Chess, unlimited

Free means free

The biggest chess site caps free accounts at a few puzzles a day, one shallow game review, and locks its stats tools behind a $119/year subscription. This page is our answer: live games with auto-matchmaking, private rooms with a shareable code, three computer levels, and a coach that suggests moves and calls out mistakes while you play — all free, all unlimited, no account. The board, pieces, and sounds are generated in your browser; there's nothing to install and nothing to buy.

How the coach works

Toggle 🎓 Coach and the built-in engine (the same one behind the Hard bot) suggests a move each turn with a blue arrow, then grades what you actually played — good move, inaccuracy, or blunder. It's an honest club coach, not a grandmaster: perfect for learning patterns without drowning in engine lines. Want a full post-game breakdown instead? Game Review analyzes any of your chess.com games with Stockfish, free and unlimited.

Cheating, and exactly how far we can see

The loudest argument in online chess is not about ads or accounts — it is about engine assistance. The largest site closes on the order of three thousand accounts a day for fair-play violations, and the public disputes are rarely about whether cheating happens; they are about whether the accusations are just. So here is our position, including the parts that are not flattering.

What we do

We never help you against a person. The coach and the evaluation are refused in game logic during a human game, not merely hidden from the interface — the engine is not asked for a move it could leak.

We measure the moves, not your machine. After a live game, both players see the same panel: how often each side played the engine's choice, counted only on hard positions — the ones where a shallow, greedy look disagrees with a deeper search. Matching the engine on a forced recapture proves nothing, so forced moves are excluded. A strong club player lands roughly 30–45% of hard moves; a sustained 75% is not a human pattern. That statistic is computed from the move list, which both players hold, so it cannot be faked by the person being measured.

It refuses to judge a small sample. Under eight hard positions it reports “not enough moves to say anything” and scores nothing. The harshest wording needs twelve. Nobody is accused mid-game, ever.

What we cannot do — including against our own tools

We also collect behavioural signals: whether the tab was hidden while you were thinking, whether your moves came from a real pointer, and whether your thinking times are suspiciously flat. Every one of those is reported by the suspect's own browser, so a determined cheater can simply lie about all of them. They can raise suspicion here; they can never be the only reason for it.

And they fail completely against a specific, entirely practical attack: a program that reads the board off the screen as pixels and plays the engine's move with the real cursor. It never hides the tab. It generates genuine pointer events. It has nothing to paste. Every behavioural signal we have is blind to it — and we build tools like that ourselves, so we are not speculating.

What survives is the statistic, because it does not care how a move was chosen — only that engine-quality moves keep appearing in positions where finding them is hard. But that too has a floor: a cheater who throttles the engine to club strength and lets it make genuine mistakes produces the same distribution as a human of that strength, and there is then nothing left to detect. That is a mathematical limit, not a gap in our code, and no site has solved it. Detection catches the careless.

Which is why the honest version of this page is not “we stop cheaters”. It is: we never help your opponent, we show you both the same numbers, we refuse to guess from a short game, and we tell you where our vision ends. There is also no rating here worth cheating for — which is, unglamorously, the most effective anti-cheat measure on this site.

So what actually happens when the numbers look wrong?

The game stops counting. That is the whole penalty. When a game's numbers clear our threshold, the result is marked disputed and neither rating moves — in either direction, whoever won. Nobody is banned, nobody is accused, and there is nothing to refund, because the points were never taken off you in the first place.

That is a deliberate choice and it follows from the two paragraphs above. If detection reliably catches only the careless, then handing out bans and reversals means confidently punishing the clumsy while the careful keep everything they won — and dressing it up as justice. We would rather take the one action we can defend: decline to score a game we cannot score honestly.

Two consequences worth stating plainly. A false positive costs you a rating point and nothing else — no mark on your account, no appeal to lose. And because either player's browser can raise the flag, someone losing a game could fabricate one to avoid the drop; they could already do that by simply reporting the result wrongly, which voids the game the same way. On a free ladder with no prizes, we think a system that cannot be abused into hurting anyone beats a system that punishes hard and sometimes punishes the wrong person.

The threshold itself needs two independent signals to agree — engine-quality moves in hard positions and something else, such as flat thinking times. Move-quality alone, however extreme, never gets there on its own, and a game too short to judge can never be voided at all.

Play with a friend anywhere

Create a room, send the 4-letter code, and your friend joins from any phone or laptop browser — 5-minute clocks, resign and draw offers included. Quick Match pairs you with whoever's in the queue, and if nobody's around within 30 seconds a Medium bot steps in so you always get a game.

Chess is a public-domain game. This implementation, its artwork and its engine are original to EveryGameMade. Want the numbers behind other classics? See our data-backed strategy guides or play MOGUL, our free property-trading board game.