How the rating works

For when you want to know why to trust the number. One rating for the team and every player, split into six categories (Impact, Threat, Control, Discipline, Defense and Duels); teams skip Impact, since a team’s goals are its scoreline.

Two sets of replays

The SSL baseline

9,000 replays, GC3/SSL

The bar. Every stat you put up is held next to what GC3 and SSL players do, so matching their average puts you on 100.

The cross-rank win signal

36,000 replays, Bronze to SSL

The rulebook. It picks which stats count and whether more is better or worse, by looking at who actually won in recent ranked games. SSL players are already good at the basics, so only games from every rank show which basics matter.

From stats to one number

  1. 1Every stat becomes a rate (per minute, per event or a share of the match), so game length doesn’t skew it.
  2. 2Each rate is compared with how SSL players perform, and related stats are grouped into categories like threat, control and defense.
  3. 3The categories combine into one number, where the average SSL player lands on exactly 100.

Reading the number

100 is the average SSL player, so below SSL your rating is supposed to sit under 100. Watch your own trend and your lowest category rather than the raw number.

135+
Peaking
Top 1% of SSL form.
115+
Popping off
Standout, well above the SSL average.
100+
Consistent
Right at the SSL average.
85+
Coasting
Slightly under the SSL average.
70+
Slipping
Well under the SSL average.
<70
Choking
Far below the SSL average.

The details

What is a Rating?
A single number per match, for your team and for every player, split into six categories (Impact, Threat, Control, Discipline, Defense and Duels), so you can see which part of your game carried the match and which part dragged; teams skip Impact, since a team’s goals are its scoreline. 100 is the average performance in the 9,000-replay GC3/SSL baseline.
What is the Baseline?
A pre-computed baseline built from 9,000 recent GC3/SSL replays. It holds the median and spread of every stat, and each of your stats is shown next to it so you can see where you sit. It only sets the bar; a separate dataset decides which stats count.
What decides which stats count toward the Rating?
The cross-rank win signal: a separate set of 36,000 recent ranked replays, spanning Bronze to Supersonic Legend. For every stat it measures how strongly it moves with winning, which settles both whether the stat counts and which direction is better. It never touches the scale. That's the baseline's job.
Why two datasets instead of one?
They answer different questions. SSL replays tell you how high the bar is, but they are a bad place to learn what wins games: a fundamental that every SSL player has already mastered barely varies at that level, so it reads as noise. Ranked replays across every rank still have that variation, so they show which stats decide matches.
How is a Rating calculated?
Every stat becomes a 'rate' first (per minute, per event, or a share of the match), so game length doesn't skew it. Each rate is scored by how many standard deviations you sit from the average in the 9,000-replay SSL baseline, and the stats in a category are weight-averaged into one number. Which stats get into that average, and which way each one points, comes from the separate 36,000-replay cross-rank win signal, measured on what wins games rather than picked by us.
How accurate is Rating?
Rating will never be perfect. The fundamentals are defined by hand (challenge for the ball, keep boost, stay goal-side), each taking an equal share of its rating category. Everything else is measured: the 36,000 ranked replays of the cross-rank win signal decide which direction of a stat is better and how strongly it moves with winning, and the 9,000 SSL baseline replays decide how far off the pace that puts you.

Read more

Seen enough? Put your own replays through it.

Find out what's holding you back

Free · 100% local · No account

Windows & Linux · currently in beta