realistic touches.  Mocap editing to match keyframe look.  Add overlap, create more impact, adjust posing.  “Stylized realistic” animation style.  Skeleton shared between all characters.  Create opportunity to replace the animations.  Male or female, thin or fat.  Did some scaling if needed (except on legs).  Headache to modeling team to have to fit those meshes onto the skeleton, but was a really good decision.

 

Layering animation – every NPC shares the same behavior system but if want to customize (more feminine) just replace the needed animations within the system.

 

Lots of bones in our characters!  55 in main skeleton (w/hands), 35 in head, 40-80 procedural for hair, swords.  36 bones in the pigeons!  Hinge bones:  simple physics – wind, gravity, momentum.  No collisions, just constraints.  Used for hair and robes. 

 

“Look at” – 3 independent targets: torso, head, eyes.  Eye blink.  Control the eyelids with the characters look up (lifting).  All NPCs doing that.  Small subset of total: IK, lipsync.

 

Movement system – complex graph.  More realistic = controls not responsive.  Really complex = too much impact on other systems.   Separate the walk into 4 stages, each with their own transitions.  Starts from every direction.  Maybe a bad decision, started to explode when linked into other systems.  3 weight poses, so all systems needed to start on those 3 weigh poses, so simplify where possible.  When you release control the animation takes an extra few steps to look realistic. 

 

Went to a simplified version, keeping maximum fluidity.  High / low profile – has a transition between the two to keep it more realistic rather than just snapping.  NPC and assassin use the same animation system.  The computer is like a player, playing against the human.  Apply all of assassin’s animations to random shopkeepers. 

 

122 movement and transition animations.  108 transitions.  14 movement, 10 wait and idle.  Left-front vs right-front versions for movement.  12 cycle breakers.  24 other.  168 total animations.  So every new major version required 168 of them.

 

Movement system – displacement node.  Drives animation forward.   (key point that drives movement).  Crowd swimming, cycle breakers, and walk stop.  Swimming – shifting shoulders sideways.  Makes them look more lifelike.  Torso and arms only animations that don’t need to disrupt the walk cycle.  Can stop for just a moment to show the flow. 

 

Oriented move.  NPCs needed to walk backwards, which the NPC does not do.  Blend from backwards to forward to turn around while fleeing.  Became problematic since the assassin did not do it, and for navigation purpose.  So not actually used in the game. 

 

Environment:  not on a grid, very large.  Navigation not constrained to ground plane.  Totally arbitrary environment, irregular, beams everywhere, not fixed heights.  Detail everywhere.  Generate navigation data from collisions.  2 layers in AI:  behavior and decision.  Behavior shared with the assassin – managing animation and doing fine-grained interaction.  Guidance system – lines with normal information to cast against and interpret those lines in run time.   Extremely expensive, so used only in specific cases like assassin climbing.  Guards chasing use the same code as assassin. 

 

Other data is more AI specific – 2.5D triangle mesh for flat spaces.  Local planning and tests.  On top of that generate a waypoint network.  A* pathfinding on top of that.  Object waypoints and metalinks – jump data encoded ahead of time.  Use guidance data generated for assassin to compute jumps offline and then stored compactly.   Also encode ladders, beams, etc.  Steering pass to use long-term nav data.  Continuous small corrections for avoiding walls, dynamic objects.  Path validation to remove waypoint placement artifacts.  Communicate with decision layer if there is trouble (pushing away other NPCs if stuck). 

 

Level design specific tools for navigation.  Fine for Pt A to Pt B, but need to specify where crowds go.  Crowd flow lines, navigation highways that the level designers place in the map.  Makes wandering very cheap, useful for fleeing.  Only thing they have to do once on the network is decide what branch to take at an intersection.  Fleeing has no specific destination, just want to get away, so they can use that same data with a different decision at the branch points.

 

Interaction with environment:  Unbalance system.  Completely animated, rag doll activated at the last moment.  Ragdoll can look like fainting, but wanted something to feel more like protecting himself and having energy.  Versions for each possible height and speed and outcome (falling, making over, recovering).  Same thing for falling off a ledge – by speed & facing. 

 

Crowd Density:  Spawning in cities – originally just in a sphere.  But as we added NPCs you wouldn’t see them, they would be in streets you couldn’t get to or on other side of buildings.  Also couldn’t see down long avenues – just increasing the bubble just put more out of sight.  2d mesh and do a flood fill.  Pick triangles out of LOS to spawn.  Don’t bias the blob in any direction because player is so mobile.  Just make sure connecting streets always full.

 

Creating variety:  Entity builder.  Mixed attributes on spawning.  Endless possibilities – head structure, textures, color, thin, fat, military, accessories, AI reactions…

 

Crowd composition:  Game play crowd.  Base walking crowds + bench, monks, beggars.  Lots of game play relevant crowd entities.  All this together gives an immersive world.  Had intended more duties but really just did kiosks.  Wish we had more things like sweeping, drinking from a fountain.  Sound faked some of this – hearing someone working metal even if you don’t see it.

 

Interaction:  Physical interaction – soft push, grab and throw, collision, fight, assassination.  Reaction system.  Communicating design to programmers by generating fake footage of the different reaction scenarios.  Reactions:  individual reaction, sound from an NPC, body posture.  Acrobatic reaction creates zones – yellow outer area to just look, grey area to stop and look.  Then people slowly peel off.  People following a fight sequence, attracted as long as no one dies.  Chase sequence with recoil, dodging the assassin.  Some just get bowled over if they don’t notice.  Then generate fake conversations when pedestrians wind up nearby.   Took a lot of work to keep the NPCs attracted to a fight from getting in the way.    Body gestures to match sounds.

 

We often focus on textures and meshes, but the AI and behavior is just as important.  Need to bring the world to life. 

 

Reactions: more than just visual.  “reaction packs” specified by level designers.  Set of response to specific events.  Draw the guards into the fight by reacting to the alert event.  Guard awareness levels – seeing a dead body changes reaction packs. 

 

How do you get lots of NPCs at 30 fps?  Not easy with no LOD on decision or behavior layers.  Lots of LOD on the animations – bone counts drop fast, drop IK, simple look at, simple procedural rigs.  Focus on making common NPCs as cheap as possible.  Don’t do stuff when NPCs out of sight.  Lots of NPCs only pathfinding to get to position or crowdflows.  Load balanced fairly well – a fair number always doing common tasks.   As long as just player and a few guards being expensive it balances out.

 

Concurrency – lots of multithreading.  Started with just render / engine, but not balanced.  Multithread the engine the Wednesday before a big demo…  Animation multithreading worked out well.   Successful so went to a general distribution system.  Filling up the physical threads on the 360.  Have to balance well, optimal result is not the goal, steady frame rate is.  As much SPU use as we could on PS3 but didn’t work quite as well.

 

Conclusion: met our goals of a believable crowd.  Quality focus with good support from management.  Lots of diversity, quality, and quantity.  Harder than planned to incorporate gameplay – emergent crowd interactions hard to manage, difficulty for player to use the crowd as a tool.  Fell short of that goal. 

 

Q:  Size of the team?  Animators, programmers… 150-180 total.  1/3 programmer / art / design split. 

 

Q: When you spawn a character, is the duty fixed for their lifespan?  Sometimes yes, sometimes no.  Different types of NPCs, most are fixed duty, but sometimes just released to the crowd.  Bench guys fixed but if they are servants they can get up.  Big system to do simulation but realized that it didn’t matter because the life of an NPC is so short.

 

Q:  Tall AIs and short AIs – how to make climbing work?  All environment detection is dynamic, all at runtime, so it all just worked.  Dynamic responsive to scale. 

 

Q:  What were goals for expanding gameplay for player?  Diversions…  no need to add new types if old types already worked.

 

Q: performance vs memory?  Measure a lot…  did limit variety.  NPCs expensive in many ways.  Tried to make them cheap as possible for common NPCs, but there is a limit to the work you can do.

 

Q: flood fill for spawning… visual test for spawning?  Yes – lots of raycasts in the game.  Raycasts for everything.  Multiple raycasts to make sure NPCs spawn out of sight or whether NPCs should react to something.  Spawning was asynchronous so didn’t matter how long it took.  Iteratively evolved flood fill triangles, since that was expensive.  Took a lot of optimization on that process.

 

Q:  how did you test the crowd system?  Humans, mostly.  That was a problem, systems breaking other systems without realizing it.  Lots of testers, that’s pretty much it.  Looking into other solutions for the future.

 

Q: reaction system, use of period events to represent continuous stimulus.  Created problem with moving player creating reaction as he arrived?  To avoid delay…  solved with brute force.  Not continuous but almost so.  When you start / stop moving sends events, fill in the difference.  All level design controlled, so they could tune and control.  Primary and secondary reaction zones, mostly data driven.

 

Q: duties planned, how many wound up with?  Not much, just orators and kiosks, used to create cluttering. 

 

50 software engineers at the peak.  AI was 15-20.

 

Q: What did you learn from AC to support much larger numbers of participants?  Crowd behavior changes drastically as you increase number of people.  40 NPCs worked, +20-30 more everything broke down. Pattern of motion extremely different as crowd gets denser.  Not just a question of detail, especially for navigation.  Each NPC has their own speed of walk which didn’t make things easy.  Suggest starting with a big number even if you can’t render and see how it works, don’t go incrementally.

 

Q: What sort of solving for complex scenarios like intersections?  Too complex to discuss here.

 

Q:  If there is a sequel, what crowd things differently to fix shortcomings?  Big problem at start was we didn’t know how it will work out.  Now we have the base and we can really build from that.  Expecting better integration with gameplay, going to be a big focus.

 

Do, Don’t Show: Narrative Design in Far Cry 2

Patrick Redding, Ubisoft Montreal

 

Far Cry 1: Invigorated setting of beauty and menace.  First third had a lot of lushness and realism, menace of armies of mercenaries.  Ubisoft decided to buy the brand, take it in other directions as well. 

 

Freedom to explore all the spaces:  systems / spatiality / humanity? 

 

Realism with focus.  Build on that notion of realism, but not for realism’s sake.  Use it in a way to let the player parse the game and its systems not just from the game but because they are human.  Wind blowing in a particular way they can make reasonable and intuitive choices based on that.

 

Immersion with Consequences.  First person can be harder for immersion because you are surrounded by this bubble.  How to selectively overcome that barrier?  Remind the player of their physical presence in the world. 

 

Meaning backed up by mechanics.  About something the player could actually experience.

 

Vision:  Themes > Intentional Play.  What does it mean to be a social animal?    A lot of far cry is around this dangerous setting.  So take other elements associated with those environments, how do they translate into a fast action shooter environment?  For humans… betrayal and subversion are your teeth and claws.  Ultimately you will run around with guns, sure, but this other element could create a human space.  Player input: every bullet counts.  Firing in the real world doesn’t just do health damage – fear, pain, uncertainty.  Causing a disruption of the social order.

 

Island of Dr. Moreau (FC1) > Heart of Darkness (FC2).  Colonial ambitions lost, failed adventures in Africa.  Any human being, no matter how civilized, when staring into the abyss will find it staring back.  Choice between false veneer of civility and the savage truth.  A journey up a river into the mind of a madman.   Modern version: Apocalypse Now.  Brings in a lot of modern shooter elements which are useful.  An irregular battle.  But making just a boat simulator wouldn’t be all that fun:  really just an allegory.  Player is in a fictional African wartorn country to track an arms dealer and terminate them.

 

Theme branches to game mechanics and story.  Red Harvest (Dashiel Hammett).  Yojimbo / Fist Full of Dollars.  Solitary gunslinger, anti-hero, strides into town corrupt and bereft of pity.  Disgusted by the pettiness of the warring bosses he doles out hard justice by manipulating the factions to take it to the guys in the charge.  Perfect model for FC2, could be broken down into great symmetry of gameplay.

 

Surviving in a social wilderness.  By the end of the game if you are piled in bodies you can lose the sense of the stakes.  The sting of betrayal is hard to ever get over.  Augment the stakes of a first person shooter by getting that sense of social survival.

 

Selective awareness of one’s physical presence in the world.  Player not aware of his avatar 100% of the time, focus on times when you DO have to do that, getting into a car, stubbing your toe.  Very scalable, could keep cutting away until we get to the core. 

 

World is not just an Africa-shaped arena.  Savannah, jungle, desert.  Don’t really all exist in one place so invent a fictional place.  All the real African conflicts are loaded with history that we don’t actually want to be tied up with.  Space where it makes a difference that it is In an African setting.  50 square km space with all these different biomes.   

 

How do you present a story when you don’t have any idea where or when the player will be?  Player controls the order + scope of missions.  Player has the hand on the throttle – either directly and consciously or just indirectly through his actions.

 

So what does this say about our story?  Premise: distinct from story.  It’s the thing we do have control over, the formal constraint we are imposing on the game.  The player is going to go to wartorn Africa and set these factions against each other.  Premise as distinct from the story, which emerges from the player’s actual gameplay.  Many ways to experience the story.

 

Enacted, embedded, emergent.  Jenkins: Game design as narrative architecture.  Enacted: as player picks up the box, the traditional story structure.  Embedded:  environmental narrative, dialogue and exploratory narrative.  Emergent:  player has a story that unfolds in their head that is profoundly personal and has everything to do with their gameplay choices and personal interpretations.

 

Illusion of continuity > anything else.  Open world with a lot of stuff, so what is primary function of the narrative?  To maintain that illusion of continuity.  Doing a lot of moving around (in space and systems), lots of chance to trip over discontinuities, so help guide the player away from those.

 

Be stochastic.  Game world that is large and filled with people and activity is a noisy place.  High fidelity environments but struck by how “quiet” everything is – verisimilitude, a dead world.  We want a live world with activity, characterized by chaos.  Not randomness – unpredictability that we still as human beings can parse.  Deliver story content using mechanisms that feel natural.

 

Exploit the urge to anthropomorphize.  Projecting intent on simple / stylized characters.  That weird “Mii” moment of projecting your soul.  Looking at a baby seal.  Very effective tool to offload some of the processing burden onto the player – not their console, but their head.  Cues to suggest a level of depth and intelligence but let them fill in the blanks.

 

Game ingredients > story inputs.  Take as much information as we could from the player based on their use of the game ingredients.  Normal FPS = moving around the world in a Teflon bubble.  Thin pipe / fat pipe bubble.  Awesome output in rendering and effects, but player only has a tiny trickle of input.  We weren’t making a conversation game.  How do we take those core mechanics of running around and blowing stuff up, doing missions, and use that as our story input?  Track what the player is doing, assign some value to it, and then use that to dynamically assemble to the story.

 

Set the game flow – then break it.  Act  > Mission > Objective >Sequence.  Devise this gameflow and critical path, and then figure out how to break that into pieces.  Find smallest pieces, then link those pieces together. 

 

Pull, don’t push.  Set beacons up that would draw the player voluntarily to places where interesting story moments could take place.  Can’t happen uniformly over 50 sq. km.  Let the player build that intuition and learn to trust it.

 

Player character belongs to this moral universe.  Giant cast of NPCs?  How to understand what they do, and how the player fits into that.  First-person game, we’re not providing them with a lengthy dossier of their background but give them an empty vessel that they can pour their own ideas into.   This is my vehicle for entering this dangerous world and drawing my own conclusion.  The Gordon Freeman route.  Leave the character a bit of a cipher – this is about universal concepts.  Show that belonging by making this family of surrogate NPCs you can encounter and interact with.  NPCs aren’t just talking heads.  Connection to friendly NPCs: buddies.  Player is picking his avatar out of a set of characters that already belong in the world, this became very important to the game.

 

Reputation is a gate.  Are we making a hybrid RPG / shooter?  Well, there are some RPG elements.  Rather than using player action in a blunt way to gate content, instead we built out a notion of reputation into “infamy”, something mechanical that the player can build through low level game mechanics.  By building that up the player can fundamentally alter how the AI in the world interacts with him.  Situations where you walk into a cease fire area, there is a fundamental attitude shift.  A source of the players power: to be infamous, and apply their creativity to being cruel and bastard-like.

 

Cultivate AI ecosystems.   How to have these scores of characters moving around and not being completely robotic?  Solution had implications for the story and the level design.  Giving player 360 degrees of freedom, can approach key locations from any direction.  Can’t create levels like in Rainbow 6.  Needed a more stochastic way to deploying enemy AI in the game world.  One approach: Sims like method, primitive set of needs and desires that the AI can address through affordances scattered in the game world.  I need to go on patrol > there’s a patrol route.  I need a rest > there’s a break spot to smoke a cigarette.  Creating a social wilderness via AI, creating social needs.  Anthropomorphize.  Dialogue between AI are about the events going on in the world, the player’s AI.  Manage challenge but also manage story.

 

The challenges:  Getting the player’s attention (continuity design).  Generating tons of content to listen to and explore.  Huge branching tree of combinatoric mess?  Dynamic drama management – buckets of parameterized micro-content.  When we do a “call” on the story, we can take a snapshot of the world and based on that state select the correct pieces of micro content.  Affords us a level of flexibility – those buckets of content aren’t wasted.  Reputation is relevant on the next mission as well, mission history continues relevance, etc.  Conserve a large quantity of content (anim and dialogue) so that we are literally not wasting it and only hearing 10%. 

 

Locations of narrative significance.  Player can express themselves stylistically in a lot of different ways.  Rambo-ing in vs tactical approach vs stealth.   Need to be prepared for the fact that the midlevel gameplay loop might be radically different so the story presentation needs to be different too.

 

Don’t show the seams:  Systemic vs Choreographed.  Need an overlap between the two extremes:  blurting out tactics vs staged theatrics.  How to avoid the empty space between the two poles?  Beautiful scenes vs random profanity.  Rely on depth of behavior.  Not about baked, scripted events that play out exactly one way but have a lot of generic content that we can use in lots of places, in lots of ways.  Tradeoffs, for sure.  But the world becomes a lot more interactive.  Bricks of micronarrative that are the smallest pieces of content:  has to be voice recorded, animated, etc. 

 

Don’t presume to know what is in the player’s heart.  We can do all this stuff but we can’t know how the player feels as a result.  Might react with antipathy or hatred towards a character we want to be sympathetic.  Don’t tell them about the human cost, make sure they can experience it first hand, something they have to deal with.  Whether they care about it or not is not our problem, we can just make sure they experience it.

 

Story is not just about the data: it is also about the systems.  Ratio of systemic vs scripted, 4:1 or 5:1. 

 

Roles?  Traditional structure:  CD > AD > LD > Writer, loops back for shoehorning in.  Not going to work for us.  Dedicated narrative designer in the center of the structure to work closely with writer as well as everyone else in that graph.  Keep the writer not doing book-keeping.  Focus on the parts you can make shine. 

 

Three act narrative structure is a trap!  Tempting to get sucked into, human beings instinctively know what this is.  Farther you carry this model harder to let go of parts you want the player to have control over.  What if the story ignores the players choice of ingredients?  Periodically abduct the player into the storyline? Not the best way.  Pinch-bottle story architecture.  Story guides the player away from the hard limits of the simulation space.  Story helps player intuit where he should be to understand the limits of his own agency.  So at those key moments the cushion pinches together and forces the player into one location.  A little jarring but not as much as being beaten on the head and abducted.

 

Authored story events.  If that island of story is just plopped in, player will move into it abruptly.  Instead we want “shallows” where the nature of their agency can shift slightly on the way in.  In general make the experience more analog, don’t want binary transitions between gameplay and narrative. 

 

Lessons from sports:  Snatch victory from the jaws of defeat > much better than some story we wrote. 

 

Decouple premise from story!

 

Exploit the narrative fallacy.  Humans interpret random noise as a storyline – focus on the exciting parts and ignore the other stuff.

 

What do the players actually do?  Playing the Red Harvest scenario – playing two factions off of each other.  Low level – shooting bullets.  But kill vs wound or putting damage in to disrupt the social order.  Subverting the faction’s goals.  Missions have simple objectives but take the suggestions from their buddies as a way to augment the outcome of the mission.  “Be careful what you wish for” – taking it further, boosting infamy, making the conflict worse.  At the high level they are altering the specific order of missions they are given through the act of betrayal.  Throw more uncertainly in the system by escalating on both sides.  Destroying the status quo.  This is how the player moves the story forward.

 

Player spends a lot of time in that space between choreographed and systemic. 

 

What do the systems do?  Infamy twists the power relationship of the survival game.  As you build infamy up your are altering your power.  Making you tougher, more terrible, more likely to be more terrible in combat – positive feedback loop to create more carnage and disarray. 

 

Infamy progression – three act structure but used in a way that is systemic.  Infamy as a stat clamped to where we want it to be in that structure.  Finish an act, suffer a defeat and reset the player’s infamy.  Lets the player compare where it is between the top of the heap and the bottom.  Survival is not just physical – status has been reduced. 

 

Malaria – a mechanical drawback.  Stricken with malaria, starts out sick.  Immediately has to find medicine and overcome the symptoms.  Helps to restrain the player’s range, but also limits the player’s maximum health.  Malaria medicine can only be gotten from civilians.  Show the player the human cost in those civilians.  Doing small tasks for them the player secures that medicine.  Whether he cares or not he is confronted by it at every opportunity.  MaxHealth progression different from infamy – when your infamy is maxed out civilians wont cooperate and then you can only rely on your infamy and you have to become a monster because you are too sick to do anything else.

 

Buddies > Objectives > Infamy.  Build history by doing those objectives for a specific buddy, increased chances of that buddy being implicated in gameplay later on.  Building social cohension – lowest level you have life and death power – you can choose to interact in a friendly or unfriendly way.  If they rescue the player in combat they are at risk and can be killed.  If you take the mission they offer that also exposes to risk.  That family is something you are forced to deal with, your social survival is tied up with the survival of these characters. 

 

Friendship isn’t an optimization problem.  Didn’t want them to be specialized, or be a burden where he was tempted to cull them out.  Develop authentic feelings, even if they are bad.  Antipathy is better than apathy. 

 

Social dynamics shaped the world. 

Last writeup of the day, as I was feeling awful by the mid-afternoon of GDC and went back to my hotel room to sleep.  Sinusitis FTL. 

 

The highlight of this was definitely a long-ish demo of Prototype.  It looks quite exciting though 2008 is really shaping up as the year of amazing open-world games so the competition will be fierce!

 

---

 

Prototype:  Open Mind, Open World: Next Generation Thinking

Tim Bennison and Eric Holmes, Radical

 

Prototype, new IP from Radical.  Radical, 17 years in business.  30 million units.  Simpsons Hit and Run, Scarface, Hulk UD.  Epic Combat, over the top locomotion in a real world setting from Hulk. 

 

Give the player something new.  Not just better, give them more.  More leads to a new experience… Something new.  The Prestige … not just X meets Y, where is it going? Value!   Why bother?  Others are exploring “better”, is that where new experiences lie?  Better taps out eventually… 

 

Wolf 3D > Goldeneye > Call of Duty > Call of Duty 4.  Visual quality of better is slowing down.  Approaching real, and then what? 

 

“Next Gen Open World”.  What is next gen?  Just hardware?  It’s a software change, something that creates a new experience that was not previously considered possible.    Some are More and Better, Elite on the C64.   What is open-world?  OWG – free roam, sandbox.  Systems interacting = fun.  Players behave badly in OWGs, so go with the flow.  OWG is a medium, can be used to tell a lot of stories.

 

Player character, simulations you can poke / collide.  Environment, player freedom with parallel story threads.  Tough to make. 

 

Protoype, a “from the ground up” OWG concept.  You are the prototype, a man without memory and shape shifting abilities.   Mind is a blank, fill his past with the minds of his victims.  Steals memories by taking them from the physical form of other people.  Dangerous + Aloof + Calculating.

 

New IP, constantly evangelizing internally.  Shapeshifter, become a mailbox – no no.

 

Tom Clancy meets Stephen King.  Military, real world + distrurbing.  Reality plus one fantastic element.  No laser beams and hover tanks, but one element, “the virus”.  Sweeping through NYC.  Make it feel real, a real place, makes it much more intense.  NYC in current day is closer to home.   “What the hell is that” – look in closer and find something disturbing. 

 

Story touches on real world issues:  viruses, martial law, conspiracies. 

 

Get team on the same page to narrow focus to the project.

 

Game Pillars:  Deadly Shapeshifting Action, Power and Agility, Deceive or Destroy (player choice), NYC is your hunting ground (NYC as a character), Unravel the Conspiracy.

 

New tools in an OWG, “consume and become”.  Not a hero, not a villain, but equipped to do things the player wants to do in that setting. 

 

“Web of Intrigue”, hundreds of memories that form the story.  Player driven.

 

Locomotion: adaptive open-world parkour.  Seamless integration with combat.  Movement on dynamic objects.  Scale of movement vs fidelity.

 

NYC: grounds the whole experience.  World evolves through the story, not just a snapshot.  “cascading effects”  Metagame – 3 way war across NYC.  Player, military, infected.  Immersed in it and you can affect it.

 

OWG production: systems deliver content.  System = engine, content = fuel.  Order of construction is important.  Visibility comes late because systems are the experience.  Production flow:  build systems then add content.  Parallel system and environment stuff.  Build game content comes late.

 

OWG, player nature is key.  “oh, can I do that?”  Make stuff up as you go along.  Players behave badly, that’s in their nature.  Can you make the player care?  Generically, no.  But about their things, yes.  So no tottering school buses, but RTS base model works.   Specific characters.

 

Basic structure:  kill guys and take their stuff.  Interconnection of systems is much richer in an OWG.  Those interactions are the good stuff.  This is where the rubber hits the road.  Goal is lots of connections between systems.  Example: bases in isolation, vs set in the middle of a city.  Identify your systems early and bake them into the IP.  Consciously connect them!  Consider scale, need to identify that.

 

More vs Better – we need to explore “more” more!

Wow.  This was *easily* the best presented and most inspirational / thought-provoking talk I've been to at GDC this year (and maybe in a number of years).  Very good stuff, Clint said he's going to put the talk up on his blog later (you can see it in the blogroll to the left) and I recommend checking it out, though given his style the slides alone won't convey the story.

 

My interest level in Far Cry 2 shot up about 1000x after this talk, even though he went out of his way to not talk about the game itself.

 

---

 

 

i-fi: Immersive Fidelity in Games

Clint Hocking

 

Here to talk about immersion.  Why immersion?  Creative Director on Far Cry 2 but not really talking about that today.  3 yrs ago doing initial conception.  What would be the core of the game?  Identified a half-dozen possible pillars.  FORMID(able). 

 

Freedom of Gameplay

Open World

Realism

Meaning

Dynamism (drama management?)

Immersion

 

DRFMO instead?  Figured by now gold master would be out, easy to show immersion.  If you think you understand quantum mechanics, you missed something.  Same with immersion.  In addition to being complex, it is very divisive – smart people with conflicting opinions. 

 

Immersion: the basics.  Not a property of a game, but a player state.  Evoked in the player, not put in the game.  Similar to other states?  PlayerImmersed=1.  Doesn’t seem to have an analog range – immersed or not.  Not partially immersed any more than partially dead?  Rapidly switch in and out, which may feel that way.

 

So if it is a state what can draw him into that state?  Formal: left brain, Sensual: right brain.  Logical vs Sensory. 

 

Sensual Immersion, via the right brain.  Passionate, emotionally charged.  Hamlet on the Holodeck – 4 properties of digital envrionemtn “procedural, participatory, encyclopedic, spatial”.  Last two = immersion.  Painting immersion as a sensual experience.  Movie theater – sense deprivation and getting all our senses from a movie.  Any sense alone can maybe do this – master painting, miles davis, pure chocolate.  Sex – total immersion in the tactile.  Incense = smell immersion?  Well, look at dogs…

 

What are we trying to do?  Close the gap between the player, avatar, and game world.  Not applicable only to these kind of games but it gets cumbersome to use the more formal language and want to talk about concrete, not abstract.  Answer?  Trespasser!  The promise was to be the most immersive game ever made:  sprite impostors, normal and specular maps.  First game to offer many features that are standard today.  Needs driven AI (if barely functional).  Only recently achieved (Sims).  Robust physical simulation (Half Life 2 used a clever fiction to dispel need for arm). 

 

Drill in on two: IK driven animations, removal of all HUD elements.  IK was very ambitious for the time, trying to do something very ambitious.  Another decade before a game would succeed at this (AC).  Asserting their failure doesn’t enlighten us – look at why those decisions were made in the first place. 

 

Goals of animation system?  1:1 between player input and avatar response.  Higher fidelity than queueing up canned animation.  More procedural, so more reliant on inherent values of the mechanism.  Success of the Wii similar parallelism?  Canned response to motion feels bad, but direct motion correspondence causes immersion.  Trigger like buttons on the Xbox controllers to dominance of shooters.  Innovations like skate driven this way.  Guitar Hero using  Guitar Hero like controller.  GH on a pad is worse than even Simon.  Improve immediacy and responsiveness, and make 1:1 correspondence between avatar and player.  What about player and avatar connection to the game world itself?  Back on Chaos Theory, vivid memory of reaching out and taking hands in co-op.  Almost felt as if he had reached through, vivid feeling of physical presence.  This was with canned and cued animation, but maybe with a fully procedural solution we can make that work – hand always grab the hand.  Very human when achieved.  Increase the connectivity of the animation.

 

What about removal of the HUD?   A lot of that information is needed, like ammo.  Vocal feedback for state of ammo, health on a tattoo on Anne’s left breast.  Not just about the tits here – most games would show you those AND give a health bar.  What were they trying to do?  Every game needs to communicate state to the player.  Chess or Pac-man, you see all the pieces and have a precise game state.  Civ requires exploration and scouting, so making decisions about what you need to know now.  In a first person view we are taking the most limiting camera, and thus limiting that information.  So guys shouting out “reloading” when they are around the corner.  Makes it seem like enemy is communicating with allies, but really is saying “now charge me and kill me before I reload”.  Convert every meter in the game into something that is perceived directly by the player’s sense.  UI goes through a mediating filter of a HUD.  Direct experience facilitates sensory immersion. 

 

Left brain: Games vs Digital Environments.  Bioshock – compelling sense of place and graphics / audio / script.  More immersive than “Norbit” or other movies on the shelves.  While immersive it is a different kind of immersion – those qualities that are unique to games.  Games as film.  Immersion that is fundamental to a game, what does that even mean?

 

Formal Immersion.  Sensory immersion is a tool but it is not unique to game.  Is chess not immersive because it has no sound or smell?  You can play chess like a list of text of course, so it is not immersive?  Of course it is, perhaps even moreso than things that just require us to immerse our senses.  So what is going on with our brains?  They are sophisticated pattern matching machines, so good at it that they have constructed in their likeness another: the computer.  We just sort automatically.  Drawn in by looking through particular types of patterns.  Sorting “small things” is a few moments, but games are complex and require deep recursion.  Fractal leaves.  Uncharted new areas, exploring deeper in known games.  Exploring aimlessly is like doing math.  Playing a game with a goal is a process of seeking good patterns.  Good patterns persist, optimal patterns win.  Chess openings, these have been very heavily recursed.  Lets look at Guitar Hero, a simple game – clear power of recursive patterns. 

 

Three axes: implication, complexity, tolerance.  Implication is number of elements implicated in the pattern -- # of frets used.  Higher degree of implication = more frets = more difficulty.  Implication in chess – 4 for Anastasias Mate.  Complexity is what the player much be able to do with the implicated elements.  GH, 1 implicated element at a time at first, but over times more chords and sustains, moving up and down the fretboard, finally juxtaposition of rapid movement and plucking.  Complexity in chess – the comprehension of simultaneous offense and defensive nature of pieces. For each piece how is it implicated for you and opponent.  Tolerance – requirement of precision and accuracy.  Literally has a meter for this in GH – how fast it fills and drains.  Chess – the competence of your opponent.  Easy opponent is very tolerant for Kasparov, but Deep Blue was not.

 

So, if well used how does this create immersion?  Start on easy, seeing patterns of notes.  Variations on a pattern within a song.  Different themes and patterns within a level, but more or less the same level of implicity, complexity, and toleration.  Easy: slightly increasing complexity.  Medium: implication and tolerance immediately jump.  Continues recursively through hard and expert.  These well tuned forces acting in concert create immersion.  Unlike sensual immersion, this kind of immersion is universal to games.

 

Now what?  We have these two kinds of immersion.  Which is better?  Clearly they must fight!  Critics of sensual immersion: not fundamental to games, so at best a supporting aesthetic goal.  But total  sensual immersion is a critical convergence point?  Critics of formal immersion say that those patterns, while delightful, are not mass market.  Grandma may play GH at Thanksgiving but she doesn’t recurse it.  Love and fear and sacrifice not challenges of dexterity.  Rock god, not Simon expert.  The formal structure of GH doesn’t cause this, it is the skin and the controller.  Sensual immersion the domain of film?  If we want to compete for cultural dominance, don’t attack on their home turf.  Movies have a lead, so play to our own strengths?  Recursion through game systems have a divisive effect, leaving behind those who cannot keep up so as games become more complex they inherently become less accessible.  Imagine if every mystery movie you watched increased your ability to watch mystery movies, by today they would be incomprehensibly complex. 

 

The opposite of one profound truth may very well be another profound truth.  Quantum vs Relativity.  WTF? 

 

Reconciliation:  How can we get them to work together and “flirt”?  Mix those elements and make them more than reflex honing.  Chess vs MacBeth – similar structure, can we theme one up as the other?  Bioshock leverages its sensual immersion by daring us to rip out the little sisters hearts, while challenging us to recurse our patterns of saving them.  Recurse through love and sacrifice.  While experiencing love while immersed in both directions, we will be able to recurse through them.  What if instead of mystery complexity it was nuances of human emotion?  What if we trained not for playing notes but bringing people together?

 

Coming together of these two forms of immersion is the future.  All media can bring people together.  Each new medium can obliterate the dominance of previous forms.  We are not on the verge of edging out films.  We are just starting and when we finish we will obliterate them – games to film as films are to radio. 

 

Q:  People don’t “level up” in mystery.  But they do, difficult by your 20^th novel its tough for the author to trick you.  Yes, that’s an oversimplification. But they don’t get explosive the way you do when you recurse patterns.  You don’t get a little bit better at GH – Expert vs Easy is orders of magnitude more difficult.  Linear media we get better at by degrees because we don’t have that feedback loop.

 

Q: Voice in immersion?  AC: world is believable but spoken voice was disrupting.  Also relevant in narrative of SC (on blog).  Difficult for avatar to talk, speech is very specific unlike throwing a grenade.  When I say something I mean what I say, so if my avatar is speaking for me there is a big lack of connection.  One technique – in the X-Men movies Wolverine comes in and makes fun of them but in fact what he is doing is acting as the bridge of the audience into this movie, expressing their incredulity to ultimately accept this because we followed Wolverine’s journey.  Unfortunately can make your avatars cynical.  Silence maybe better to not break immersion.

 

Q:  Learning is a real barrier to immersion but those are not really mass media – millions don’t play chess.  But Chess is mass media – more people play it than GH.  More people play sports than videogames.  But those things are not about media, and those kinds of sensual immersion experiences.  Q: In reality, we don’t have to learn the details of physics to operate.  We need to leverage that sensual immersion but then recurse them – it’s not enough to just build them, we have to attach the warp drives as well.  Bioshock is the best example of this.

 

Q:  Is IK less of a 1:1 relationship because you’re not jumping?  Difficult choice to strip away a lot of the intent that players can express – pressing button to dive through a wall or off of a stall.  The problem with that approach is the recursive problem.  We started makng games where the controls are too complex, so many things you need to do that it becomes inaccessible.  By making movement where I want to go, but abstracting the jumps and leaps, you keep it abstract. 

 

Q:  Competitive MP can be very immersive.  If we want to enhance that we need to do that by displaying as much information as possible as accurately as possible.  Doesn’t losing the life bars hurt games like Street Fighter?  Philosophical answer:  MP and SP are two different problems.  In MP tolerance for not having that information is very very low.  In SP we have become addicted to giving more information than we need.  Information that is mostly junk.  258 bullets, or I just need to reload?  More interesting question is how do you ask when you *do* need to know?  What is that threshold?  Ways to give the player better sensory experience than relying on a HUD.   Consistency between MP and SP is always a challenge, hard thing to balance.

 

Q:  Intersection of formal and sensory immersion.  Started talking about love and doubt and sacrifice.  Is empathy a sense we can engage via immersion?  If we postulate synthetic emotion, then we can recurse.  What does that look like?  An example starting to build:  my girlfriends in GTA San Andreas.  Give mechanical bonus but there is no sensual immersion in that relationship.  Maybe Hot Coffee would have increased that immersiveness?  If managing that relationship as a running component of the game and sensually invested and can really “feel” that bond in a non vacant and stupid way (cinematic like currently), then we can explore it on a super-accelerated timeframe.

 

Q:  HUD as mediator of sensory immersion – isn’t an avatar this as well, since it isn’t you?  Avatar can be a way of physicalizing certain elements of the HUD.  The Tattoo was just copy and paste onto the avatar so it doesn’t really work.  Done in 3^rd person games frequently these days, seeing wounds and blood and stance.  Avatar itself can be a meter but if expressed physically can feel different.  Q:  does even presenting the avatar cause that disconnect?

Kurzweil is a great speaker, and I could tell that this talk really stirred up the audience.  It was mostly a rehash of The Singularity Is Near, however, which is a great book but a book I've already read.  That wasn't true for most people in the keynote audience, I think, so I hope maybe this sparked some thoughts on the matter.  I *highly* recommend the book if anyone reading this blog has not read it yet. 

 

Honestly, the talk had very little to do with gaming.  But it was still a good time.

 

---

 

Ray Kurzweil

The Next 20 Years of Gaming

 

Gaming fits in well with the acceleration of progress.  PS2 identified as a “weapon of mass destruction” because of equivalent to a super computer.  Music, movies, are becoming a computer industry.  The word “game” is unfortunate because it makes it sound like it’s not real, but there is real romance, real learning, real commerce there.  Artificial Intelligence and Virtual Reality are unfortunate names too.  Telephones are virtual reality, agreements you make over it aren’t just “virtual”.  Games are the cutting edge of what is happening and we are going to be spending more time in VR environments, and eventually they will be fully real environment.

 

What barriers to next step, fully immersive virtual realities?  If someone is a virtual reality they are going to lose track of real reality if they are moving around (Virtua Boy).  Seeing this already even just with the Wii where you can see real reality!    Maybe we’ll put real reality in a little window over there so you can keep track of it.  Ultimately going to be very competitive with real reality.  You can become someone else.

 

To see the next several decades, first need to look at the last several decades.  Wanted to be an inventor since age 5.  Key to being an inventor is to get things to work.  Even if you build 90% of successful devices, 90% will fail due to timing problems.  So these trends and evolution of technology is very important.  Principly use this time his own technology problems.  But with these models we can invent with the technologies of the future.  Isn’t the future unpredictable?  That’s true for specific projects.  Hard to predict Google stock or the #1 game or the next wireless standard.  But if you ask “what will the cost of a MIPS of computing be in 2010” or brain scanning in 2014 we can generate a figure that is very likely to be correct.  Age of Intelligent machines in the 1980 made hundreds of predictions that are tracking quite well – Arpanet doubling every year to become the internet.  That seemed ridiculous then but came right on schedule.  Same with super computer power and chess score.  This is a very democratizing technology – predicted soviet union swept away by decentralized technology.    Communication swept away totalitarian control.  Creating new games can be done on a $1000 laptop, or a full length movie with a $500 HD camera.  Tools of creativity are democratized, ultimately the tools of production will be as well with nanotechnology. 

 

Revolutionary.  Exponential, but also explosive.  Rate of exponential growth in processors is actually slowly growing.  1965 MIT was so advanced to actually have its own computer.  The computer in your cell phone today is a million times smaller and cheaper, 1000 times powerful.  Billion-fold increase in price-performance over the last 40 years.  We will see another billionfold increase in just the next 25 years.  Also shrinking them.  Today, a pea-sized computer in your brain for Parkinson’s.  Apply a billionfold increase to that, thinking about blood-cell sized devices.  The point is, it is predictable.  Very smooth progression from the 1890s.  Health and medicine didn’t used to be a computer technology, but now it is becoming one.  Used to be drugs were hit and miss drug trials, but now we have the software that life runs on and the tools to reprogram it.  We have software in our bodies that evolved 10,000 years ago.

 

For example, we have genes that hold on to fat, which was a great idea 10,000 years ago.  Ancient gene from animals roaming.  But that now causes an epidemic of obesity which leads to heart disease and diabetes.  What if we could turn this off?  RNA interference on animals to turn this gene off.  4 companies rushing to bring this to the market.  Current drugs work by inhibit appetite, this is like birth control by inhibiting interest in sex!  

 

Accelerating pace of change.  Last 50 yrs vs next 50 years.  Other people said next 50 like last – having fat drugs in 1 decade not 5.  Intuition is linear, reality is exponential.  Social security issues projecting using a linear life span.  That’s hardwired from our linear projection of life on the savannah.  Doesn’t work as well today projecting technology.  They are the same for a short period of time, which is why govt models appear to work OK for a year or two.  But now so fast that even a few years is becoming a big change.  Think back 6-7 years.  No search engines – that seems like ancient history!  3-4 years ago blogs, podcasts, MMO games popularity.  World changing dramatically quickly. 

 

Reading machine 30 years ago, very large. Someday a blind person will be able to take a device out of their pocket and read text in real time from the world.  In 2002, asked “when is this feasible”?  Models say requisite hardware in 2006, 2Q to be exact.  Software development?  More than just shrinking, need a new layer of software because holding the device – shadows, curvature, etc.  So how long?  4 years… lets get started!  Got started in 2002 for a device that wouldn’t be feasible until 2006.  Cumbersome, but worked.  Just recently introduced this in a cellphone, which is now a reading machine for the blind and of course all those other things that an advanced phone does.  5000 times smaller than 1979 device.  Other people seeing this is feasible now.  Cool demo of this actually working with synchronized highlighting.  Helps dyslexic kids as well.  Good example of timing projects.

 

Games take several years to develop, gaming tech completely different in 2-3 years.  Games are the harbinger of what we are doing now, creating real romance, real commerce, in them.  So what are the next 20 years of gaming?  See how pervasive and exponential these trends are – and will eventually affect our biology.  Industries like energy will become information technologies.  Plan with Larry Page of Google.  Nanoengineering is an information technology.  Today 85% of our energy from fossil fuels, as is solar – these are last century industrial technologies.  Already a next generation of these coming out, and in 5 yrs tipping point where energy will be cheaper per watt than oil and coal.  1 part in 10,000 of the suns light is 100% of our energy needs.  Already billions of dollars of investment in this.  Projections about 100 years running out of fossil fuels, but these are missing the exponential trends!  Solar needs only 7 more doublings, so within 20 years we will have replaced fossil fuels.  Predictable results from information technology. 

 

Doubling paradigm shift rate – 32x more progress in the next 50 years.   Took 50 years for phones to be adopted by ¼ of US pop.  Cell in 7.  Theory of evolution:  these processes accelerate because it develops a capability then adopts that capability.  Singularity chart of major tech progressions and paradigm adoption.  Life / DNA took a billion years to evolve.  Cambrian explosion went 100x faster.  Homo sapiens, a few 100,000 years.  Only 3 simple changes from our predecessors, only 10,000s of bytes  -- larger skull but weaker jaw, more brain in cerebral cortex for “what ifs”, opposable appendage that works well.  Chimps can’t quite do it the same way.  Can’t create tools well enough to use those tools to make new tools.  Use the latest tech to make the next tech.  Computers made with pen on paper, but today uses 12^th generation computer assisted design software.  Some disagreement, but not a lot happened in a million years a billion years ago, which clearly is not true today.

 

Exponential growth and linear very similar.  In fact can be sublinear at first, which can be discouraging.  Actually growth is overlap of S-curves, because old techs do run out of steam before they get taken over by the new growth.  1890 punch card machines up to cracking enigma code up to vacuum tubes.  Eventually couldn’t shrink vacuum tubes any more and that was the end – of vacuum tubes, not of computer tech.  On to circuits , which will end… 2022?  Won’t shrink any more and that will be the end of Moores law, but not of computing – 3d chips.  Now a lot of progress here, this is becoming a mainstream concept with multilayered circuits and self-assembling 3d nanocircuits.  Crossover in the teens?  This whole curve is trillionfold, and we’ll get another trillion in a lot less than 100 years.  Pong in 1972.  Even with world wars and depressions this is still quite predictable and smooth progression.

 

Singularity is Near, predicted 10^16 by 2016.  Brain is between 10^14 and 10^16 needed for sim.  Now several computers slated to do that by 2010.  By the 2020s most of the economy is information technology.  If that has 50% deflation rate like a technology, that will change economics.  18% growth in currency despite way more than that in IT growth.  People didn’t buy ipods for 10,000.  Mobile phones were bricks and meant power elite – 10 yrs first billion cell, 3 yrs next billion, 1 yr next billion.  DNA sequencing cost similar. $1000 genome a few years away.  Model, simulate, and reprogram our biological processes.  Cure for pulmonary hypertension.  Not just designer babies but designer baby boomers.

 

We will end up in the Uncanny Valley for a while – close enough to be creepy.  Going to seem like demented humans, hopefully we get past that stage quickly.  Key to human intelligence is language, the Turing Test is entirely language based.  Simulation and modeling the human brain – able to make verifiable predictions about audio processing from brain models.  Cerebellum models – solves many differential equations in parallel using basis functions subconsciously.  Simulating intelligence as applied to game characters?  Eventually we won’t be able to tell the difference.

 

1900s: 33% in farms and factories (each), 3% now – but that didn’t cause unemployment, it caused new jobs at the top of the skill tree.

 

Automatic translation in cell phones?

 

2010: Devices will disappear – into our belt buckles, displays in our eyeglasses. Images direct to retina.  Full immersion virtual reality games.  Very convincing illusions. 

 

2029: Intimate merger – turing test capable computers.  Compete reverse engineering of brain.  These will be integrated with us, integrated with our biological neurons.  Nonbiological intelligence will grow exponentially but biological intelligence is fixed.  15 years  from now, life expectancy growing more than a year per year?

Interesting talk from Soren Johnson.  We grapple with many of these same issues in designing the AI for the Age games so it is neat to see a similar perspective on a different (and more complex game).  It does remind me how many, many hours I've lost to Civ.... when's my Civ 5 coming, eh? 

 

---

 

Playing to Lose: AI in Civilization

Soren Johnson, EA

www.designer-notes.com

 

AI programmer for Civ 3, Civ 4, now working on Spore.

 

Good vs Fun AI.  Good = beat the player at own game, a human substitute.  Fun = algorithms are the content of the game.  Not necessarily looking to beat you, but rather to provide an interesting experience.  Good AI = Chess, Deep Blue.  Checkers is a “solved” game.  Oldschool wargaming often falls in this mold because it is sometimes hard to find human opponents.  Fun AI is different: Black and White, Sims.   Desktop Tower Defense – the game is about pathfinding.  Focus for Fun AI is on the player.

 

WoW:  Aggro is “Fun” AI.  Tanks / Healers / DPS.  Aggro controls AI – heal the tank, but not too much.  DPS the mob, but not too much.  Trivial AI problem to “solve”.  Aggro is puzzle, not intelligence?  WoW mods that show aggro.  Commodotized, like HP or some other core game mechanic.  Not trying to make it unpredictable. 

 

Most strategy games fit along this spectrum – Chess vs DTD.  Starcraft a bit closer to Chess.  Partly because it is a symmetrical game – all players trying to basically do the same thing under basically the same rules.  No real diplomacy.  Heroes of Might and Magic a little closer to the “fun” side of the scale.  A bit more asymmetrical, RPG aspects.   Civilization is right around the middle of the scale, which actually causes some problems.  Deep diplomacy, but crossed with symmetrical game design.

 

Rules over time:  Fixed vs Evolving.  DTD changing frequently, SC changing infrequently.  Symmetrical vs Asymmetrical.  Best suited for MP vs suited for SP.  Tactics available to the AI?  Everything vs intentionally limited.  Measuring performance:  objective vs subjective (winning vs having a good time).  Passing the Turing test:  “Good” = pass, “Fun” = irrelevant. 

 

Good AI = playing to win, Fun AI = playing to lose?  Where does Civ fit in that?  Win or Lose?   Mix of fixed and evolving rules.  Symmetrical.  Single player focused.  Limited options for the AI – a lot of tactics the AI simply won’t use, particularly in the “diplomacy problem”.  Common strategy for players is to take all the techs of a defeated AI then break your deal and crush them anyways, something that would be very frustrating for the AI to do to a human.  Objective testing?  A little unclear, but there are some basic objective measures.  Used automated testing during Civ development to see how the AI was performing.   Flat out fails the Turing test – not even trying.   So Civ is somewhere in the middle.  This is the art of writing strategy game AI, to find that best balance.

 

Lots of different players: Challenge, Sandbox, Narrative. 

 

Narrative players: aim for personality, maintain memory, fall for traps.  Tokugawa plays different from Washington, Gandhi vs Montezuma.  Personalities who are isolationist, even though this is clearly a suboptimal strategy.  It’s OK to make the player feel clever.  We want players to win – ideally right at the last minute.  But we’d rather err on the side of them winning.  If they do lose, they need to understand why very clearly.

 

Sandbox vs Challenge: Difficulty levels.  Sandbox players can get off easy, but how to ratchet up the challenge and include new tactics?  What about AI cheating?  Good AI should never cheat – who wants to play the chess AI that can teleport his pieces around.  Fun AI has irrelevant cheating.  You are playing the game in that world, it is just consistent in that way so not thinking about cheating.  So what happens in the middle? 

 

Civ 4 AI production modifiers:  +50% settler, to -40% deity.   Noble is even.  Then 5 / 10 / 15 / 20 as ascending the scale.  Noble does have other cheats though: animal  / barb combat, inflation, unit support, war weariness – the AI just needs help in these areas.   The player is in control of the game so they can quit / load, etc.  They do what they want and we are trying to help them have an interesting game.  That freedom is things where the AI can just never do – like leaving all their cities empty because they know they can reload or know they can trust their neighbors.  If the AI did this it would be a disaster.  So the balance needs to take that into account.   To make something where it is 100% even is just an academic exercise, it doesn’t make a better game for the player.  Cheats don’t need to be linear, sometimes the AI needs more help in other areas.

 

But it should never “feel” unfair.  Civ 1 and 2 had issues with this:  free Wonders, ganging up on humans.   Sid just wrote it that way, probably didn’t realize the level of scrutiny the game would come under given its success.  People know it was cheating, but this cheat felt wrong.  It’s not much different to say it is half cost since they can build wonders, but just getting something out of the blue is much worse.  Civ 1 literally always declared war on the human in 1900 if they were ahead. 

 

Civ 3 and 4 tried to put in human-blind diplomacy so that the human is treated the same as other players, same for testing the “value” of a tech.  There are information cheats in the game, however.  Basically comes down to limited AI development resources – fog of war is a difficult AI problem and the benefits didn’t justify it vs to a few information cheats.  These cheats can backfire though!  AI choosing cities to invade looking at defenders – and players realized that the AI would always invade at a lone undefended city, so they could “juggle” the AI.  So Civ 4 does some random updates of assault targets, and no longer looks at temporary data like nearby units since the player can manipulate it too easily.

 

The “Diplomacy” problem.  When is it ok for an AI to ally with another AI?  Why not just have them all gang up on the human?  Should the human be trusted / feared like AIs, even though they are less reliable?  Should the AI do anything to try and win? 

 

Cheating is relative – the “tech trading problem”.  1.  The AI must trade techs, 2. The AI must trade fairly, 3. The Human can trade techs cheaply.  But only 2 of these can be true at the same time.   Same as the “problem of evil” (theodicy) – if god is all powerful, and good, why does evil exist?  So should the AI sell techs cheaply?  Civ 3 probably went too far in this, too aggressively trading techs with other AIs and really trying to extract value out of its techs.  Made it feel like the game was rocketing forward too fast and that all the AIs were at the same tech levels due to arbitrage.   In Civ 4, AI can undersell by 33% but tries to make up difference in gold, and only trades with other AIs on random turn intervals.  Uses same “refuse trade with” logic as with a human.  Not necessarily solved the issue, but maybe closer?  Art not science.

 

Remember that the point of cheating is not to write the “best” AI or beat the human or be fair.  Cheating is for challenge, which is one source of fun. 

 

How do we design for the AI?  Design help for AI:  global unit support, closed borders, collateral damage, enforced peace treaties, maintenance vs infinite city spam.  But make sure not to design just for AI – all these decisions made sense for humans too, like really enjoying closed borders. 

 

Civ IV AI:  all civ AIs been written by the designers.  10 files (game, team, player, city, unit).  25,000 lines of code.  Soft-coded decision making.  Tiny slice of the overall code base.  Traditional testing fails, so need automated testing and hardcore fans to analyze – 1.5 yr closed beta (peaked at 100 users).  No Ai scripts (warrior, then temple, etc.).  No enums.  Everything is made generic.  Less brittle code, less predictable AI, works well with mods and expansions.

 

Probablisitc reasoning:  Assign weights to various factors, adjust to current situation, add some noise and then go.  Allows a lot of data driven mods: “Double Your Pleasure” for Civ 3 – tons of extra units and buildings.  Civ 4 Fall from Heaven. 

 

Game / AI code released on April 06, standalone project compiles into a DLL that replaces game code. 100% independent of engine, not even platform specific.  Very powerful, FFH added a whole magic system.  Extends game functionality.

 

Ai projects: community enhancement project, genetic AI project, Blake’s better AI (civ 4 better AI on source forge).  Eventually Blake actually joined to work on the Civ 4 expansion. 

 

Definitely the most technical talk I attended today (in a good way).  We had arrived at some similar conclusions about an approach for some skunkworks tech research inside Ensemble.  I was hoping for a little more Mercs 2 demo goodness, to be honest, but the talk was quite engaging on a difficult subject.

 

---

 

Mercenaries 2: Networked Physics in a Large Streaming World

Glenn Fiedler

 

Large world 8k x 8k.  Havok physics.  Every object destructible.  Emergent gameplay, not scripting.  Drop in co-op at any time (2 players).  Which has been quite a challenge!

 

Most network games today are static worlds.  Not a derogatory term, just analysis – static BSP with characters running around.  Mercs 2 is a step toward a dynamic world.  This requires a different networking model.

 

Trend towards this – more dynamic today than it was back in the 90s. 

 

Static world to establish context:

 

Input, state.  Determinism:  given same state + inputs, same simulation results.  Not necessarily hard determinism like in an RTS though.   

 

Pure client / server:  Client is dumb terminal, but input has to wait for latency.

 

Client side prediction:  Hide latency on the client – move fwd immediately without waiting for latency.  Server needs to be able to correct the client (accumulated errors, anti-cheat tech).  Correction is in the past, so how do we apply it?  Rewind and replay is one solution.

 

In a dynamic world – why do we have to change?  Player interacts with physically simulated objects.  It is a two-way interaction!  Player > objects > player.  This make things complicated!

 

Blocks… not too bad.  But then you have stacks!  Joints, linked bodies.  Not going to work if you apply those one at a time, need an integrated simulation. 

 

Client side prediction:  We still want this, but can’t afford to rewind & replay.  If you just did one object you would get horribly divergent, something way incorrect.  But what if we just didn’t apply corrections?  Limited client side authority.

 

Authority management:  Client takes authority over objects he interacts with.  These objects are client side predicted along with client player object.  Client no longer accepts server corrections for these objects.  Authority propagates.  Works really really well when two people aren’t both working on the same islands.

 

Client can turn white objects blue.  Client authority trumps default authority. Server can turn white and blue objects red – server authority trumps all.  Objects turn white again when they come to rest.  Player cubes are always colored.

 

Fixed delta time.  Variable delta gives no good way to explain exactly when to apply things.  Client runs ahead of server, delivers input/state early.  Apply state at a known time on a given machine.  Only stepping forward a multiple of integer frames (v-sync).  Everyone using 60hz sim internally.  Havok doesn’t let you move things independently, you have to sim the whole world at once.  Jitter is a little trickier, but you can filter to provide estimates.  Input should always extrapolate from known state on the other machine.  All clients ahead in time, but each can be differently ahead.  So when input arrives early it just leaves it in the input buffer.

 

Speed up / slow down time on clients relative to server.  Don’t actually scale time with fixed delta timestep.  Very subtle client effect.  Dynamically adjust to network conditions. 

 

To set physics state, snap it hard (to support physics stacks).  If you interpolate, the stack will fall.  Correct one meter to the left, you’d topple the stack unless you do this with a hard state change.  Visual smoothing on top. 

 

How to compress physics state?  Compress quaternion at reasonable precision.  Low precision vector, known radius, quantization.  Linear velocity / angular velocity, bounded in Havok and sent as compressed data.  Quantize on both sides though!  So clients have to apply the compression locally in their sim, not just in transmission.

 

How do we extend this to a large streaming world?  “Object scoping”.  Bubble around each player in Mercs 2.  Actually on each object.  Objects within this bubble wake up an simulate.  This is different than creating and deleting objects, just waking them up.  Separate treatment in Mercs 2 of dynamic vs pre-placed objects to help with this. 

 

Even with this scoping, you can’t fit all objects in your packet.  What are the N most important?  Standard object prioritization ( by distance, distributive priority – longer without update more important). 

 

Authority based priority?  Blue objects need to be sent from client to server, and the red ones need to go from server to client.

 

Conclusion:  client side predict with authority management.  Using scoping and authority management to determine objects to send. 

 

Corrections arrive in latent time, then transmitted into predictive time on the client.   If there is a conflict (a blue item that turned red), you see a latency-time pop / warp on the client. 

 

Q: How do you handle stacks of objects that are larger than the packet-limitation on objects?  In the future this scale won’t be a problem.  For now the mitigation is that the state updates go into predictive time.

 

Q: Security issues?  It’s a co-op game so we don’t care. 

 

Q:  How do you handle non-determinsm in Havok?  As long as it is roughly close enough, we don’t care.  A second of approximate determinism is all we care about.  Running whole simulation, just slamming in a subset of state for things we receive updates on.

 

Q.  Havok’s islands, or roll your own?  Demo just blasting all the state, can’t talk about M2 since still in development.

 

Objects that are rolling stay in authority because they don’t come to rest.  Weak authority where you aren’t directly touching objects.  These are some complications, but generally nothing smoothing can’t deal with.  Worst case you wind up snapping back.

 

Some things aren’t distributed – like scripting execution. 

 

No explicit state update when a client joins – the “washing machine” model, eventually objects cycle around.  In an initial streaming state wouldn’t let the client grab authority yet.