Win a match and your rating climbs. Sometimes it leaps; sometimes it barely twitches. Sometimes you win and the reward feels almost insulting. None of that is mood or randomness — there's a small, well-behaved machine deciding exactly how much your rating should change, and it has been running quietly in the background of competitive games for sixty years. This is the whole thing, taken apart on the table.
01One number, not two
Every Quizzer player carries exactly one skill number: an ELO rating. It is the number on your profile, the number the leaderboard sorts by, and the number matchmaking uses to find you an opponent. Nothing is layered on top of it and nothing is hidden from you — the raw rating is what you see.
That is a deliberate choice, and it hasn't always been the case. An earlier version of Quizzer wrapped the rating in a separate trophy count and ran a parallel XP track beside it. Both are gone. Two numbers that always moved together were really one number wearing a costume, and the costume mostly made it harder to answer the only question that matters: am I getting better?
02The formula
The engine is standard ELO, the system chess has used since Arpad Elo published it in 1960. Before a match settles, it asks one question: how likely was this result?
That produces your expected score — a number between 0 and 1. Two players rated the same expect 0.5 each: a coin flip. A player rated 200 points higher expects roughly 0.76, so the system thinks they should win about three times in four.
Then the result arrives, scored as 1.0 for a win, 0.5 for a draw and 0.0 for a loss, and the rating moves by the gap between what happened and what was expected:
The 32 is the K-factor: the largest swing a single match can produce. It's the dial that decides whether ratings move quickly and noisily or slowly and stubbornly, and it's why no single match can move you more than 32 points in either direction.
The consequence is the part worth internalising. Beat someone rated well above you and expected was small, so the gap is large and you take most of that 32. Beat someone well below you and you were supposed to — the gap is nearly nothing, and so is the reward. Losing works the same way in reverse: losing to a stronger player costs little, losing to a weaker one stings.
03Why you start at 1600
Every new player begins at 1600. There's nothing magic about the value — it's a middle with room for the maths to move in both directions. What matters is that everyone starts in the same place, so the number is comparable across the whole player base from day one.
Ratings are clamped at a floor of zero, so a loss can never take you negative. In practice that only matters below about 32 rating, since a single match can't swing further than the K-factor.
One more thing about starting out: the leaderboard only lists players who have completed at least one match. Sitting at the 1600 starting rating doesn't place you above people who have actually played — you appear on the board after your first match.
04When it moves
The rating changes once, after the match completes. Not per question, not mid-match, not when your opponent finally takes their turn. One match, one adjustment.
Before you start, the app shows what the pairing puts on the table: how much you stand to gain by winning and how much you stand to lose by losing, as two separate numbers. Because ELO is asymmetric, those two are equal only when both players are rated the same. They're computed with the same maths that settles the match, so they can't drift from the real payout — but they're still an estimate, because the final calculation uses each player's rating at the moment the match ends, not when it was created.
05What doesn't count
A few things people reasonably assume affect the rating, and don't:
- How fast you answered. The clock is pressure, not currency — it never touches your rating. (There's a whole article on what the clock does and doesn't do.)
- Your score margin. Winning 10–0 and winning 6–5 move your rating by exactly the same amount. ELO cares that you won, not how comfortably.
- Using Brainpower. Striking a wrong answer off a question is a legitimate part of the game. Matches where either player spent Brainpower count fully.
- Achievements. They're a record of things you've done. They never feed back into the rating, matchmaking or the leaderboard.
06Where it shows up
The same number does three jobs. It's the rating on your profile and in your match history. It's the value the global leaderboard ranks by, with ties broken alphabetically by username. And it's what matchmaking reaches for when you ask for an opponent, because the fairest match available is the one closest to your own rating. That last one is its own story.
07Questions have ELO too
Here's the part that surprises people: every question carries an ELO rating as well, and it plays matches against you. Answer correctly and you beat the question. Get it wrong or time out and the question beats you.
Questions start at 1600, use the same K-factor of 32, and update after every answer — using your rating as the opponent. Your own rating is untouched by this; it only ever moves when a match completes.
Run across thousands of answers, this sorts itself out: questions that strong players keep failing drift upward, and questions almost everyone gets drift down. The difficulty label you see is read straight off that number — Easy below 1450, Medium from 1450 to 1750, Hard above 1750. Which is the whole point: "Hard" is a measurement taken from real answers, not a guess someone typed in. More on that here.