Chess960 Stats

How we sample the population

To estimate where you sit in the global Chess960 population we first need to sample that population. There is no chess.com endpoint that returns "everyone who has ever played 960", so we walk the player graph ourselves.

We pick six seed players spread across the rating ladder — roughly one each near 500, 1000, 1500, 2000, 2500, and 3000 rating — and we collect every opponent each seed has faced in a live Chess960 game in 2026. Then we collect every 2026 live-960 opponent of those opponents, and so on. The result is a breadth-first crawl across the active 2026 player graph, balanced across the rating spectrum so we don't oversample any one neighbourhood.

How the database grows over time

Every player we visit contributes a handful of opponents we have not seen before. The rate of new discoveries slows naturally as the crawl continues — we start running into players we already know. Plotting the size of our observed population against the number of API pulls gives us a saturation curve. We fit it with a Hill function so we can compare today's reach with where the curve eventually levels off, which gives us a working estimate of the size of the active 2026 live-960 community.

How your percentile is computed

Once we have the rating distribution, we compute your percentile using the mid-rank convention: half of the players at exactly your rating count as "below" you and half count as "above". So if 80% of observed players are below your rating and another 4% are exactly at your rating, your percentile is reported as 82%.