Future Video Games May Tailor Themselves to Players' Desires
CREDIT: Activision Blizzard
Some video gamers inevitably favor epic adventures of taking down huge dragons, even as others enjoy crafting and selling goods through in-game auction equivalents of eBay. In the future, games may take notice of such player behaviors and change their storylines or missions to offer a more personalized experience.
Those future games could rely upon the past behaviors of players to predict their future interests. A new model has succeeded in doing just that by predicting player actions 70 to 80 percent of the time in the online fantasy game "World of Warcraft." The early success hints at how game developers could tailor virtual experiences on the fly to what gamers want.
"They could dynamically adjust a game to make it more or less specific to a player type," said David Roberts, a computer scientist at North Carolina State University.
Some games already adjust their difficulty based on the number of players or how powerful a player's character has become. The game "L.A. Noire" even allows players to skip combat or chase scenes if they prefer to focus on sleuthing and interrogating suspects. But future games might make that decision automatically.
What gamers want
One huge challenge of tracking player behaviors to predict future actions comes from the massive amount of game data involved. To get around that, Roberts developed a two-step algorithm approach with Brent Harrison, a Ph.D. student, to make the number-crunching much more manageable for computers.
To test it out, they turned to an online database called the "World of Warcraft Armory" that lists game achievements accomplished by individual players. Such game achievements can range from catching a rare fish called "Old Crafty" to completing 200 "Battleground" matches involving massive fights against other players.
[Read More: Video Games May Shield Soldiers During Nightmares]
First, the algorithm looked at one random group of 7,000 "World of Warcraft" players to find out which players tended to accomplish certain achievements. Related achievements were organized in groups as large as 80 achievements, and became the foundation for the algorithm to create behavioral models of different types of players.
Second, the algorithm used the behavioral models to predict the achievements of a second random group of 7,000 players.
Not all the predictions were obvious. One popular belief is that players fall into two groups; those who enjoy fighting other players, and those who enjoy doing quests or high-level dungeons where they fight artificial intelligence (AI) enemies. But researchers found that players who completed Battleground achievements also completed high-level dungeons where as many as 40 players cooperate to take down powerful AI monsters.
From game to game
The achievements may not perfectly reflect player desires; some players choose to do something just because they earn the achievement, rather than out of enjoyment. But the "World of Warcraft" test suggests the algorithm's predictive power should work equally well with more relevant data, if game companies would open their data up to researchers.
"If we had access to enjoyment data, we could certainly run the model on that," Roberts told InnovationNewsDaily. "That information is just not available, but there's nothing in our approach that would prevent us from targeting enjoyment specifically."
The researchers presented their work at the Foundations of Digital Games Conference held at Bordeaux, France from June 29 to July 1. But they already have future plans to try predicting the behavior of online shoppers engaged in the virtual economy.
As for video games, Roberts envisions a future where gamer preferences might even transfer between different games, if game developers can agree on allowing such data to shape the flow of their individual games. Some games, such as those in the sci-fi franchise "Mass Effect," already carry over data about player actions from previous games in the series.
That project could take shape over the next three or five years.
"You could imagine having a personalized content model of your particular style of play that follows you from game to game," Roberts said.
You can follow InnovationNewsDaily senior writer Jeremy Hsu on Twitter @ScienceHsu. Follow InnovationNewsDaily on Twitter @News_Innovation, or on Facebook.
10 Sci-Fi Predictions That Came True
Sex in the Digital Age: Just One Human Required
9 Cyborg Enhancements Available Right Now