How Verdict decides what's true.
Every read in Verdict has to clear the same four bars before you ever see it. None of this takes a statistics degree to understand — that's the point. But you deserve to know the rules we hold ourselves to.
Every pattern stands trial against its closest matches in your own history — never against everyone's average. We give luck a chance to explain every pattern before we call it real. Every fact carries the receipt of how it was known. And when your data is too thin to judge, we say so, plainly, instead of inventing a story.
Your games, compared like with like
Most sites compare you against everyone's average — millions of players who aren't you, on champions you don't play, at ranks you've never seen. Verdict never does that. When something looks like a help or a hurt, that pattern stands trial: the games carrying the habit on one side, their closest matches in your own history on the other. Same you, same style of game — the habit is the only difference on the table.
That's what we mean by a fair fight. It's the difference between "players who take the first dragon win more" — true of everyone, useless to you — and "you win more when you take it" — true of you, and worth acting on.
Before anything is called a driver, luck gets a turn
Look through enough games and you will always find patterns — and most of them are coincidences. So before Verdict names anything a driver of your wins or losses, it asks one more question: could plain luck have produced this? Verdict shuffles the deck and checks. If luck alone could paint the same picture, the pattern stays unnamed.
Only what survives gets a verdict. That's why Verdict hands you two or three things instead of forty — most of the forty were noise, and we're not going to pretend otherwise.
The same test guards every link. When Verdict connects two events — a death, then a lost objective — it first checks the timing against your own games' natural rhythm, and even then it will only ever say one thing followed the other. "Caused" is a word the app refuses to use.
"Not enough evidence" is an answer
When your data is too thin to judge — a champion you've played six times, a matchup you've met twice — Verdict says so. A win rate off eight games isn't a win rate; it's an anecdote. We won't dress an anecdote up as a number, because the moment we do, you'd be right to stop believing the rest of the page.
That's rarer than it sounds. Most sites will show you 38% off eight games as if it were gospel. Verdict holds the claim until your games can actually carry it — and tells you what it's waiting for.
Every claim carries its receipt
Ask Verdict how it knows, and it can answer. Every fact the engine states about your game carries the record of how it was known — read straight from the match's own events, or estimated from its timeline — and an estimate is never allowed to dress itself up as a fact. The receipt travels with the claim, enforced by the machinery itself, not by good intentions.
That's the standard the whole product is built toward: not just true things, but true things that can show you why they're true.
Why we work this way
Anyone can show you numbers. The real work is deciding which numbers deserve to be believed — and having the discipline to throw the rest away. That discipline is the product.
It's a discipline we've already paid for. Our first playstyle engine looked right — until we measured it and proved it was quietly detecting whether you won, not how you played. So we scrapped it and rebuilt playstyle from the choices you make in a game — how you spend your damage, your kills against your assists, your vision, your tempo — each one screened to stay silent about winning. When our own work fails these bars, it goes too.
If anything Verdict tells you ever fails these bars, that's not a quirk, it's a bug — tell us.