Use This Algorithm, Win 'StarCraft II'
Trick question: there is no single best strategy, but a new algorithm developed at North Carolina State University (NCSU) can examine the unique circumstances of your game and tell you exactly what you should focus on to maximize your chances of winning.
Computer science professor David Roberts and Ph. D. student Pu Yang analyzed thousands of professional-level matches in team-based, real-time strategy games like "StarCraft II," "Warcraft III" and "Defense of the Ancients."
From these game logs, the researchers came up with a complex set of rules and probabilities to help players develop the best possible strategies for any situation.
"Not all strategies work against all opponents," Roberts told Tom's Guide. "There's a lot going on in the environment that can contribute to the success or failure of the" match.
As professional strategy gamers know, certain strategies are more important at different times in the game. To help players determine which strategies they should employ at a given time, Roberts and Yang's program analyzes the various attributes by which a game judges progress, such as population, resources and attack power (the attributes vary from game to game). The program then tells players what level each attribute should be.
The rules generated by this program work on an "if-then" format. For example, if you're playing as X and your partner is playing as Y, then in the first fifteen minutes you should focus on A, in the second fifteen minutes on B and in the third fifteen minutes on C.
The program also tells users their likelihood of victory if the specified conditions are met.
For example, in a game of "StarCraft II," if you're playing as the Terrans and your teammate is playing as the Zerg, and you keep your population growth rate low while your teammate keeps population growth rate high, your chances of winning are over 70 percent.
"With 'Starcraft II,' the things that we found were most predictive [of victory] had to do with population growth rates," said Roberts. But "it very much depended on team composition as well."
The researchers found that in certain games and situations within the games, some attributes are more predictive of victory and therefore more important. To help players manage their time, the program tells players which attributes will most increase the likelihood of victory in a given situation.
The algorithm also shows that sometimes you want certain attributes to be lower than your opponents'. For example, in "Defense of the Ancients," more commonly called "DotA," if the player-characters on Team A have a combined damage-point total that's 59.7 points more than Team B's, the first team has an 80 percent chance of winning.
However, if Team A's damage-point advantage drops by just six points, to 54 more than Team B, then Team A's likelihood of victory plummets to 10 percent. In that case, Team A might want to switch strategies to something other than damage point growth.
This all might sound vague, but that's because there's no one "best strategy" — it all depends on what's happening in a given game. The program helps players juggle the many variables at play.
Roberts told Tom's Guide he's working on creating plugins for "StarCraft II," "Warcraft III" and "DotA" to help players develop the best possible strategies.
Right now, the plugin works, but it's not very easy to read and, as you can see from the above picture, takes up a large part of the screen. Roberts is now looking into usability testing to develop a user-friendly interface.
Roberts and Yang's full paper is available from NCSU's website. Professional and hardcore strategy gamers: What do you think? Do these strategies ring true for you?