Rated like chess, styled like the IQ scale β 100 is average. A game estimate, not a clinical measurement.
Five fresh puzzles at your rating. One attempt a day β the streak is the cure for brainrot.
Rated like chess, styled like the IQ scale β 100 is average. A game estimate, not a clinical measurement.
Five fresh puzzles at your rating. One attempt a day β the streak is the cure for brainrot.
Play in order, or jump straight into the deep end β harder levels move your estimate much faster. Bravery is statistically rewarded.
You played . Waiting on their reply code β paste it below the moment it lands.
Get matched with an opponent at your rating and race them through five random puzzles. One winner.
Play five puzzles, then send the link. They face the exact same five β your times become the ghost they're racing.
Got a link or a code? Paste it here β challenge codes start the match, reply codes settle one you sent.
Shown in duels and, once leaderboards go live, next to your rating.
Everything Brainmaxx knows about you lives on this device. You can take a copy or wipe it β no request form, no waiting.
Privacy Policy Β· Terms of Service
Questions about your data: privacy@brainmaxx.gg
Where you sit β against the general population, and against the people who play this game.
No dailies yet. Today's challenge is waiting on the home screen.
Live rankings against every player need an account and a server β that's the next build. Until then, the player-field percentile above is a model of the Brainmaxx population, not measured data, and duels settle peer-to-peer through shared codes.
0 of 14 collected β the shelf won't fill itself
Brainmaxx is a training ground for the skills IQ tests measure: pattern recognition, rule induction, trend-finding and planning. Every puzzle is procedurally generated β no two runs are alike, so you can never memorize answers, only get genuinely better. Think of it as a gym for your gray matter: short sessions, honest feedback, and a rating that moves the way chess ratings do.
IQ β the intelligence quotient β is a score from standardized tests of reasoning: spotting patterns, working with abstract rules, and solving unfamiliar problems. It doesn't measure knowledge; it measures how you handle something you've never seen before. That's exactly what this game generates: puzzles you can't have memorized.
Scores sit on a bell curve with the population average defined as 100 and a standard deviation of 15. About 68% of people land between 85 and 115; roughly 1 in 7 reach 115β130; about 1 in 50 pass 130 β Mensa's entry bar is the top 2% β and 145+ is about 1 in 1000. The matrix puzzles here are modeled directly on Raven's Progressive Matrices, the most culture-neutral format ever devised.
Honest answer: training makes you dramatically better at the skills β rule induction, trend-finding, classification, planning. Whether that fully transfers to your "true" IQ score is debated. But the skills transfer everywhere:
Your rating uses an Elo system β the same math as chess ratings. Every puzzle is an opponent whose strength matches its level. Beat harder puzzles, gain more; miss easy ones, lose more. The range narrows as you play, because more answers mean more evidence. It's calibrated to feel like the IQ scale, but a real IQ score requires a proctored test.
You're shown where you sit twice: against the general population (the standard curve, mean 100, SD 15) and against Brainmaxx players. Those differ a lot β people who choose to train reasoning are a sharper-than-average crowd, so the same score is rarer in the world than it is here.
The honest caveat: until online play ships, the player-field percentile is a model of that population (centred slightly above average), not measured data from real players. It's labelled as an estimate everywhere it appears, and it gets replaced by real numbers the moment there's a server counting them.
Every puzzle here is generated from a seed, and the same seed produces exactly the same puzzles for everyone. So a duel is just a shared seed: you play five, your run is packed into a link, and whoever opens it faces those same five puzzles while racing your times as a ghost. More correct wins; a tie breaks on total time.
Duel rating is a genuine chess Elo (K-factor 40 while provisional, then 24) kept separate from your IQ estimate β beating someone far stronger moves it a lot, beating someone weaker barely moves it. Results are self-reported between friends, which is fine for friendly play; verified ranked matches arrive with the online arena.
Quick match pairs you with an opponent within 80 points of your duel rating and races you through five random puzzles pitched at that level β one winner, decided on correct answers first and total time as the tiebreak.
Right now those opponents are bots, and the game labels them as such everywhere they appear β in the match card, the race bar, your duel history, and anything you share. They're not fake people: each one is a rating and a play style, and its run is simulated from the same Elo maths the rest of the game uses, then replayed live so you watch it solve alongside you. Sparring moves your rating at half weight, so the ladder still mostly reflects matches against real humans.
Queueing for human opponents at your rating needs a server to hold the queue β that's the next build. When it lands, the race screen stays exactly as it is; only the opponent changes from a simulated run to a real one.
Your name shows up in duels and on the leaderboard. You can change it any time.
Accounts, cloud saves and global leaderboards aren't built yet. Leave your email and we'll write to you the day they launch. Nothing is sent now β no confirmation email, no account is created, and the address stays on this device until then.
By playing you accept the terms and privacy policy. Brainmaxx is for entertainment and training β not a clinical IQ measurement.