Abstract & spatial reasoning practice · all ages
A matrix puzzle is a grid of shapes that changes by a hidden rule — the format used in many school and entrance reasoning assessments. MatrixMind generates them endlessly and for free, and explains the arithmetic, modular arithmetic and geometry that each one is built from in 39 written guides.
39 written guides · Free to use · Built by one student
The learning behind the puzzles, in plain terms.
A matrix puzzle shows a grid of shapes that change by a hidden rule — you work out the rule and complete the pattern. It's a classic abstract and non-verbal reasoning format, the same style of question found in many school and entrance reasoning assessments, and MatrixMind lets you practise it as much as you like, for free.
Each puzzle exercises one skill: pattern recognition (how a sequence grows), spatial reasoning (how shapes rotate and move), and logical deduction (combining rules to find the one answer that fits). The Train & Learn lessons teach each rule step by step before you practise it.
Your rating rises as you solve harder puzzles and dips when you miss, so improvement is something you can see on a graph — not a promise. MatrixMind doesn't claim to raise IQ or general intelligence; it offers honest, trackable practice at the reasoning puzzles themselves.
MatrixMind is a free educational practice tool suitable for all ages — built for learners, parents, and teachers who want screen time that's genuinely practice.
Three kinds of reasoning to practise — pattern logic, spatial rotation, and rule-overlay thinking. Every mode feeds one rating graph, so your progress is measurable, not a promise.
The classic 3×3 reasoning grid. Read how every row and column evolves — then work out which shape belongs in the empty cell.
Arrows and shapes spin 90° with every step. Track the rotation in your head and land the final position — no spinning your screen allowed.
Colorful polyomino planning. Fit the pieces, clear the lines, and think three placements ahead — spatial packing under pressure.
Every puzzle MatrixMind generates is built from a real piece of mathematics. Here is what is actually happening underneath.
The vertex puzzles are built on an arithmetic sequence — a list of numbers in which each term differs from the one before it by a fixed amount, known as the common difference. MatrixMind extends that idea across two dimensions at once: the number of sides on each polygon is set by its position in the grid, following sides = start + d × (row + column). Because a step to the right and a step downward both add the same common difference, the shape in the bottom-right corner always sits four steps away from the one in the top-left. Reading that rule off the board means recovering a linear relationship from nothing but its outputs, which is the same reasoning used in algebra when you fit a straight line to a table of values and solve for its slope.
A complete turn measures 360°, which makes rotation a natural home for modular arithmetic — the branch of number theory concerned with remainders and wrap-around. The spatial puzzles place each arrow at an angle determined by (start + column × step + row × shift) mod 360. The modulus is what makes the puzzle work: a rotation of 450° and a rotation of 90° leave a shape in precisely the same orientation, because 450 mod 360 equals 90. Solving one of these grids means composing two separate transformations — one running across the row, another running down the column — and then reducing their sum to a single position on the circle.
The colour-logic puzzles run on Boolean algebra, the two-valued logic that underpins every digital circuit and every conditional statement in computer code. Each quadrant of a cell is either lit or unlit — true or false — and the bottom row is produced by combining the two rows above it with a logical operator. OR returns true whenever either input is true. XOR, the exclusive or, returns true only when the two inputs differ, so a square lit in both rows cancels back to darkness. XOR is the parity operator, the same operation behind error-detecting checksums and the carry logic of binary addition. Working one of these puzzles backwards is, quite literally, reading a truth table off a picture.
Ratings here use the Elo system, which is applied probability rather than a running score. Before you answer, the engine computes an expected result from the logistic function E = 1 / (1 + 10^((Rp − Ru) / 400)), converting the gap between your rating and the puzzle's into a probability between 0 and 1 — a gap of 400 points corresponds to roughly a ten-to-one expectation. Your rating then moves by K × (result − E). Beating a far stronger opponent shifts it sharply because the outcome was improbable; beating a much weaker one barely moves it at all, because the result was already expected. The K-factor itself shrinks as ratings climb, so the curve flattens towards the top — diminishing returns, made visible on your own graph.
The Daily Challenge hands every player in the world the same three puzzles without any server deciding what they are, using a seeded pseudorandom number generator. An ordinary random function cannot be reproduced on demand; a PRNG is instead a deterministic algorithm that unfolds a fixed sequence of numbers from a single starting value called the seed. MatrixMind seeds it with the calendar date, so the same date always produces the same sequence, and therefore the same puzzles, on every device that computes them independently. It is a compact demonstration of a large idea in computer science: given identical input, a deterministic process must return identical output.
15 rated rivals from a bubbly 600 all the way to the unbeaten 4600 Legend. Race to 5 points — every win banks real Elo.
✨ Rotation Puzzle
7 guided units, from your first shape to the Boss Round — each one an animated demo, then real drills. Nothing here touches your rating.
A guided path of units — each one opens with an animated demo of the rule, then drills it for real. Nothing here touches your rating; this is the practice room.
Animated demos · Real drills · Always free
Jump into any unit — go straight for the one you want.
More units unlock as the engine grows — every puzzle type in Tactics becomes a lesson here eventually.
MatrixMind #—
Three puzzles — one from each discipline, easy to hard. One attempt per puzzle, against the clock. Every player on Earth gets the same three today, generated from the date itself.
Your rating isn't at stake here — the daily is all about the streak.
Your rating, streak, and history are saved in this browser (localStorage) — no account needed yet. Resetting starts you fresh at 1200.