What Is the Statfeed Score?
Short answer: the Statfeed Score (SFS) is a 0–10 rating Statfeed computes for every player, in every match, from the events Rocket League's own Stats API reports — goals, saves, shots, demos — weighted by the scoreboard and clock at the moment each one happened. It's frozen the instant the match ends, and it comes with a receipt: the exact events and weights that produced it, exportable, checkable. The rest of this page is the plain-language version; the full methodology, the blind spots stated up front, and the version history live on the Statfeed Score page.
Is it the same as my Rocket League rank or MMR?
No, and this is worth being direct about: the Stats API Statfeed reads doesn't transmit rank or MMR at all — Statfeed can't see it, so it can't rate you against it. The SFS is not a rank, not a hidden ladder, and not a competitor to Bronze-through-Supersonic Legend. It's a grade for one match. Rank answers “how good are you, broadly, over time, against similar players.” SFS answers a narrower question: “in this specific game, weighted by what the scoreboard and clock actually demanded, how much did you contribute?” Different question, different number, no attempt to replace the one Psyonix already owns.
How is it different from the score you see in-game?
The in-game score counts volume — goals, saves, shots, assists, each worth a fixed number of points, all the time, no matter when they happened. Three garbage-time goals in a 7-1 blowout score the same as three goals that dragged a game back from 0-3. The Statfeed Score weights the same events by context: the scoreboard and the clock at the instant each event happened, capped so no single moment can swallow the match. It also divides by minutes actually played, not wall clock, so a match padded with kickoff pauses and a long goal replay doesn't quietly dilute anyone's rating. The final score never enters the formula directly — no win bonus is added anywhere.
How is a rating like this calculated, in plain terms?
Four ideas, no formula:
- Start from what the game actually reports. Goals, assists, saves, epic saves, shots, demolitions — real events from Rocket League's own Stats API, not an estimate.
- Weight each one by the moment it happened. The same save matters more at 2-2 with 20 seconds left than at 5-0. Context is what turns a raw event count into a rating.
- Divide by minutes of actual play. So a longer or more paused match doesn't inflate or dilute the number for reasons that have nothing to do with how anyone played.
- Map the result onto 0–10 against real matches. The scale is anchored to percentiles of an actual population of past performances, not to a gut feeling about what “good” looks like — see the exact population size and how that anchoring works on the Score page.
The exact price of each event — how many shots or demos buy how much rating — isn't published as a menu on this page. That's deliberate: a public price list turns into a farming guide. The full numbers for your match are in the receipt Statfeed exports for it, where they have context and an audit trail instead of being a recipe.
Every change to how the rating is computed — and what it did to ratings that already existed — is logged on the changelog.
Does a higher score mean I played better?
Usually, but read one honest number before you trust it too far: on the calibration base measured so far, about 42% of the spread in a single match's rating comes from which side of the lobby you landed on, not from you personally — teammates end up correlated, because the score is zero-sum within a match (one player's big goal is, structurally, taken from the pool the rest of the match had to work with). One match is a noisy read of a player. A run of matches averaged together is where the number starts being about you specifically rather than your teammate that game. The full disclosure — what the score structurally can't see at all, like positioning or shadow defense — is stated up front on the Score page, before anything else on it.
What does “SFS” mean, and how do I cite a number?
SFS is short for Statfeed Score — say the full name once, then use the short form, the way people say “Elo” without naming the federation behind it.
- The rating is a number, spoken plainly: “SFS 8.1,” “my SFS,” “an SFS diff of 1.4.” Always with the decimal — a rating doesn't round itself up when the moment matters.
- The version has a
v: “SFS v2.1,” never “SFS 2.1” bare on its own — that gap is what keeps a rating (“8.1”) and a version (“v2.1”) from reading as the same kind of number. - A percentile always names its population: “better than about 55% of this base (n=1,115).” A percentile with no stated population is a number that sounds precise and isn't — the Score page states its base every time, and any citation of SFS should too.
Is it peer-reviewed, or independently audited?
No — it's self-published, on purpose, with the audit trail attached instead of a third party's stamp. What that means concretely: the method is public (the mechanics above and the full detail on the Score page), every match exports a receipt with the exact events, weights, and version that produced its rating, and the headline validity numbers are published next to the population they were measured on — measured across the full local calibration base (271 matches):
- 91.9% of matches: the highest Statfeed Score on the field is on the winning team — with zero win bonus anywhere in the formula. The correlation with winning wasn't built in; it emerged from scoring individual moments.
- 87.5% agreement with the game's own in-match MVP crown on the winning team (a coin flip would land at 50%). The disagreements aren't treated as errors — they're published as the rating having an opinion, with a receipt attached to each one.
- ICC 0.42 — see the “does a higher score mean I played better” answer above: this is the measured share of a match's rating that traces back to your team rather than to you alone.
Nobody outside Statfeed has re-run these numbers yet. The goal of publishing the method and the receipt format is to make that possible, not to claim it's already happened.
Can I use the Statfeed Score in my broadcast or tournament?
Yes — that's the point of publishing the method and the citation format above instead of keeping either one private. If you're calling a series or running a bracket and want a number that comes with its own audit trail, use it, and cite the population alongside it the way this page does. Nothing here is offered as an official or endorsed rating system, and nothing here claims any tournament, broadcast, or organizer has adopted it — if and when one does, that'll be named, with a link, not implied.