Anson the Gnome

I missed attending PAX this year, because we've been pretty busy at work... perhaps that's a good thing. I've been hemming and hawing about the extent to which I want to participate at the conference in my own backyard, but now... well, a lot of people I missed seeing at PAX will be there. Possibly with the flu.

I will try to see at least a few people coming into town for AGDC, but preferably outside the context of the conference itself :-)

I know I'm not the best at maintaining a constant stream of new content up here, and I know that's important for really making this blog widely read. I try, but I also try to stay focused on interesting technical topics, and that just doesn't permit the kind of post volume required.

On days like this, though, I let out a sigh of relief: if this is the kind of interview I can expect from having a blog everyone knows about, perhaps I don't need that.

Through some roundabout way, I heard about this AI competition today; I can't tell exactly how long the competition has been going on, but according to the mailing list the cash prizes were announced today so that's probably why it showed up on my radar.

The basic gist, if you didn't follow (or have not yet followed) the link, is pretty simple: write an AI that plays Super Mario World-era Mario well, on random levels. Also, the cash prizes are restricted to conference attendees (the IEEE Games Innovation Conference in London, and the IEEE Symposium on Computational Intelligence and Games in Milan). So the contest aspect of it as a whole seems a bit uninteresting to me. Admittedly, it might be a bit more interesting to students who, cash prizes or no, would like to get their name out to the wider world.

On the other hand, it's also a platform for learning AI, regardless of contests or prizes: the game engine, the graphics, and simple examples already done and presented for you to tinker with. Whether your day job is in the game industry or completely outside it, AI is a hard thing to fall into on the job; this seems like a great opportunity to learn and demonstrate proficiency outside concerns of job performance. :-)

EDIT: Sorry, folks, I meant to save this as a draft to finish later and accidentally published it. Please re-read it now if you last saw it ending in an incomplete sentence :-)

Robin D. Laws, pen-and-paper game designer (including the cult classic Feng Shui), explains why his blog is an achingly slow source of of rules support for the games he has designed. While I was working on a live product I occasionally fielded support questions on my blog, but I saw and experienced the issues he talks about far more on the official forums. You don't have to look far in official game forums to find players asking why, even with developers participating, game design questions are answered infrequently and slowly.

Robin has the advantage of being the sole named designer for Feng Shui and HeroQuest, and he still can't answer these questions off the top of his head. When you're instead talking to a community manager, or a random non-designer (such as myself), or even the wrong designer, it gets even slower. On the bright side, when you're talking to a developer or community manager on the forums, or going through other official support channels... answering your questions is part of the job. The community manager might have tools to help track and manage questions to make sure they get answered, people in support definitely do, but for people like me it's a matter of personal discipline, memory, and note-taking to be able to get back to these a week or two (or more) after they're asked with answers.

I've written a bit here before about giving up on traditional relational databases for gathering, storing, and analyzing game metrics data. All three steps have serious scalability issues for the "one big machine" model.

However, SQL is often much more pleasant to write analytical queries in than, say, raw Java (or C++ or Python or...). So progress on a hybrid MapReduce/SQL system based on Hadoop and PostgreSQL is pretty interesting. I haven't had a chance to read too far into the announcements, or research the project itself, but it sure sounds interesting.