Each time you register a match result, your play level is updated. The better you perform, the more stars you earn. Most players have 4 stars or fewer, while top club players can reach 6 stars, and world-class players reach up to 10 stars.
Your level increases if you perform better than expected. For example, if you are a 3-star player and you win against another 3-star player, your level will go up and your opponent's level will drop. When you lose a match, your level may still go up. For instance, if you have 3 stars and narrowly lose to a 4-star player, your level will increase while your opponent's level will decrease. On the other hand, if you are a 3-star player and win by a small margin against a 2-star player, your level will decrease, and your opponent's level will increase. In the latter case, you'll need to win by a significant margin to maintain or raise your level.
Category neutral
Our play levels are 'gender-neutral', meaning a match between a woman and a man should be balanced if their level is the same. We use different level descriptions (e.g. Beginner, Intermediate, Advanced) for men, women and junior players so that levels are accurately described while maintaining a unified play level.
Team play level
Individual results and team results are tracked separately. In competitions with fixed teams, match results only update the team level. Your individual level is not affected.
Open matches and 'Americanos'
Your individual play level is adjusted when you play in temporary teams — with smart adjustments based on your certainty level and the sport you play.
New players
Your level does not change when you play against a new player with an unreliable level, unless you are a new player yourself.
Yearly corrections
At the start of each year, your level may change even if you haven't played any matches. To keep the system fair and meaningful, we analyze the spread of all player ratings (their statistical distribution and standard deviation). If the overall scale has drifted during the year, we adjust levels for all players at once, while keeping the relative order exactly the same.
For nerds :)
Bayesian inference meets racket sports: we model player skill as a dynamic normal distribution, continuously updated from match outcomes. Z-scores, star ratings, and vanishingly rare 10-star outliers — the math is worth a look.