One engine.
Two lenses.
Every number.
Every rating, rank and win-probability on KickServe comes out of a single system — a dual Elo + Glicko-2 engine that has rated 23,000+ ranked players across four surfaces, and re-runs itself every morning. No ensemble of black boxes. Here's what happens inside.
Follow one result through the engine
The same match, scored two ways. Watch an upset ripple through both models — and see exactly why one of them hedges.
def. #8 seed
Illustrative — real shifts depend on the gap between the players and how settled each rating already is. Elo moves decisively on a surprise; Glicko-2 waits for the evidence to catch up.
Why the engine runs two models
Head-to-head expectation
One number per player, nudged after every match by how the result compared to what the model expected. Fast, battle-tested, decisive — it reads current strength cleanly and reacts the moment a result lands.
Expectation, with a confidence band
Carries a reliability deviation (RD) that widens on thin or stale data and pulls its estimate toward 50%. So a three-match hot streak can't masquerade as elite until the record backs it up.
When the two disagree, that gap is the story — the model telling you how much to trust its own read. You can watch it move on the interactive methodology demo →
Four ratings, not one
Clay isn't grass. The engine keeps a separate Elo and Glicko rating on every surface — plus a rolling recent-form window — so specialists read as the different players they are.
What's running, and how often
The engine, on the court
Everything above turns into these — updated all day.