Is FreeCell Always Winnable? The Math Behind the Myth

You have probably heard that FreeCell is almost always winnable. The full statement is more precise: of the original 32,000 numbered deals (deals 1 through 32,000) included in Microsoft FreeCell, exactly two are provably unwinnable with perfect play. Deal #11982 and deal #146692 cannot be solved regardless of the sequence of moves made. Every other deal in that set can be won if you play correctly.

Where the 99.999% stat comes from

When Microsoft FreeCell shipped with Windows, the game came with numbered deals. The original implementation (written by Jim Horne and later updated) included deal numbers roughly 1 through 32,000. Over years of play, players and researchers worked through these deals systematically.

The critical verification came from computer analysis. FreeCell's fully visible board makes it amenable to exhaustive search algorithms. A program can evaluate every possible move at every point in the game and confirm whether any path leads to a win. For almost every deal, such paths exist. For deals #11982 and #146692, exhaustive search confirmed that no winning path exists.

Later versions of Microsoft FreeCell expanded the deal set to over 1 million numbered deals. Within that larger set, a small number of additional provably unwinnable deals were identified, but the fraction remains extremely small, well under 0.01% of all deals.

How computer solvers proved this

FreeCell is a finite, fully observable game. All 52 cards are visible from the start. Every possible game state can be represented as a node in a search tree, with each legal move representing an edge to a new state. A solver explores this tree, keeping track of states already visited to avoid infinite loops.

When the solver has visited every reachable state from the starting position and none of them is a winning state (all 52 cards on foundations), the deal is proven unwinnable. This is not a statistical argument. It is a mathematical proof by exhaustive search. There is no move combination a human player could discover that the solver missed, because the solver checked all of them.

The original inventor of digital FreeCell, Paul Alfille (who wrote the first implementation for the PLATO system in 1978), studied the mathematical properties of the game extensively. His early analyses suggested that almost all FreeCell deals were solvable, a property he considered one of the game's defining features. The computer-verified results confirmed his intuition with precision.

What "winnable" actually means

A deal being "winnable" means there exists at least one sequence of legal moves that leads to all 52 cards on the foundations. It does not mean the deal is easy to win. It does not mean a human player will find the winning sequence. And it absolutely does not mean every sequence of moves will succeed.

Many FreeCell deals have exactly one winning sequence, or a very small number of them. A deal where only 3 out of billions of possible move sequences lead to a win is technically winnable, but in practice no human player working without computational assistance will find it without tremendous effort or luck.

This distinction matters because some players interpret "FreeCell is always winnable" as "you should always be able to win." That interpretation is incorrect. The correct interpretation is: "unless you are playing deal #11982 or #146692, the deal is not lost before you start, but you can certainly lose it through your own choices."

The correct interpretation is: unless you are playing deal #11982 or #146692, the deal is not lost before you start — but you can certainly lose it through your own choices.

Why FreeCell feels harder than Klondike despite being more solvable

Klondike's win rate for skilled players is around 43% on Turn 1 and around 83% with undo allowed and optimal play. FreeCell's win rate for typical human players without hints is often reported in the 70 to 80% range, despite 99.999% of deals being technically solvable.

The paradox resolves when you consider what each game hides. In Klondike, many cards are face-down. If the deal buries a critical card under five face-down cards that cannot be exposed, there is nothing skill can do. The deal kills you. In FreeCell, all cards are visible. There are no lucky flips. You cannot blame a bad deal. If you lose, it is because you ran out of moves after a sequence of choices that led to a dead end.

FreeCell puts the weight of failure entirely on the player's decisions. Klondike shares that weight with the deal itself. Players often find fully-visible games more mentally demanding because there is no external explanation for a loss.

There is also a planning depth requirement. Winning a complex FreeCell deal may require seeing 15 or 20 moves ahead, keeping track of which free cells you can afford to fill and when. Klondike rarely demands that level of forward planning because card reveals break the planning horizon naturally.

Paul Alfille and the origins of FreeCell

Paul Alfille wrote the first computer implementation of FreeCell in 1978 as a medical student at the University of Illinois at Chicago, programming it on the PLATO educational computer system. He used a card game description language called TUTOR to build the implementation and deliberately designed the game with all cards face-up to make it fully solvable by a patient player.

Microsoft developer Jim Horne later encountered Alfille's PLATO version and implemented it for Windows, shipping it with Windows 3.1 in 1992. The game became enormously popular partly because the fully visible board made every game feel fair: no random card flip could kill you.

The FreeCell board layout

Understanding the board structure is essential to understanding why almost every deal is solvable. Four free cells in the upper left provide temporary card storage. Four foundation piles in the upper right receive completed Ace-through-King sequences. Eight tableau columns hold all 52 cards at the start, all face-up.

Frequently asked questions

Which FreeCell deals are unwinnable?

In the original Microsoft FreeCell 32,000-deal set, deals #11982 and #146692 are provably unwinnable by exhaustive computer search. In the expanded million-deal set used in later versions, a small additional number of unwinnable deals exist, but the percentage remains well under 0.01%.

How was it proven that certain deals are unwinnable?

Exhaustive computer search. A solver explores every possible sequence of moves from the starting position. When no sequence leads to a win state, the deal is mathematically proven unwinnable. This is not a probabilistic argument but a complete enumeration.

If FreeCell is 99.999% winnable, why do I lose so often?

"Winnable" means a winning sequence exists, not that every sequence wins. Many FreeCell deals have only a small number of winning paths through billions of possible move combinations. Human players without computational assistance frequently miss the winning path.

Who invented FreeCell?

Paul Alfille wrote the first computer implementation in 1978 on the PLATO system. Jim Horne later brought it to Windows, where it shipped with Windows 3.1 in 1992.

Is FreeCell harder than Klondike?

From a pure solvability standpoint, FreeCell is easier: 99.999% solvable vs. 43–83% for Klondike. But players often find FreeCell more mentally demanding because all failure is attributable to their own decisions rather than the deal, and because winning complex deals requires deeper forward planning.