Puzzle meets YouTube

Puzzle videos

Videos around our puzzles — where they came from and the ideas nearby.

Latest

"Only 3 left" and "5 minutes to go" — those messages are deliberate [Dark patterns and design ethics]

"Only 3 left." "5 minutes to go." "20 people are viewing this." A 2019 crawl of 11,000 sites found these devices on roughly 11% of top shopping sites. Dark patterns are good design turned inside out — the same knowledge, pointed the other way.

Recommended

"Solve This and Win $1,000" — the Mathematically Impossible Puzzle That Gripped America 140 Years Ago [15 Puzzle]

The 15 puzzle that gripped America 140 years ago. Sam Loyd offered $1,000 to anyone who could swap just the 14 and 15 tiles — but that arrangement can never be solved. Here's why, told through the puzzle's history and the idea of parity.

People can't play rock-paper-scissors at random — the habit experiments revealed, and what it teaches game design

The optimal strategy in rock-paper-scissors is a perfectly random one-third each. Experiments show otherwise: winners repeat, losers shift toward the hand that beat them, and the whole group cycles in one direction. That habit isn't weakness — it's evidence of learning.

That vast world isn't stored anywhere — procedural generation, and the 1985 paper behind it

The sprawling terrain in a game world isn't stored — it's computed the moment you approach. How Ken Perlin's 1985 paper resolved the contradiction of "random yet smooth," and why a single seed can reproduce infinity.

The shortest solution is 20 moves — so why do people take 40? [A path you can follow beats an answer you can't memorize]

Any scrambled cube can be solved in 20 moves — proven in 2010. But with 43 quintillion states each needing its own sequence, no human can memorize them. The layer-by-layer method trades optimality for something better: a process you can actually learn.

A fair board is a boring board — bias is what makes "where to buy" a decision

On a 40-square board game, one square gets landed on about 4.7 times more often than average — the culprit is forced movement. But that bias is exactly what gives properties different values, and turns buying into a real decision.

Your opponent was chosen for you — the history of rating systems and what they were really for [Elo → Glicko → TrueSkill]

From Elo to Glicko to TrueSkill in 2006. The real goal of a rating system was never ranking — it was minimizing the number of matches needed to learn how strong you are, so a good match arrives sooner.

2048 isn't a game of luck — the proof that optimal play wins 99.9% of the time, and the 9-to-1 ratio that sets the difficulty

In 2023, optimal play in 2048 was proven to win at least 99.9% of the time. What makes it hard is a single number — 90% twos, 10% fours. Good randomness doesn't decide the outcome; it nudges your plan just enough.

About 18% of Solitaire deals are unwinnable from the start — it may not be your fault

A 2019 solver study put Klondike's winnability at about 82% — meaning 18% of deals can't be solved even with every card visible. Knowing that a loss was yours to own is what makes improvement possible, which is why modern games deal only solvable hands.

Last place gets the better items on purpose — rank-based drops and the design of not giving up [Game design]

Racing games quietly change the item pool based on your position. The real aim of this "rubber banding" isn't to help whoever is losing — it's to keep everyone, leader included, paying attention to the end.

The best move in Memory is to turn nothing over — the hidden strategy mathematics found [Game design]

When Zwick and Paterson analysed Memory in 1993 under perfect recall, they found that in some positions the best move is to flip nothing at all. If taking everything available is always right, it's just labour — the dilemma is what makes it a game.

One tile shape that never repeats — the "hat" that solved a 60-year problem [The 2023 discovery]

A single tile shape that fills the plane completely yet never repeats. Mathematicians hunted this aperiodic monotile for over 60 years; in 2023 an amateur in the printing trade found it. Also: what a never-repeating pattern is good for in backgrounds and boards.

Why amnesic patients dreamed of falling blocks — the brain's two memories and game design [The Tetris effect]

In a 2000 study, amnesic patients who could form no new episodic memories still saw the falling blocks — without remembering having played. The gap between declarative and procedural memory is why a great first level needs no manual.

Tridem

The single screenshot that made Wordle a phenomenon — three conditions for a result card that spreads for free [Game design]

Wordle's result card spread worldwide on zero ad spend. We break down why into three conditions — no spoilers, readable at a glance, tells a story — then design Tridem's own share card live, from a bad first draft to options A and B.

The Real Reason Loading Screens Stayed "Boring" for 20 Years — One Patent Held the Industry Hostage [Game Law]

Playable loading-screen minigames were locked up for two decades by a patent Namco filed in 1995. Here's the legal reason loading screens stayed "boring" until it expired in 2015.

Gomoku Is a Proven First-Player Win — Yet the Reason People Still Play It Is Brilliant [Renju/Game Design]

Gomoku without forbidden moves has been mathematically proven to be a first-player win. Here's why it's still played anyway — through the history of that proof and the origins of Renju's rules.

Why You Can Still Jump After Leaving the Ledge — the Invisible Kindness of "Coyote Time" [Game Design]

Even after stepping off a ledge, you can still jump for a few frames — it is not a bug, it is a design feature called "Coyote Time."

Gacha — the invention of the "pity ceiling": how design changed after the complete-gacha ban

After Japan banned complete gacha, how did gacha design evolve? A history of probability and desire, told through the invention of the "pity ceiling."

16-Clue Sudoku

Why Sudoku can't go below 17 clues — a 7-million-hour supercomputer proof

The minimum number of clues a uniquely-solvable Sudoku can have is 17 — and proving it took a supercomputer about 7 million hours. The story behind our 16-clue Sudoku, which can never have a unique solution.

Cointossweeper

Was Minesweeper just luck?

Is Minesweeper really won or lost on luck alone? Chasing that question is what led to Cointossweeper.

Tetris is mathematically doomed to lose

Why an endless run of only S and Z pieces makes Tetris mathematically unwinnable — the idea behind Tetris for Masochists.