Here is a recent interesting article on Social Games:
I have not been able to play social games lately (like Restaurant City), although I strongly agree that the future of social games are very promising. The only reason I’m not playing those games is that I know I’ll spend a lot of time playing them if I start, and it will ruin my first year in the graduate school as a Ph.D student (still, I have to practice Starcraft 1, and prepare for Starcraft 2, and there are a lot of iPhone games… and NDS… Man, am I really in the graduate school?).
The article mentioned above is particularly interesting, because it mentioned Zynga uses “metric-based” design method. It did not talk about details of their method, but I’m pretty sure data analysis SHOULD play an important role in maintenance of social games. No game designer is so perfect that he/she can design a game with perfect balance. That is the reason of those now very usual alpha/beta testings. But not only in social games but also in every other online games, we accumulate data on user behavior 24 hours a day. Intuitively, we can make a good use of those data. There are a lot of good newses. There is no measurement error (although, there SHOULD be other types of error such as mistype / ill-classification) at all! It’s no sampling, but we get the data from the whole population! I don’t know how data analysis is used currently in the game industry, but I’m confident it’ll create a lot of value at least in the near future.
However, I’m curious what models and methods we should use for that kind of data. Maybe in the initial stage, certain intuitive basic statistics may work, but to use the data more effectively and to incorporate data analysis into the center of game design, I believe more sophisticated models and methods of Statistics may play some role. Because the data SHOULD be massive and constantly accumulated, however, I think it might be a challenging task, though.
I really want to work on a real project on those issues. It’s of no use to imagine how it can be used. Currently, beta testing of Starcraft 2 is going on. What methods do they use to monitor the balance between races? I’m not sure. But if we can make a good database of replays and also get APIs to access information on those replay files, I suppose we can do some interesting data analysis projects. I already have some idea on Starcraft 1… Hope I can show you some result in the summer vacation (until then, I have no energy to do so, unfortunately).