← Back to Blog

In April 2007, a research team at the University of Alberta led by Jonathan Schaeffer published a paper titled "Checkers Is Solved." After eighteen years of computation, their program Chinook had proved that the game of American checkers, played perfectly by both sides from the standard opening position, always ends in a draw. The result made checkers the most complex game ever solved — and it will likely hold that distinction for a very long time, because chess, the next obvious target, is almost certainly unsolvable by any technology we can foresee.

What "solved" actually means

A game is weakly solved when we know the outcome of perfect play from the standard starting position. Checkers is weakly solved: perfect play by both sides results in a draw. This does not mean that every possible checkers position has been evaluated — it means that from the opening position, we know the value (win, loss, or draw) of every move, and every response, and every response to that, all the way to the end of the game.

A game is strongly solved when we know the value of every possible legal position, whether or not it could arise from the starting position. Connect 4 is strongly solved. Tic-tac-toe is strongly solved. Checkers is not strongly solved — only weakly — but even that required extraordinary computation.

Why checkers was solvable

Checkers has approximately 500 billion billion (5 × 10²⁰) possible positions. That sounds impossibly large, and it is — no computer could search every one. What Schaeffer's team did was use endgame databases (precomputed solutions for positions with ten or fewer pieces) to anchor the search, and then work backward from solved endgame positions to prove the value of the opening.

Several features of checkers made this tractable. The game has only one type of piece (plus kings), pieces can only move forward (until promoted), captures are forced, and the number of legal moves per position is relatively small. These constraints limit the game tree and make it possible to prune enormous branches without examination.

The forced-capture rule was especially important. In checkers, if you can jump, you must. This drastically reduces the number of choices at each turn, shrinking the effective game tree to a size that aggressive search techniques could handle.

Why chess is different

Chess has roughly 10⁴⁷ possible positions — a number so much larger than checkers that direct comparison loses meaning. If you could evaluate a trillion positions per second, solving chess would still take longer than the age of the universe. The number is not just big; it is computationally intractable by any known method.

Several features of chess create this complexity. There are six different piece types, each with different movement rules. Pieces can move in any direction. There are no forced moves (except when in check). A single position can have thirty or forty legal moves, compared to checkers' average of around eight. And the game can last hundreds of moves, with complex middle-game positions that resist simplification.

Even endgame tablebases — the technique that anchored the checkers solution — illustrate the gap. As of 2025, chess endgame tables are complete only for positions with seven or fewer pieces. The seven-piece tables required 140 terabytes of storage. Eight-piece tables are estimated to need petabytes. Getting to the full thirty-two pieces that start on the board is not merely difficult — it exceeds any plausible technological advance.

What chess engines actually do

Since chess cannot be solved, engines like Stockfish and Leela Chess Zero use heuristic evaluation and deep search to play at superhuman levels without knowing whether any given position is truly won, lost, or drawn. They are incredibly strong — far stronger than any human — but they are not perfect. They occasionally disagree with each other, and there are positions where the correct evaluation is genuinely unknown.

This is what makes chess perpetually interesting. It is too complex for certainty, so there is always room for discovery, creativity, and surprise — even for computers.

Other solved games

Connect 4 was solved in 1988: the first player wins with perfect play. Gomoku on a standard 15×15 board is solved: the first player wins. Nine Men's Morris is solved: it is a draw. Hex on boards up to 9×9 is solved: the first player wins. Awari (a variant of Mancala) is solved: it is a draw.

The pattern is interesting: most solved games are draws. This makes intuitive sense — a perfectly played game between two perfect opponents usually results in neither side being able to force a win. The exceptions, like Connect 4, tend to have a significant first-mover advantage built into their rules.

Try both

Play Checkers on Hard difficulty and experience a game where the computer can play perfectly. Then try Chess on Hard difficulty and play a game where even the strongest engine in the world is still, in a mathematical sense, guessing. Both are excellent games — but only one still has secrets to give up.