Those new to watching competitive eSports games like Dota 2 and League of Legends often complain that the action is hard to follow. Sometimes, it's hard to follow even for people who been playing it for years.
But a team of researchers from Sweden, Denmark and Germany has developed an analytical model that can help commentators (or casters, as they're known in eSports) break down exactly what's going on - and even predict who's going to win.
The model is based on the players' behaviour throughout the game - specifically, the results of every small skirmish from the very start. It considers damage dealt, abilities used and the rewards gained by both sides in the aftermath of the fight. Those statistics are then combined with machine learning techniques to predict who's likely to win.
It was configured and tested using replays of more than 400 games of Dota 2. The game's digital nature means that it's possible to get much more detailed statistics from it than from, for instance, a real-life football or tennis match.
The details of its inner workings were published in a paper that was presented at the MIT Sloan Sports Analytics Conference in Boston.
"The model provides analysts with ways to better communicate to their audience what's happening in a game. Sometimes the game moves so quickly that it is hard to see exactly what's happening", said co-researcher and cognitive scientist Tobias Mahlmann from Lund University in Sweden.
Mahlmann added that many teams will find the results interesting and be able to use the data it generates to improve their play. "The opportunity to be able to analyse games in depth and evaluate tactics is interesting to analysts and game developers, but also to the players themselves."