So for some people, these two talks might come out of the blue. Both of them were in tough slots, and both of them were in small rooms. One was about level design and one was about writing. Two topics that speakers have a hard time impressing me with. But as was the case with most of GDC this year the talks surprised me with how great they were.

Susan O’Connor on Storytelling Challenges in Gears of War

The only thing I didn’t like about this talk was the part of the title I omitted from above. Grrrr. I don’t even care if a literal Hero with a Thousand Faces was central to the talk… simply reading those words makes me wanna punch someone. Campbell’s book seems to have as firm a stranglehold on game writing as Tolkien’s books have on game design. I’m fucking sick of magic elves, and I’m fucking sick of Campbell. Oops – rant over… back to the talk.

Actually, I don’t have a lot to say in particular about the talk, other than to point out that it was illustrative of the fact that O’Connor is in the camp of game writers who ‘get it’. Writing for games is not about having better stories or plots. It’s not about having better, less cliched characters and dialogue. Yes we need all of that too – but it’s not about that. Writing for games is not even really about writing at all. It’s about mechanics. More specifically, game writing is structuralist and the structure of a game story is malleable.

Susan gets that, and was able to address a handful of writing challenges she confronted in Gears where her awareness of this fact clearly saved what was otherwise a collection of cliche characters in a cliche sort of story from falling totally flat. I’ll be the first to admit that Gears – for all of its cliches, rises above them in many places. Obviously design-wise, artistically, graphically and visually it is worlds better than any of its competition and the same goes for its story, largely because O’Connor was clearly able to identify levers in the structure that she could pull to get me to laugh or smile or growl along with or in juxtaposition to my avatar in (mostly) the right ways.

Brian Upton on Narrative Landscapes

Talk about a tough slot… last hour of the last day. I almost didn’t even go to this for a number of reasons; I was exhausted and it was ‘another level design talk’ – I have not had much luck with level design talks at GDC. But Upton was lead designer on the first two Rainbow Six games, which I played the shit out of back in the late 90’s, and the R6 games sure had unusual level design. I’ve also had the luxury of seeing that level design sensibility evolve over the years with countless of my co-workers working on different Rainbow titles down the hall from me, so I figured I’d give it a shot. It was great. If anyone can pinpoint Brian’s slides, let me know.

The heart of Upton’s talk was an analysis of Kevin Lynch’s seminal treatise on urban planning, The Image of the City (which I have yet to read). Specifically, Upton contextualized Lynch’s concepts of Paths, Edges, Nodes, Districts and Landmarks to make them relevant to level design, then looked at the inherent dramatic and mechanical values of these concepts. He showed how we can use Lynch’s ideas to conceptualize our designed spaces (whether interior or exterior) to get the player to experience specific feelings when encountering these spaces. The presentation was long on details and examples, and I’m not going to throw up my lenghty notes here, other than to just comment on one.

He used the example of doorways in Rainbow Six as Edges, and pointed out how Edge transitions are full of dramatic tension. Clearly that’s the case in every Rainbow game from the original to Vegas. I did wonder (and I asked the question after the presentation) whether Upton felt that with improvements in AI and changes in mechanics and in overall focus over the course of the Rainbow series of games had altered the impact of the Edge transitions at doors in Rainbow.

In early Rainbow games, the teammate AI was particularly vulnerable at doors. Partly this was ‘realism’ – from the folks I’ve spoken to who know about this stuff, a huge percentage of casualties in CQB situations occur at a doorway – but partly this was also an AI failing – getting four AI characters to move through a doorway and cover their angles correctly, with all of the animation, navigation and synchronization required is a very, very hard AI problem.

In playing the original Rainbow Six, doors were certainly dramatic… but they were also potentially very frustrating because a minor AI failure could turn six guys into six corpses in point-six seconds. Any time I took a team through a door in the original R6, it was a gamble that I would fail the mission. In the intervening years, and over the course of several titles, the problem of having AI breach a door effectively seems to be completely solved. At the same time, however, the designers of more recent Rainbow games have also changed the way the games’ encounters tend to unfold. They have de-emphasized the ‘binary’, or ‘life-or-death’ nature of doors.

Breaching a room successfully in the early Rainbow games tended to end with a pile of dead terrorists. Failing tended to end with a pile of dead Rainbow and a reload. Now, rooms in Rainbow tend to be much more analogue. The ‘breaching’ of the room is just the first ‘act’ – if you want – in each room, the ‘clearing’ of the room is the second act. Often there are (scripted) reinforcements, or enemies hidden out of line of sight of the door or doors who then engage the team once they have entered the room and gotten to cover. Failing to breach perfectly generally means a complication pushed to the second act of the fight.

My question for Upton was whether he felt this was more or less dramatic, or whether it was a better use of Edges (and/or Nodes) than in the original highly ‘edgey’ doorways of earlier Rainbows. I think, technically, the dramatic punch of the edges in the early Rainbow games may have got more bang-for-the-buck so to speak. While the conditions were binary, and the challenges were often frustrating, there was really clear edge-based drama at every door. In recent Rainbow games it’s a lot more complex. In terms of what is fun and what fits the context of the overall play experience, the edge/doorway challenges of more recent Rainbow games are doing exactly what they should. At the same time, I think you could say that the pure dramatic intensity and the clear design focus on edges in early Rainbow games was exceptionally well realized and worth learning from. Very probably this clarity of design vision (in regards to doorways and other factors as well) is what made the original R6 games so successful, and is what still inspires smart designers to continually try to wrestle these hard problems ot the ground.

Anyway, that wraps it up for my two ‘second favorite’ presentations of this GDC. In a day or two I’ll post my favorites – which I think are probably pretty predictable. If I have time befre I jet off to Paris next week, I’ll try and fire up a sort of GDC wrap-up as well.

So, I wanted to post some thoughts about my favorite GDC sessions and comment on some of the takeway. Unfortunately, this was such an excellent GDC, and I tend to post huge tracts anyway, that I think I had best not do it all in one post.

There are six sessions I wanna talk about in particular, so I’ll do them two at a time, starting with my two ‘third favorite’ talks.

Chaim Gingold on Spore’s Magic Crayon’s

I thought Chaim’s talk was engaging and thought provoking. Even better is that I’m not really entirely sure if I agree with the underlying principles. Basically – and maybe I’m getting it wrong here – but Chaim was saying we can use the computer to help the player make better decisions about (for example) where to put a foot. The reasoning is that it’s better to have most people make pretty awesome creatures (or spaceships or buildings or whatever) than it is to have almost everyone be able to make shitty creatures, while an elite few make really awesome creatures. Thus, we give the computer rules that – for example – make sure all feet attach to the ground, or that eyes have bi-lateral symmetry or whatever.

Cool. Smart. Clearly the right way to do it. But at the same time, isn’t it kind of like saying ‘we should not allow the player to delete the door to the bathroom while his Sim is inside’? It seems to be a small but important step back from what I would expect from Maxis which would be – ‘hey if the player wants to put the creature’s feet on its head – why not?’

To me it’s kind of like the old Lego ‘problem’. Lots of kids who play with Lego make spaceships. So sometime in the ’80s Lego made space Lego, which to a 9 year-old me seemed like the best invention since – well – since ‘plain’ Lego. Lego said ‘kids wanna build spaceships, but our vanilla Lego bricks only do a mediocre job of that – let’s make space bricks in blue and grey and black with lots of angles and cones’.

In effect, space Lego allowed more people (including me) to do a better job of the thing they were trying to do anyway, which was build spaceships. And that is exactly what I took Chaim’s talk to be about. But in my mind, there is a cost associated with this. Call it a lost opportunity cost. Call it pandering to the lowest common denominator. I don’t know. The point is that a little bit of imagination is killed when the step from ‘pile of Lego’ to ‘spaceship’ is short-looped with more affordant Lego… maybe.

It’s possible – I suppose – that the imagination isn’t ‘lost’ it’s just spend solving different, higher order problems about making ‘awesome spaceships’ instead of solving the more trivial and less interesting (and less entertaining) problems of just getting something that feels like a spaceship. I’m not sure. Part of me wants to believe that’s true, but another part of me looks at Lego today and says – shit, Lego today sure looks a lot easier and a lot less fun than it did when I was a kid. Maybe I’m wrong. Maybe I’m a cynic.

Anyway – the talk was brilliant and I learned a lot from it. Questions of this sort are overly harsh and nit-picky. Mostly I think the way they’re solving the (very hard) sorts of problems that Spore seems full of looks pretty fucking ingenius to me. So don’t let me get in the way.

The Game Design Challenge – The Needle and Thread Interface

So ‘tied’ with Chaim’s talk was Eric Zimmerman’s Game Design Challenge. Having been up there before, I know for real how brutally hardcore this challenge is, and to be completely honest, in my opinion this was the hardest one yet – I think it may actually just have been unreasonably hard. Not to be all flakey and waffly about it, but I really think all three of the contestants did an admirable job under what seemed the direst of circumstances. The most important take-away for me was after David and Alexey had both gone, and before Harvey presented, Eric pointed out that seeing the differences of approach between Alexey’s drive for elegance and David’s ‘guts+fun mechanics’ approach was worth the show right there. And it was no surprise after that statement that Harvey’s solution was clever and strove to say something meaningful.

While Alexey won, I think he had the +2 Fan Favorite Bonus on his roll. His design was clearly elegant, and he was obviously handicapped with the language and powerpoint barriers – it’s possible that Stitch and Cross would be the most compelling game in a decade – it was hard to tell. I did really like his observation that because the moves were stitched, that every second move was a hidden move. He used the mechanics of the interface to generate the classical game theoretical concept of hidden information. Crafty.

As far as demonstrating the most robust understanding of the domain of the challenge, I actually think David Jaffe should have won it. I thought the ‘paper airplane game’ part of the design was weak, but his understanding of and utilization of the interface was the strongest. Things like detecting the topology of a fold in the fabric by passing the needle through two points simultaneously – fuck that was brilliant. Detecting bad stitches and good stitches to determine the mechanical strength of attachments to virtual hardpoints was also brilliant, though it’s a clear example of what I mean when I say that, while his demonstrated understanding of the interface was stunning, his resultant design was a little weak and forced. I don’t particularly think it would be good design to make the player really spend hours making strong stiches in order to have his virtual plane have more hit points.

Harvey’s game was the most believably possible as an actual game in my opinion, but I felt he overreached a bit by trying to generalize the controller to all games, and ended up with what was mostly a remapping of conventional controls to something kind of like a touch-pad with some added functionality. Some of the unique gameplay elements he designed to more robustly support the controller (like stitching bits of fabric to your doll to give it new powers) were great.

Anyway, like I said, the Challenge was (possibly) unreasonably hard this year. Keep in mind that when I did it, the driving focus of my talk was about hunting for good constraints because I found the challenge too open in many ways. That clearly did not seem to be the case at all this year and I think all three of these fucking geniuses had to wrestle a little freedom out of a very harsh constraint.

Tomorrow or the next day I’ll try and get around to posting my two ‘second favorite’ sessions. I think both of them will be surprising.

The slides and the accompanying paper for my GDC presentation are up for anyone interested.

There were three questions people asked after the presentation that I think really contributed to the talk and enhanced even my own understanding of the topic (there were other very good questions too, but these three were the ones I really remembered).

First, someone who I don’t know asked how these ideas applied to multiplayer games. This is a great question, and I had intentionally left it unanswered in the presentation because I don’t play enough MP games to really consider myself qualified to tackle the problem. But I did have the opportunity to notice while I was working in the presentation that there is a certain useful parallelism between the kinds of exploration I was talking about and Nicole Lazarro’s Four Kinds of Fun.

Rougly speaking, I think we can say that System Exploration maps fairly well to Lazarro’s idea of Hard Fun – fun that is about optimization and winning. Spatial Exploration could be said to map to Lazarro’s Easy Fun, which is a playful fun emerging from rewarded curiousity. The analogy is weaker here – but in some ways, my notion of Self-Exploration has some overlap with her ideas of Serious Fun (though Self Exploration as I talk about it seems kind of orthogonal to her ‘plane of fun’ – maybe echoing back to what I was talking about at Futureplay with a new axis of ‘meaning’.)

Anyway if we accept this rough (though admittedly incomplete) mapping, it becomes clearer than ever that there is a fourth kind of exploration that has to do with exploring interpersonal relationships in MP games. This is obviously what Lazarro calls People Fun and what I would call ‘Social Exploration’.

Again – I don’t play enough MP games to really examine this ‘kind’ of exploration. Maybe someone who really gets MP games can use what I’ve done here as a jumping off point to talk about that fourth kind. Ironically, it would also probably be a lot more useful to developers than this stuff, because so many more MP games than SP games seem to offer rich opportunities for exploration. 

The second awesome question came from Frank Lantz, who caught the sleight-of-hand by asking if in fact Self Exploration was not just another kind of System Exploration. Frank is right – it is. In fact, in a game, everything is System Exploration. Even the representational 3d space of a game is still a system. If I’m playing Oblivion and trying to get to the top of some mountain but the max climbing angle is 50 degrees and I hit a fifty-one degree slope, my brain immediately stops seeing the world as a space, and starts seeing it as a system. I start wiggling the stick and jumping and otherwise trying to hit that one triangle in the terrain that is at 50 degrees just as I tap jump so I can launch myself over that 51 degree triangle and continue upward. I did a pretty bad job of answering Frank’s question, but the next question kind of helped.

The third question came from Jesper Juul, who was also my teammate in The Metagame. Jesper asked – well actually, he more kind of pointed out – that while everything in a game ultimately distills down to being a system, and all kinds of exploration are really by defintion (or at least by the definition I was using) another form of System Exploration, the reality is that we (as in us pathetic humans) have a whole shitload of hardware (brains) devoted to chunking our perceptions up to look like spaces.

That is why it is easy for us to make a conceptual difference between systems and spaces, and harder for us to see my so-called ‘self-exploration’ as anything other than just another system. But really – space itself is just a system. It happens to be a system that is best perceived at a higher level – as a bunch of solids and gaps, rather than as a bunch of interacting particles and energy responding to a few fundamental forces. If we perceived space as pure physical interaction instead of perceiving it through a kind of higher-order lens, we might be distracted by the noise in the system and not see that the fascinating cluster of organic molecules over there is a hungry tiger about to eat us. In other words – while Frank is right, it all distills down to systems, Jesper was savvy enough to remind us not to lose track of the forest in the trees.

The real question that emerges from my presentation and from those two super-smart comments is ‘how do we really make a forest of self-exploration’. We need to do a much better job of it, because right now we only have a few sickly little trees.

Finally, Chris Butcher nailed me with the dreaded ‘where’s the feedback’ question. That’s a hard one and I know it’s a weakness. It’s probably the biggest weakness of the presentation and the biggest hole in my own thinking on the topic. The best answer I have right now feels incomplete and not very rigorous, but it has something to do with leveraging Wright’s so-called ‘Second Processor’ (the brain of the player). Basically my response to Chris’s question goes something like this:

The game does provide certain elements of low-level feedback – of course. We need to have the animations of a disappointed AI, and his accompanying bark to tell us he thinks we have done a bad thing (or whatever). That’s self-evident (it’s also insanely expensive as the possiblility space around so-called self-exploration gets richer – but that’s another topic). The other part of the answer is that the objective in some ways is not to have the game tell the player explicitly ‘you are bad’ or ‘you are kind’, but to have the game present the output of the interaction such that the player tells himself ‘I did good’ or ‘I could have done better’. It’s about making the game more reflective.

We do this in systems all the time. When I play a racing game, the game tells me I get Gold for 01:00, Silver for 01:10, and Bronze for 01:25. At the end of the race, the game reflects my system exploration back at me by telling me I came in at 01:07.78. Now I say ‘I could have done better’.

But what is important in this example is not that I know I could have done better overall – but that I know I could have beat that one car at that second turn if I had started the turn from two meters further out, and passed him on the outside instead of trying to squeeze past him on the inside when I grinded along the barrier. Yeah there was lots of low-level feedback in the second-to-second when was trying to make that turn – tons of sound and sparks and intensity everywhere – but it wasn’t until I stopped to think about what I had done that the game became reflective and I was able to compare myself as I was to myself as I had hoped to be. The processor that’s providing the feedback here is not in the machine – it’s in the player. Again – I think while interesting, the formulation of this idea is incomplete and it is certainly absent in my talk. This is something we need to figure out.

Anyway – thanks to the folks above and to all the other smart people who nailed me with hard questions. I hope there was material in the presentation that will help you with whatever you’re working on, and I hope the slides and paper over on the right will be of some value to you.

Just got out of bed from a nap after my red-eye back from San Francisco and thought I’d fire up my first few thoughts about GDC this year. I’ll probably do a few posts about GDC in the coming days, as well as find a chance to put up the slides for my talk, though I won’t get to that today (Eric-Jon Waugh covers it on Gamasutra here if you want a summary that includes some spoilers).

Overall, this was the best GDC for me in 2-3 years (though I always love it). That’s especially impressive when I consider the problems that needed to be overcome. For one, I prefer having the conference in San Jose. Not because I have some love for the city or anything, but really, since I spend the entire time at the conference, the city that it’s in is irrelevant to me. It could be on the moon for all I care. So while I hear a lot people say ‘San Fran is so much nicer’ – well, that’s totally true… but how much of San Fran did those people see? I didn’t see any of it.

It’s better in San Jose because people scatter less, and it’s easier to find or run into people who you want to talk to. My impression of San Fran is that most people disappear at 5:00 back to their rooms to freshen up, get changed and hook up with their co-workers, but because they are so much more scattered, and because there is no central ‘hub’ (like the Fairmont) then they end up hanging out with their coworkers and networking less. I certainly found I had to go out of my way to track down certain people in particular to find them this year, and I came back with about 20 business cards, instead of the ususal 100. So I definitely meet fewer people.

The second thing it had working against it was the simple fact that I had a crappy hotel room way down on Fisherman’s Wharf, so instead of havign a 10 minute walk each day to the Moscone Center, I had a 15 minute, $10 cab ride. Suck suck suck. Oh well.

The third thing it had working against it was the fact that I am now ten days from a major milestone, and my entire team was back here crunching on my – still unnannounced – game. More than anything I would rather have been in the office nailing down the last dozen things that need to be fixed before we hit the deadline, but such is the difficulty of getting your work schedule in sync with GDC.

The good news is – as I said – even with these things working against it, it was the best GDC I’d been to in a few years, and the reason for that simply was the quality of the sessions. Except for one or two things, the sessions I saw this year were uniformly top-notch, and I learned a lot of stuff. For me, the best part about a good session is not (usually) that it opens my eyes to something that I never thought of – but rather that it fills in a lot of blanks about stuff that I had thought of, but had not had a chance to think of in any detail. In a twenty minute talk, a guy like Randy Smith can kind of upload two months of research and thought on save-load system design into my brain and effectively save me two months of work. Now I have a working understanding of a complex problem that I always wanted to tackle but never had the time to, and because it comes from a highly trusted source, I don’t need to question the quality of the thinking that generated the conclusions I can start from. In fact, I’ll likely question those conclusions less than I would if I was the one making them (how’s that for pressure Randy?)

Anyway, more details about specific sessions – as well as my slides and presentation materials – to come soon.