Leave it to higher education to devote quality time to research that will actually help gamers become better players in games like Dota 2, StarCraft II and League of Legends. Computer science researchers from North Carolina State University have developed a technique to determine which strategies give players an edge at winning in real-time strategy games like Defense of the Ancients (Dota), Warcraft III and StarCraft II. The technique offers extremely precise information about how a player’s actions affect a team’s chances of winning, and could be used to develop technology for use by players and developers to improve gameplay experiences. Check out the papers here and here.
“We sincerely hope that pro gamers will find our work beneficial, and enable to them to gain insights into their play that makes them more competitive,” said Dr. David L. Roberts, an assistant professor of computer science at NC State and co-author of two papers on the research. “One of the exciting things about our results is that they can also be helpful to the average gamer, especially once we’ve made more progress on the visualization tools we’re working on.”
Researchers used the technique, which makes use of various analytic tools, to evaluate logs of player actions from thousands of ARTS games. They then used that information to develop a set of rules governing team gameplay strategies, in order to identify which approaches give teams the best chance of winning. Although League of Legends wasn’t part of this research, Roberts said this methodology can be applied in any setting where there are time-evolving attributes that describe progress in the game. The quality of the insights will depend on the mechanics of each individual game.
“Real-time feedback about what players should focus on to increase their chances of success can help gamers learn more effective strategies,” said Roberts. “It will enable them to learn about how their approach to game play affects their progress, and to identify new goals for increasing their odds of winning.”
Researchers focused on Dota for three main reasons: 1) Being a multiplayer real-time game, Dota gameplay is an example of the types of behaviors they were interested in examining, 2) Dota replay log files are readily available on the internet, making it feasible for them to get enough data for computational inquiry, and 3) Dota is very popular, which they hoped would make these results interesting to a large audience. The other games were selected to compliment the characteristics of Dota and show the applicability of the technique.
“In short, these games are extremely complex,” said Roberts. “Players are making 10s or 100s of decisions per minute (depending on the level at which you look), and it can be exceedingly difficult to do ‘temporal credit assignment.’ How is a player supposed to know that purchasing an item 12 minutes into the game ultimately sends them down the path to failure 30 minutes later? The types of insights we can now provide will hopefully enable players to get a better understanding of the relationships between their goals and their success.”
When you add in the thought processes of human opponents, sometimes as many as five per team, these games become very even more complex. ESports adds an extra layer of drama with millions of people watching through livestreams and thousands of live spectators taking in the virtual action in huge venues.
“Modeling the human mind explicitly is just not feasible in such a complex scenario, so our techniques handle the human mind through data,” explained Roberts. “By collecting a large number of replays of games, we get examples of the range of things that human players can do, and we use machine learning techniques to identify and leverage any subtle patterns.”
Roberts sees this research not only benefiting gamers of all skill levels, but the very developers of MOBAs and other games. Game developers are constantly tweaking the mechanics of their games (e.g., the rules that govern interactions) in an effort to promote a certain gameplay experience. For example, the way that scores are calculated in Scrabble are a game mechanic. The choice to locate double and triple word score square where they are is one way of tweaking the mechanic.
“It can be very difficult to understand the relationship between game mechanics and gameplay experience, especially in complex games,” said Roberts. “Techniques, like the ones we developed, can help give developers insights into the relationship between mechanics and game play. So, in the context of Dota, for example, our techniques indicate that gold alone is not sufficient to predict success, it’s how gold is used to gain intelligence, damage, etc. that is important. That information can be invaluable for a developer.”
The ultimate goal for this team is to develop real-time visualization tools that could train game players to play more successfully. These tools could be incorporated into games by game developers, or could be developed into stand-alone training modules. With increasing prize pools on the line each year for many of these games at big events like Intel Extreme Masters (IEM), World Championship Series (WCS), Major League Gaming (MLG) and others, there could be big business potential for this type of research for gamers who want to turn pro, pros who want to stay competitive and game developers looking to beat the competition in this crowded eSports field.