2.21.2007

Jonathan Blow Interview

This is an interview with Jonathan Blow, who has had my respect for a long time based on his work for the IndieGamingJam. He's about to release Braid. We're hoping that Braid will be to Prince of Persia what Portal is to Prey - plunging deep into an interesting mechanic where previously developers have only dipped their toes.

Up until now, details have been a bit thin on the ground surrounding the game. It's very interesting to hear the guy's creative process. He talks about a lot of issues which we can certainly relate to. In working on our current game, chaos has always been a big issue. We're trying to model something normally modelled with deterministic systems, which result in really pretty simulations, but which also very immediately explode into unpredictable possibilities. I've always called this "Perceivable Randomness" - when logical causal steps are obfuscated to the player, and it becomes difficult to understand how the system works, and therefore, how to manipulate it - like a magician hiding his tricks behind a cloth.

It's simply not very helpful for a player to be able to call on a system (deterministic or not) which he or she can't anticipate the result of, even after trying through trial and error (or even explicit guidance) to build a mental model of it. Where random elements are concerned, we can at best gage the typical deviation of an outcome from its mean, but only after many, many iterations (part of the reason random Rolling still exists in RPGs is because of this - we feel as though we're still trying to grok combat systems, even after the 500th WheltSnipe has been mashed. Hmm. I'm getting blog deja vu, so I'll get back on topic now).

We try to ensure that our system comes to rest quickly, and makes its immediate causal links obvious and easily predictable: a quality which Doug Church calls Perceivable Consequence (and according to Google, I am the world's most prolific misspeller of that phrase).

When it's just a few game elements you're dealing with, Perceivable Consequence is fairly easy to maintain (both cognitively and computationally) by simply not introducing random factors, or indiscrete values mapping onto wildly dissonant outcomes. Our game has hundreds (possibly thousands?) of elements all springing off each other in real time, which you'd think would cause massive causal explosions. However, everything is well "frictioned" so that we can bound our possibility space.And because friction is a passive property of the physics (resistant force increases with velocity), that bounding feeling doesn't feel sudden and arbitrary (like a glass wall). It's a rare case where technology (in this case, lots of multithreading) has enabled a game play mechanic. All the brute force physics we're using in our game is there to ensure we have a system which doesn't explode quite so readily - pulling something which could easily become perceivably random back into something with great perceivable consequence.

God, this is just too vague to be useful to you, isn't it? Point is, Jon is expressing a lot of things that we've come to learn during this development, so it's reassuring for us. Really looking forward to this.

1 comment:

Randal said...

This is awesome!