It is not an average and not a head-to-head. Every event is a set of comparisons: each athlete finished ahead of some of the field and behind the rest. The rating is the strength that best explains every one of those comparisons in the division, across every season on record. Beating a strong athlete moves you more than beating a weak one; a strong field that beats you costs less than a weak one. That is what takes the strength of each field out.
Recent results count more. A result loses half its weight every 18 months: a year old counts 63%, two years 40%, four years 15%. The "Evidence" column is the total weight an athlete's comparisons carry, in full-weight events. An athlete whose results are all three or four years old can have a long record and little evidence, and the rating is pulled toward 1000 when evidence is thin.
So two athletes can disagree with their own head-to-head. If A beat B twice in 2022 and B has won eleven events since 2023 against the current field while A has not competed, B rates higher today. The two meetings still stand (they are on the compare page), but they say who was better in 2022, at 15% weight now; the rating says who the evidence favours now.
Reading the numbers. 1000 is an average rated athlete; a 250-point gap is about a 73% chance of finishing ahead, 500 points about 88%. Adjusted % is the other view of the same results: what the athlete would be expected to score against an average field, in the % of possible you already read. Minimum 3 events to be rated; fewer than 5 is provisional. No result in 24 months: inactive, rated but unranked. Season points and standings are unchanged by any of this.
