The most interesting element of esports is the wealth of potential data available from gamers and their systems to determine what physical performance traits separate elite pros from just participants. It's also interesting how little of this analysis is actually done, as it would be enormously helpful in devising training programs. Luckily, one group of researchers in Russia conducted this research - and it is fascinating.
The researchers analyzed data collected from an eye tracker, keyboard and mouse while subjects (athletes, players and newbs) played CS:GO Deathmatch. Then they investigated the features contributing to classifying the subjects according to their gaming skills.
Eye-Tracking Based Features
Deathmatch simplifies the user interface on the screen, which should greatly reduce where subjects are looking. Interestingly, there is a large difference in where players look based on their skill level:
It turns out that the higher skill the player has, the more the gaze is concentrated on the screen center, as illustrated by the heat map.
There are two main reasons why high skilled players spend more time looking into the center of the screen. First of all, they have better knowledge of the game map and always know how to position themselves and where to aim. Therefore, they do not look around the screen frequently. Secondly, in the situation when an enemy appears (not in the expected screen center position), the skilled players much more quickly move the in-game crosshair to the enemy and then again look at the crosshair keeping the enemy on sight.
Keyboard and Mouse Based Features
Keyboard and mouse data is explored in two ways. First, the percentage of time when a combo of keys and mouse buttons were pressed. Second, the average length of time when specific keyboard and mouse keys were pressed.
Here are the distributions of the most important keyboard and mouse features for each group of subjects:
Here's how this breaks out:
- (a) The usage of forward and backward motions
- (b) the duration of forward and backward motions
- (c) the usage of left and right motions
- (d) the usage of the “shoot being in the duck pose moving to the left” technique
Unexpectedly, professional players use forward, and backward motions on average like newbies and the usage of these motions decreases as the experience of the subject grows from the low-skill amateur to the professional. The duration of these motions decreases as the subject's skill increases.
Professional players also have less usage of forward and backward motions, they move left and right more often than the players from other groups. The higher this feature, the higher the skill of the subject.
The use and duration of more exclusive combinations peculiar to professional players, such as “A & Ctrl & MOUSE1”, are considered (see Figure 4d). This feature is comparatively exclusive: the newbies almost do not know about it and only a few amateurs exploit it. Even some professional players use it occa- sionally, which results in high variance on the box-plot. Nevertheless, the high value of this feature indicates the highly skilled player.
Predicting Player Performance
After analyzing all this data, what are the most important factors in predicting performance? It comes down to a few key features:
What's needed is an analytics package, preferably open source, that can collect this type of data for individual and offer prescriptive training plans to improve performance.
Want to improve your performance? Level up with the Official Nootropic of Esports™!
Esports Athletes and Players: a Comparative Study. Khromov, Korotin, Lange, et. al.