Georgia Tech researchers have developed an algorithm that uses big data in games to anticipate players' performance and customise difficulty levels accordingly.
"People come in playing video games with different skills, abilities, interests and even desires, which is very contrary to the way video games are built now with a 'one size fits many approach,'" said Mark Riedl, co-creator of the model.
The idea is to do away with gamers simply selecting difficulty levels or AI using rubber banding when things go wrong. Instead, the algorithm selects appropriate in-game events for the player that will bring their ability level in-line with what the developer wants.
"For those good at certain skills, the game can be tuned to their particular talents to provide the right challenge at the right time," says Riedl.
The model uses collaborative filtering, a technique used by the likes of Netflix and Amazon for making recommendations. It uses an established algorithm called tensor factorization to build up a picture of the gamer's ability level over time.
The model was put to work in a turn-based game, from which the developers pulled participant scores to predict how players with similar skillsets would develop over time. Alex Zook, a Ph.D. candidate working on the project, said they were able to predict players' performance with up to 90% accuracy using the model.
The researchers also see potential for the model to work in educational and training programs. Reidl said that they've been in touch with the army, working on ways to moderate the difficulty of 'virtual missions'.
The model was presented at the 8th AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment in Palo Alto, California.
You can't read more than an article and a half on a tech blog without seeing the phrase 'big data' these days, so it's no surprise to see it being used in games. This idea, of course, is that players don't end up walking away from games in frustration. Some gamers may object, however, to playing a game that is always being adjusted to be just the right side of doable.