The True Skill algorithm estimates the “true skill” of each individual player with respect to other players while taking into account as many favourable and unfavourable conditions as possible. TrueSkill is very similar to the ELO rating system, except it doesn’t just take into account the results of games but also considers overall performance and other nuances. This is extremely useful for games that require match-making and has been used in a lot of esports as well as conventional sports. It is also very accurate when it comes to predicting the outcomes of games. If you are interested and want to learn more in-depth about True Skill, you can click on this link here and here.
As per usual, we decided to take it a step further! We have been working on a “Team True Skill” algorithm which goes beyond skill estimation of individual players but adds internal and external factors to make a more complete estimate of the level of a team and the individuals within it. For example, we take into account team chemistry, specific matchups against other teams, home-court advantage, etc. It lets us shift between considering how good an individual player is and how good an overall team would be. Of course, this provides a new set of highly complicated mathematical and computational problems, but our team of brilliant mathematicians is more than happy to challenge and solve those!
So far, we have used this model to predict the results of horse races and tennis matches, and we are planning on extending this to major esports such as League of Legends, Data 2 and CS, as well as to basketball and football.
