How the GOE Index rating works

One number per skater: the total score we expect at their next ISU event

What it measures

The ISU World Standings reward placements over two seasons, so a skater who misses events drops, however well they skate. Season's best rewards one great day. The GOE Index rating asks a narrower question: what total does this skater post on a typical day right now? It is measured in points, on the same scale as the scores you see on the kiss-and-cry screen.

The formula

For each skater we take every completed ISU event with a free skate in the last 2 years and average the totals, giving recent results more weight:

rating = Σ (wᵢ × totalᵢ) / Σ wᵢ,   wᵢ = 0.5 ^ (days since event ᵢ / 240)

A result from 240 days ago counts half as much as one from today; a result from two years ago counts about an eighth. Events where a skater did not reach the free skate are left out of the average, because a short-program-only score is not comparable with a full total.

Who gets a rating

Skaters competing as Individual Neutral Athletes appear under “AIN”, as in ISU results.

How we tested it

We built each candidate method on events before 1 July 2024 (2,503 head-to-head pairs) and then asked it to predict who finishes ahead in every pair of skaters at the events after that date (5,142 pairs), using only results known before each event.

MethodPairs called rightBrier score
Rating combined with Elo80.7%0.137
GOE Index rating (time-weighted average total)80.6%0.139
Best total in the previous 365 days78.9%0.146
Elo on head-to-head results77.3%0.159
Points by placement, ISU World Standings style75.7%0.170

Brier score measures how good the predicted probabilities are; lower is better. Combining the rating with Elo adds 0.1 points, which is within noise, so we publish the simpler method.

What it does not do

The full code is versioned with the data. If you think a number is wrong, tell us.