What gamers want: New tool predicts player paths

One of the great things about social gaming, whether it’s role-playing on Twitter or playing an MMORPG (massively multiplayer online role-playing game) like World of Warcraft (WoW) or just traditional table-top gaming is the uncertainty. When other people are involved, you never know what they are going to do. You might make some good guesses, based on what they have done before, but there’s always that chance that they will surprise you.

WoW all the lonely people
Inhabitants of WoW

Now researchers from North Carolina State University (NC State) have developed a new method of predicting player behavior in online role-playing games.

“We are able to predict what a player in a game will do based on his or her behavior, with up to 80 percent accuracy,” said Brent Harrison, a Ph.D. student at NC State and co-author of the paper describing the research.

The implications are vast for a tool with such predictive accuracy. It could be used to help develop new game content or steer players to existing content that they will enjoy the most. It would also give game producers more confidence when expanding existing games or introducing new ones.

“A good game stands on its own,” explains Dr. David L. Roberts, an assistant professor of computer science at NC State and co-author of the paper. “If you want to improve it, you have to make sure players will like what you make. This research can help researchers get it right, because if you have a good idea what players like, you can make informed decisions about the kind of storylines and mechanics those players would like in the future.”

“[Y]ou could develop a program to steer players to relevant content,” Robert adds.  “Because it is data-driven modeling approach, it could be done on a grand scale with minimum input from game designers.”

In developing their predictive method, the researcher collected data on 14,000 players and the order in which they earned their achievement badges in the popular MMORPG World of Warcraft. The researchers then  identified the degree to which each individual achievement was correlated to every other achievement. The researchers used that data to identify groups of achievements − called cliques − that were closely related. Those cliques could then be used to predict future behavior.

Dice Copyright Dawn Turner

One of the more interesting insights yielded by the research is that highly correlated achievements or those that are part of the same clique,  do not, necessarily, have any obvious connection. For instance, an achievement in unarmed combat is highly correlated to the achievement badge associated with world travel despite the fact that there is not clear link between the two badges to the outside observer.

Will such insight make role-playing games more enjoyable, as the researchers posit, or have the opposite effect? The 20 percent error rate would seem to preclude the industry from eschewing the human creative factor entirely in the game development process, which is good news. Eighty percent still leaves plenty of room for rigidity, however.

Researchers will be presenting the paper Using Sequential Observations to Model and Predict Player Behavior at the Foundations of Digital Games Conference in 29 June through 1 July in Bordeaux, France.