A number of videos games are currently available (in some cases for free) that pit teams of human players against each other in occasionally chaotic combat. Determining which team will win can be quite difficult when the end is not near, because of the number of variables involved. Researchers at North Carolina State University though have recently developed a technique to review logs from action-RTS games, such as DotA, to find what determines victory.
In games like DotA, there are a myriad of stats to be concerned with, including how much damage a specific character can inflict and how much gold and experience a player or team collects. Because of the interaction between the different stats, the impact these data will have on victory can vary greatly, though the stats themselves do not. For example, in DotA, if Team A develops 59.7 more damage points than Team B by the second quarter, Team A is more than 80% likely to win. Change that to 54 damage points though, and the likelihood of winning drops to under 10%.
Armed with this research, it may be possible for players to determine optimal strategies for winning action-RTS games, as well as helping developers identify how gameplay mechanics will affect their game. Now the researchers are working on visualization tools to tell players how they are doing in real-time, compared to those strategies predicted to be successful.
Source: North Carolina State University