Current Issue
This Month's Print Issue

Follow Fast Company

We’ll come to you.

2 minute read

Kicking Butt At Starcraft II, With Science

Researchers at North Carolina State University are using machine learning to decode the optimal strategies for multiplayer games.

Kicking Butt At Starcraft II, With Science

A team of researchers at North Carolina State University loves Starcraft II and Defense of the Ancients (DotA) just as much as the rest of us do. So much so, in fact, that they published an academic paper on how to create the perfect strategy for multiplayer games. Authors Pu Yang and David L. Roberts wrote the paper, "Knowledge Discovery for Characterizing Team Success or Failure in (A)RTS Games," and claim they’ve found clear patterns players can use to increase their score.

"Our goal is to use these data to develop tools that could train game players to play more successfully," Roberts said in a release. "These tools could be incorporated into games by game developers, or could be developed into stand-alone training modules. (They) could also be used by game developers to help them understand whether the game mechanics they are putting in their games are having the desired effect, and to fine tune their games accordingly."

Starcraft II

Yang’s and Roberts’s study used thousands of game logs from Starcraft, Defense of the Ancients, and Warcraft III that were then parsed through "machine-learning-style evaluations" to find patterns related to the success of multiplayer teams. The study was interesting to read, with lines like "Gankers do not invest resources to develop their Strength attribute, they invest resources to develop their Intelligence attribute by which they use magic to stop opponents from farming resources and to enhance their teammates’ farming, especially resource-hungry Carries. Moreover, they save strength resources which are less important to the Ganker for a team win and provide opportunities (farming lanes) to other teammates." By generating decision trees and generating statistical analyses, the duo were able to break down a series of distinct strategy styles for multiplayer games.

DotA 2

A big part of their research consists of the idea that multiplayer games have coherent rules, structure, and internal logic. In a previous paper, the pair compared the internal knowledge rules of multiplayer games to similar patterns in hurricanes and international currency exchanges.

For gamers, there’s good and bad news. The good news is that Roberts and Yang were able to find clear statistical patterns in the way teams approach playing Starcraft and DotA. However, no tools to improve team performance have actually come out of the study yet. Their study was publicly presented earlier this month at the IEEE 2013 Conference on Computational Intelligence in Games. In the paper’s conclusions, they argue that these statistics-based analyses can help develop new strategies for gamers to approach multi-player games. Next up? A plug-in for these multiplayer games which takes game logs and offers teams instant advice based on machine learning.

[Image: Flickr users Sebastian Anthony, and SobControllers]