AscentStatsBoulder

Methodology

Rankings here come from a Bayesian model fitted to IFSC competition boulder results, 2021–2026. Unlike a simple tops-and-zones tally, it reads the full result of every attempt: whether a climber topped, whether they zoned, and how many tries it took.

One attempt at a time

A climber attacks a boulder as a sequence of attempts. On each attempt the outcome is ordered: they top, they reach only the zone, or they get nothing. Topping passes through the zone, so the three outcomes are nested rather than independent. The chance of each depends on the gap between the climber's grade and the boulder's zone and top difficulties, on a shared scale.

The competition grade

The grade we report is a competition grade: the boulder a climber would top, under competition conditions, about half the time: onsight, a 4–5 minute clock, and a handful of attempts. It is not a flash grade (a comp send allows several tries, so it is a touch higher than a true first-try flash), and it is well below a redpoint maximum worked over sessions. The very best sit around 8A+ (V11–12); most of the elite field runs V9–11. These sit below the outdoor grades these same athletes are known for, because a competition boulder must be done on demand, in any style, under pressure. That is the point: it is a like-for-like measure of competition form. Throughout the site the grade is shown as a plain number on this nominal V-scale (so 11.5 means roughly V11–12); we drop the leading V to keep the tables uncluttered.

Two independent sources anchor the scale. Its spread is calibrated through the previous model's graded-boulder reference (a handful of early-competition boulders with known grades) and the standard font-to-V table; its level is pinned so the world #1 sits at 8A+. It is corroborated by a completely separate analysis of outdoor 8a.nu logbooks, with one point worth stressing: that analysis estimates a climber's median flash grade, the boulder they would flash about half the time, not the hardest they have ever flashed. By that measure an elite boulderer sits around 8A / V11. Their best-ever flashes run far higher (Adam Ondra has flashed boulders up to V15), and on an off day they fail well below; 8A / V11 is the typical level, not the ceiling. That is exactly where a competition send should land, since the extra tries in a comp roughly offset the onsight and the buzzer.

Internally the model works on a per-attempt scale (a fixed slope of log 2), also exposed as a chess-style Season Elo (baseline 2000 men / 1750 women). The competition grade is that per-attempt difficulty converted to a per-session result using each boulder's spammability (below), which is why the same boulder can be a hard per-attempt move yet a moderate comp send.

Attempts and spammability

After any non-topping attempt a climber may go again, until the buzzer. How readily a boulder gets re-attempted, its spammability, is estimated per boulder: power boulders allow only a few maximal pulls, while coordination boulders can be tried many times in the same window (the record in this dataset is a top on the 19th attempt). Modelling this per boulder, as a pooled random effect, markedly improves fit and lets the attempt penalty be interpreted correctly. Spammability turns out to be nearly independent of difficulty; see the boulders page.

Event strength

Boulders at different tiers (World Cups / World Series, Europe Series, other) are set at systematically different levels; the model estimates these offsets so a grade earned at a weaker event is not over-credited. A climber's grade is informed by every boulder they touch, with their previous season as a prior.

What to keep in mind