For each game, the current rating (and accuracy) of each player is used to determine the odds of winning. The higher your odds of winning, the less impact the win will have on the rating.
The accuracy of the rating (stars) is based on the number of game results. If a player never played in a CompuSport event, his rating accuracy will be 1 star. As more results are included, the number of stars increases. An average of 60 game results are needed to obtain 4 stars and about 140 game results to obtain 5 stars.
A power factor is calculated for each player using wins vs losses, points differential, flawless games. For each division, the standard deviation is calculated, depending on the distribution of each power factor, and a starting rate is evaluated for each player.
To establish the starting rate, the history results of each player was integrated into the calculation, a minimum of 100 league games or 25 tournament games was used. If there are not enough games, the rating is set to the average of the CSR of all players in the divisions with an accuracy of 4-5 stars.
Each week, a formula is used on each game result to determine points to add or remove to each player’s CSR. This formula uses the accuracy of each rating, the odds to win, and the results of the game (win or loss, flawless game). For a flawless game like ERO, NRO or Break and run, no points are subtracted because the loser didn't really play.
To prevent cheating, the CSR rating calculation is done so it cannot move rapidly over a few weeks. This means a player can’t lose a bunch of games before a tournament to make his CSR lower.
