The simple answer first, the full math below for the curious.
Everyone starts at 1000. When you win, your rating goes up. When you lose, it goes down. How much it moves depends on who you played.
If you're rated 1050 and you beat a 1200-rated player, you gain a lot โ you upset the expected result. If you beat a 900-rated player, you gain less โ it was expected. Losing to a much stronger player costs little; losing to a weaker one costs more.
Unlike a round robin where everyone needs to play everyone, Elo handles partial schedules through transitivity: if Alice beats Bob and Bob beats Carol, that information propagates through the ratings even if Alice and Carol never play. Saturday groups with 50 players rotating through courts work perfectly โ after 15โ20 sessions, ratings converge to an accurate picture.
New players have a Provisional badge. Their rating moves twice as fast (K-factor 64 vs 32) for the first 10 matches, so they land at their real level quickly instead of slowly drifting there over months.
Club admins can designate one or a few Anchor players โ typically established members whose level is well-known. Anchors use the stable K-factor from day one and serve as reference points that the whole group calibrates against.
| Rating range | Approximate NTRP equivalent | What it means |
|---|---|---|
| 1000 | โ | Starting point / not yet calibrated |
| 1050โ1150 | 2.5โ3.0 | Developing player, consistent rallying |
| 1150โ1250 | 3.0โ3.5 | Club competitive player |
| 1250โ1350 | 3.5โ4.0 | Strong club player, tournament-ready |
| 1350โ1450 | 4.0โ4.5 | Experienced competitive player |
| 1450+ | 4.5+ | Advanced / former competitive player |
These are rough approximations โ your club's internal ratings are relative to each other, not to a global scale.
ClubLadder uses the standard Elo formula developed by physicist รrpรกd รlล (pronounced "AY-low") and used worldwide in chess, table tennis, Go, and increasingly in tennis.
Before a match, we compute what score each player is "expected" to get based on the rating difference. Expected score is a probability between 0 (certain loss) and 1 (certain win):
Where RA and RB are the two players' current ratings.
Example: Alice (1200) vs Ben (1000).
Alice is expected to win 76% of the time. Ben is expected to win 24% of the time (1 โ 0.76).
The actual score (S) is simple:
The rating change is:
K is the K-factor โ it controls how much a single result can move a rating:
Example continued: Alice wins (S=1, E=0.76, K=32):
Ben loses (S=0, E=0.24, K=32):
If Ben had won (the upset), the swings would be much larger:
In doubles, ClubLadder uses the average rating of each team as the "player rating" in the formula above. Both players on the winning team gain the same ฮ; both on the losing team lose the same ฮ.
This means partnering with a much stronger player does move your rating, but the effect is shared โ your individual rating always reflects your personal performance history, not just who you happened to partner with.
Following ATP/WTA convention: a retirement (player stops mid-match due to injury) or default (player cannot start) counts as a standard win for the opponent in the rating calculations. A walkover (opponent never appears) does not count for ratings since no ball was struck.
A key property of Elo: the sum of all ratings in the pool is conserved. Every point Alice gains comes directly from her opponent. This means the average rating in your club stays near 1000 as long as new players enter at the baseline.
It also means ratings are relative โ a 1250 in one club doesn't necessarily mean the same as 1250 in another club, unless the two populations have played each other.
For large groups with rotating partners (like Saturday sessions), ratings converge reliably after ~15โ20 sessions per player. The more cross-group matches occur, the faster and more accurately the entire pool calibrates.