Connect Four has a correct answer.
Not a best strategy. Not a rough guideline. A provably correct answer for every position on the board. A sequence of moves that, played from the start, guarantees a win for the first player no matter what the second player does.
The game was solved in 1988. Here's what that means.
What "Solved" Means
A game is strongly solved when a computer has analyzed every possible position and determined the optimal outcome — win, loss, or draw — from any state, with perfect play from both sides.
Chess is not strongly solved. The number of possible chess positions exceeds the atoms in the observable universe. No computer has mapped it completely.
Connect Four is small enough that it can be solved. The board has 7 columns and 6 rows. A maximum of 42 discs. The total number of legal positions is around 4.5 trillion. Large, but tractable.
Strong solution means: if both players play perfectly, the outcome is determined before either player touches the board.
The 1988 Breakthrough
Two researchers solved Connect Four independently and almost simultaneously.
James D. Allen completed his analysis first, in October 1988. He used a method called threat analysis — identifying the positions where each player could complete a four-in-a-row, then reasoning about which threats were reachable and which cancelled each other out. His result: the first player wins with perfect play, starting with the center column (column 4).
Victor Allis arrived at the same result independently in 1988, publishing his findings in a master's thesis at Vrije Universiteit Amsterdam. Allis used a knowledge-based approach — encoding strategic rules about threat pairs, odd/even parity, and zugzwang into a solver that could reason about positions without searching the full game tree.
Same answer from two different methods. First player wins.
How the Solution Works
Neither Allen nor Allis searched every possible game. The full game tree is too large even for 1988 hardware. Instead, they used pruning — cutting away branches of the search tree that don't need to be explored.
The key technique is alpha-beta pruning. A regular minimax search checks every possible move and counter-move. Alpha-beta pruning keeps track of the best result each player is guaranteed to achieve. When a branch can't possibly improve on what's already been found, it gets cut. In practice, this reduces the number of positions to search by a factor of millions.
On top of pruning, both solvers used strategic knowledge to reduce the search further. Allis encoded 9 rules about when certain positions were automatically winning, drawn, or losing — independent of the full game tree. This let the solver recognize critical positions instantly instead of searching ahead.
The result: a complete map of Connect Four, compressed into a form a computer can evaluate in real time.
John Tromp's Complete Solution
In 1995, John Tromp took the solution further. He computed the complete value of every legal Connect Four position — all 4.5 trillion of them — using a more powerful approach based on bitboard representations and exhaustive retrograde analysis.
Tromp's database doesn't just say "first player wins from the opening." It says, for any board position you place in front of it, whether that position is a win for the current player, a draw, or a loss, and what the optimal move is. Every position. No gaps.
His dataset has been used as the reference truth for Connect Four ever since. The Allen vs Tromp match — six games played between the two solvers in 1995 — was built on this foundation. Two people who knew the game's mathematical structure, playing at its outer edge.
What the First Move Proves
The first player's winning move is column 4 — the center.
That's not a heuristic. It's a proof. Every other opening move either loses or draws with perfect counter-play. Column 4 is the only first move that guarantees a win if followed correctly.
The second player's best response is also known. And the third player's best response. The entire principal variation — the line where both players play perfectly from move 1 — is a forced win for the first player.
How many moves does it take? The forced win is at least 41 moves in some lines, fewer in others. It depends on how the second player fights. But the result never changes.
Why This Doesn't Make the Game Trivial
Knowing that Connect Four is solved doesn't mean every game is predictable.
The solution requires perfect play. Perfect means not a single mistake across every one of those 41+ moves. No human plays perfectly. Even very strong human players deviate from optimal play regularly — sometimes because they don't see the right continuation, sometimes because the optimal line is counterintuitive.
The search depth required to navigate the solution during a game is enormous. The winning lines are correct but not always natural. Many positions that look fine to an experienced player are actually lost. Many positions that look scary are still drawable or winning.
This is why the game is still competitive at high levels. The solution exists in theory. Humans live in practice.
What It Means for the Second Player
The second player cannot win with perfect play from the other side. That's the proof.
But the second player can draw if they respond correctly. And drawing against a first player who doesn't know the perfect continuation is a real skill.
The odd/even threat strategy is part of this. P2's draw technique — building a counter-odd threat to nullify P1's winning threat — is a practical expression of the theory. If P1 deviates from the optimal line, the draw becomes a win for P2. Good second-player play means knowing when that moment arrives.
Where Solvers Come From
The open-source solver ecosystem traces directly back to this research. Tools like Pascal Pons' connect4.gamesolver.org use the same alpha-beta pruning techniques Allen and Allis pioneered, now running in milliseconds on modern hardware. Enter any board position and get the exact evaluation instantly.
FourRow's engine is built on the same principles. It plays at full solver strength — perfect play from any position. Use it to test your opening preparation, to practice defense as the second player, or to find the exact refutation of a line you've been struggling against.
The engine doesn't just play the best move. It knows why it's the best move. When you review a game after playing the engine, you can trace exactly where a position turned and what the correct continuation was.
The Practical Takeaway
Three facts worth carrying into your games:
1. Play column 4 first. It's not just the strongest practical move. It's the only provably winning first move. Every other column is either a draw or a loss with perfect counter-play.
2. Perfect play never loses as P1. If you're the first player and you make no mistakes, you win. Every mistake you make is an opportunity for the second player to draw or win. Every mistake they make while you're ahead just extends the inevitable.
3. The game is deeper than it looks. A solved game doesn't mean a simple game. It means the game has a correct answer. Finding that answer in real time, under time pressure, against a smart opponent — that's the game. That's what skill means here.
Want to test yourself against perfect play? Play the engine at full strength and see how far you get. Then review the game to find where you drifted from optimal. The theory of the solution shows you what's possible. Playing into it shows you where you actually are.
More on the math: the odd/even threat guide covers the parity theory that underlies the endgame, and the Allen vs Tromp match shows what these ideas look like in actual games between two solvers of the game.