The Bell game's stranger sequel. Three players, sealed apart, win a cooperative game at most three times in four — a wall that isn't statistical but proven. Share a three-way entangled state and they win every round, forever. No message passes. It only looks like telepathy.
One room over, the Bell game showed two separated players beating the classical limit with entanglement — winning 85% of a game capped at 75%. This is the version that goes all the way. Now there are three players — Alice, Bob, Charlie — each sealed in a room, none able to signal the others. A referee hands each one a bit (x, y, z), promising only that an even number of them are 1. Each player answers with a single bit. They win the round together if their three answers contain an even number of 1s when all the questions were 0, and an odd number otherwise.
Plot in advance all you like. In any universe where each particle simply carries its answers with it, the best the three can do is win 75% of the time — and unlike the Bell game, the proof here is a single line of parity arithmetic, not an inequality. You can build a strategy in the page and watch it stall at three-in-four, every time.
Then give each player one particle of a GHZ state — three specks entangled together, prepared once and carried, unread, into the three rooms. Each player, depending on the bit they were handed, chooses which way to measure their particle. They win every single round. Not 99%. Not "on average." Every round, forever, with nothing passing between the rooms. Physicists call it quantum pseudo-telepathy, and it is real: a three-photon version was built by Pan, Bouwmeester, and Zeilinger in 2000.
The reason is the sharpest fact in the foundations of physics. The four things the entangled particles reliably deliver cannot all be true of objects with definite, pre-existing properties — multiply them together and you get +1 = −1. No averaging, no loophole, no "well, usually." This is Bell's theorem without a single inequality (Greenberger–Horne–Zeilinger 1989; Mermin 1990), and the last instrument lets you try, and fail, to beat it by hand.
Run the classical game and the bar parks at the amber line: 75%, never higher. Switch to entangled (GHZ) and it fills the whole bar — 100%, and the "rounds lost" counter never moves. The per-input cells tell the difference at a glance: a classical plan wins three inputs and is forced to lose the fourth, while the quantum strategy wins all four. The next instrument shows why the fourth loss is unavoidable.
With no way to talk, each player's answer depends only on their own bit. Pick what each says on a 0 and on a 1. The table shows, live, whether your plan wins each of the four promised inputs. Try to win all four.
The proof you can't. Add the four win-conditions together (XOR). On the left, each player's two answers each appear in exactly two of the four rows, so everything cancels — the left side is always 0. On the right, the required parities are 0, 1, 1, 1, and 0⊕1⊕1⊕1 = 1. So the four conditions demand 0 = 1: they can never all hold at once. At most three can — 75% is the exact ceiling for any classical plan, with or without shared dice.
So a classical universe must lose at least one input in four. The entangled players lose none. They are not sending signals — each player, alone, still sees a fair coin. They win because the particles deliver a set of correlations that, the moment you try to explain them with particles that have definite values, contradict themselves.
Each player measures their particle one of two ways — call them X and Y — each giving +1 or −1. The GHZ state guarantees four facts about the products of the three outcomes (right column, cyan — recomputed live). A local-realist world would have to assign each particle a definite X-value and Y-value in advance. Set those six values below and see how many of the four facts you can match.