How the results score works

The short version

For each student a coach taught, we measure how fast that student's rating moved while they were being coached. We compare that against players of a similar standard, in the same sport, who were not being coached. A coach's score is the average of their students' comparisons, pulled toward the middle when there is not much data, and shown as a number out of 100.

1. What counts as coaching someone

A single unit of measurement is one coach, one student, one sport, over one stretch of time. That stretch is built from sessions the student actually attended — booking and not turning up counts for nothing.

It takes at least 3 attended sessions to register at all.
The stretch runs from the first session to 45 days after the last, because coaching does not stop mattering the moment a lesson ends.
A gap of more than 120 days splits it into two separate stretches rather than treating a returning student as one long engagement.
If the student later starts with a different coach, the stretch is cut short there, so improvement is not credited to the wrong person.

2. Which matches count

Only matches inside that stretch, in that sport, and only ones that survive the checks below. A student whose data cannot support a fair comparison is left out entirely rather than counted badly.

The student must not have been provisional when coaching began — a brand-new player's rating moves fast on its own.
Their starting rating must be above the floor, since a rating that cannot go lower can only go up.
At least 6 matches in the stretch, and no more than 40 count.
At least 3 different opponents.
No single opponent may account for more than 40% of the matches.
At least 80% of their matches must be against other Amathlon players.
No more than 2 coaches at once.

Deliberately excluded

Three ways of inflating a score are removed by construction:

A coach playing their own student. Those matches do not count. Otherwise a coach sitting at the rating floor could feed wins to students at no cost to themselves.
Two students of the same coach playing each other. Also excluded, so students cannot be paired up to trade wins.
Matches against people who are not on Amathlon. An off-app opponent has no verified side, so those matches never enter the calculation.

3. How improvement is measured

For each student we take the total rating change across their counted matches and divide it by the square root of how many matches there were. Dividing by the square root rather than the count is what stops a student who played six matches and a student who played forty from being measured on different scales.

4. What it is compared against

That figure means nothing on its own, so it is ranked against a reference group rebuilt every night: players who were not being coached, in the same sport, who started from a similar rating — bucketed in 50-point bands, so an advanced student is compared with advanced players rather than with beginners. The group is drawn from the last 180 days and is only used when it holds at least 20 players; otherwise the comparison widens to the whole sport, then to all players at that standard.

The output is a percentile. Half of uncoached players sit at 50.

5. From students to a coach's score

A coach's score is the weighted average of their students' percentiles, pulled toward the median by an amount equivalent to five average students. That is what stops one exceptional student from producing a top score. A coach with three students has their average pulled strongly toward the middle; a coach with thirty barely at all.

Where a student was working with two coaches at once, the credit is divided between them rather than counted twice.

6. When a score is shown publicly

Both conditions have to hold:

At least 3 effective students after weighting.
A score of at least 60.

A coach with a high no-show rate — more than a quarter of decided bookings, once there have been at least eight — is not shown a public score regardless.

Below the bar, nothing appears. Not a low score, not an empty badge — the profile simply carries no results figure. Coaches can always see their own number and which students are counting toward it. This is deliberate: a public number that can only help you is a weaker claim than one that can hurt you, and we would rather be honest about that than pretend otherwise.

What this does not claim. This measures that a coach's students improved faster than comparable players who were not coached. It does not prove the coaching caused it. Motivated people seek out coaches, better coaches attract better students, and neither effect can be separated out from observational data. Until enough coaches qualify for us to check how often the score is wrong, we describe it as measuring improvement, not as ranking coaches by the results they produced. The distinction matters and we intend to keep making it.

Recalculation and changes

Scores are recomputed nightly, and whenever a counted student plays a rated match. The qualifying bar and the comparison groups are calibration choices on a young platform, and we expect to adjust them as more data arrives. When a threshold on this page changes, the page changes with it.

Questions or a disputed score

If your score looks wrong, your coach dashboard shows each student and the reason any of them is not counting. If that does not explain it, get in touch and we will look at the underlying data with you.