Checkers Endgame Database

Every position with four pieces or fewer, solved exactly · Updated August 9, 2026

Most checkers endgame advice is a rule of thumb: two kings usually beat one, king against king is a draw. This page replaces the guesses with a solved database — 6,408,836 positions, each with an exact result and an exact distance to the end. Every ending below can be played out on the board, move by move, as perfect play by both sides. Nothing to download.

The longest forced win: 55 moves

This is the hardest position in the entire database. Two plain men against two plain men — no kings, nothing that looks decisive — and Red, to move, wins by force in 55 moves. That is 109 half-moves in which any deviation throws the win away.

No engine with a clock will ever find this. A search that looks ten moves ahead is not within sight of the end; it sees a balanced position and calls it a draw. The only way to know is to have the answer written down, which is what a database is.

What is in the database

Every position with one to four pieces is in there, in every legal arrangement of men and kings, as long as both sides still have something on the board. The rules are American checkers — see the complete checkers rules if you need them.

The classic endings, solved

Each card below is the hardest case of its kind: of all positions with that material, the one that takes the most moves to win. That is a deliberate choice, and it is not the same as a "typical" position. Beat the hardest case and every other position with those pieces is easier. It is also a fact from the file rather than someone's opinion about what makes a good example.

Press Play and both sides play perfectly: the winner takes the fastest win available, the loser puts up the longest possible resistance. That is why the moves can look strange — the defender is not blundering, it is surviving as long as arithmetic allows.

Test yourself: the only move

Sixty exercises taken straight out of the database. In every one of them exactly one move preserves the result and every other move throws it away — that is not an opinion about the position, it is what the table says.

Most of them ask you to hold a draw, which is the half of endgame play that almost never gets taught. It is also the half no other puzzle collection can set: a search program cannot prove a draw, it can only fail to find a win, and those are different statements. Only a complete table knows the difference.

Difficulty here is read, not guessed. For a win it is the distance to the end — "win in 3" and "win in 23" are not the same exercise. For a draw there is no distance, because a draw does not end, so the ruler is how long the punishment takes: if the quickest refutation of a wrong move finishes in two moves anybody sees it, and if the fastest loss still takes fifteen, the mistake is invisible over the board. Both numbers come out of the file.

Three things the table says that the rules of thumb do not

1. Two kings against one is not "usually" a win — it is always a win. Zero drawn positions out of all of them, and the worst case is 17 moves. Most endgame guides hedge here; there is nothing to hedge.
2. King against king is drawn only two thirds of the time. A third of those positions are won by whoever is to move. The rule everyone repeats is a good default and a bad certainty.
3. Men are slower than kings, by a lot. Two men against one takes up to 31 moves; two kings against one takes 17 — same material edge, nearly double the work. The race to crown is not just about strength, it is about time.

How a database like this is built

Backwards. You start from the positions that are already over — one side has nothing left — and work outwards: a position is won if some move leads to a position that is lost for the opponent, and lost if every move does. Repeat until nothing changes. What remains, still unlabelled, is drawn.

The subtle part is the distance. Storing only win, loss or draw is not enough to play the ending: every winning move keeps the win, so a table without distances shuffles forever without making progress. Each entry therefore stores how many half-moves remain, the winner picks the shortest, and the loser picks the longest. That is exactly what happens when you press Play above.

This database was generated with the same rules engine that runs the live board and validates every move in online games — no separate "backwards move generator" was written for it, since new untested code at the most delicate point is how a table ends up confidently describing the wrong position. It is verified from four independent directions, including an exhaustive re-check that every one of the 6.4 million entries follows from its own successors, and a test in which the table plays itself and must deliver the win in exactly the promised number of moves.

How this compares to Chinook

Credit where it belongs: Chinook, from the University of Alberta, is the reference work. That team solved checkers outright in 2007 — proving the game is a draw with perfect play — with endgame databases reaching ten pieces and hundreds of billions of positions. Four pieces is a small corner of that.

What this page offers is not size but immediacy. Chinook's databases are a multi-gigabyte download; here the answer is on the page, the table is 1.8 MB and only fetched if you press Play, and every ending can be watched rather than looked up. If you want the authoritative science, go to Chinook. If you want to see two kings actually beat one, it is above.

Questions about checkers endgames

Is two kings against one a win in checkers?

Always. Not one of the two-kings-against-one positions in the database is drawn, and the hardest of them is won in 17 moves. If you have two kings against one and the game ends in a draw, the win was there and it was missed.

Is king against king a draw in checkers?

Usually, but not always: 66 % of king-against-king positions are drawn, which means a third of them are won. Those are the positions where the side to move can simply jump the other king, or force it against an edge within a few moves. The longest of those wins takes six moves.

What is the longest forced win in a checkers endgame?

With four pieces or fewer, 55 moves — and it is two plain men against two plain men, not a position full of kings. Every one of those 109 half-moves is forced in the sense that any other move throws the win away. No search with a time limit will ever see the end of it; only a table can.

What is a checkers endgame database?

A table holding the exact result of every position with a given number of pieces — win, loss or draw, and how many moves it takes. It is not an engine's opinion: the values are computed backwards from positions that are already over, so they are proofs. This one covers every position with four pieces or fewer, 6,408,836 of them.

How is this different from Chinook?

Chinook is the reference and it is far larger — the University of Alberta team solved checkers outright, with databases up to ten pieces, and proved that perfect play is a draw. This database is much smaller at four pieces. What it offers instead is immediacy: it answers in the browser, with nothing to download, and every ending can be played out on the board.

Do I have to download anything?

No. The verdicts are on the page. The table itself is 1.8 MB and is only fetched when you press play on an ending, so reading the page costs nothing extra.

The checkers solver uses this same database automatically: set up any position with four pieces or fewer and it will answer exactly instead of searching. For the phase before the endgame, see checkers strategy, and for a daily test there is the daily checkers puzzle. If you want an opponent, play checkers against the computer — though note that the playing engine does not use this table, which is exactly why it can be outplayed in a long ending.