A player ranking system is a way of measuring and comparing the skill level of players in a game or sport. It assigns a rating or score to each player based on their performance in matches or games against other players, and uses it to rank them from highest to lowest. It also updates the rating or score after each match or game, and tries to predict the outcome of future matches or games. Player ranks are used to create closer races in sim racing or matches in other games.
First and foremost, let’s clear the confusion. SGP Pure Performance ranking system is not an ELO based ranking system, and therefore is not comparable to any other ELO based system.
ELO was originally developed for Chess, 1v1, clear and even rules for either player where the outcome was solely based on the player's skill against the other player.
Race results in sim racing are depending on a multitude of factors: driver skill, weather and track conditions, car performance, number of drivers, so every race is unique in a certain way. It is not a 1v1 game in the same conditions.
For example, there are 2 communities - one has very fast pro drivers and the other is for casual racing. The drivers of both communities are racing only inside their communities. Given the same amount of races and similar conditions, top drivers of both communities would have similar ELO as they won the same number of races and opponents, but in reality they have different skills.. Only if drivers of both communities raced together - their ranks would be corrected and ELO rank would differ. This can happen even in a single large community where groups of drivers do not periodically race with one another. Those communities need to do additional work to evaluate their driver level.
Another example - if a very fast driver joins rookies’ race, he can win the race without much effort, not even trying to match his top performance and ELO would still give him some rank increase for the win. Basically some fast drivers can “wisely” choose opponents.
SGP ranking is a proprietary driver performance evaluation system that is built upon a machine learning model and uses race data like car, track, track conditions and game settings, opponent count, race duration. It evaluates driver performance taking into account all these conditions. Besides, SGP ranking evaluates race history, so one bad race would not change rank completely. SGP ranking supports Assetto Corsa, Assetto Corsa Competizione, RaceRoom and Automobilista 2 (soon) - games that are run on Simracing.GP server infrastructure.
SGP rank is evaluating drivers performance with given conditions therefore it doesn’t have limitations of ELO based ranking. If we take the example of faster and slower communities, the top drivers of a faster community will have higher ranks of top drivers compared to slower community drivers even if they haven’t raced together.
Following the second example of a fast driver winning a rookie's race without performing to the max of his abilities, SGP rank would reduce the rank for the winner as he was driving much slower than he could in given conditions. Therefore to keep his rank a driver will need to choose the races and opponents of his or higher level, or race at the max of his capabilities. No sandbagging.
SGP ranking has built in penalties for drivers who sign up and then skip the races or quit before race end (completing less than 80% of a race). The rank drop would be as much as a driver had raced at 0 speed. So, even the slowest driver who had accidents, but finished the race would get a lower drop than the one who had dropped or did not participate.
SGP ranking has a decay of 10% of rank per 4 weeks (28 days) break.
