Welcome to AscentStats

We build statistical models and software tools for climbing analytics.

2026 season opens: Kaunas and Keqiao

Sorato Anraku and Zélia Avezou won the World Climbing Series opener in Keqiao on May 1; Sam Avezou and Lina Funa took the Europe Series opener in Kaunas a week earlier. The two Avezous are siblings. Janja Garnbret returned to competition with a 0.2-point silver behind Zélia.

The 2026 fit also introduces a categorical event-tier model, with separate route-grade offsets estimated per tier. The data so far suggests Europe Series routes have played roughly 125 ELO (men) or 225 ELO (women) easier than World Cup routes.

Read the season opener →

Competition climbing data science

We have analysed the last 19 years (2008–2026) of IFSC bouldering competitions using a full Bayesian dynamic Bradley-Terry model to find out which climbers are the most skilful, and which boulder problems have been the most difficult. Check out our analyses and elo rankings for climbers and routes here.

Recent Posts

Kaunas and Keqiao 2026: event tiers and the new model

The 2026 season opens with two events one week apart, one a tier below the other. We use the contrast to motivate a change in the Bayesian model: route-grade priors now carry separate offsets per e...

View all posts →

FlashGrade

FlashGrade is a mathematical grading system for climbers! It leverages probabilistic data modeling to accurately assess the skill level of climbers based on their ascent data.

Adam Ondra’s bouldering flash grade

We applied the flash grade model to Adam Ondra’s bouldering ascent data on 8a.nu. Check out the results.

FlashGrade tool

We have developed a simple experimental tool that allows climbers to input their ascent data (from thecrag.com) and estimate their climbing ability. Currently it only uses the simple Bradley-Terry model, but a version that uses the dynamic Bradley-Terry model is under development.

Cumulative Performance Grade

The cumulative performance grade is a universal index of climbing performance over multiple ascents that weights each ascent proportional to the expected difficulty compared with other ascents.