FreeCell by the numbers: almost every deal can be won

FreeCell deals all 52 cards face up into eight columns and gives you four "free cells" to park cards in while you sort them out. Nothing is hidden and there is no stock, so every game is a pure puzzle.

How winnable is it? Our solver played 10,000 FreeCell deals and solved 99.68% of them. It did not prove a single one impossible; the remaining 0.32% simply ran past its search budget. Published research, based on far larger samples, puts the winnable share at 99.9989%. In plain terms: if you lose a game of FreeCell, the deal was almost certainly winnable.

The more interesting question is what makes it winnable, and our data answers that clearly. It's the free cells. Take even one away and the game changes completely.

At a glance: FreeCell in numbers

What each free cell is worth

We re-ran 3,000 deals with fewer free cells to see how much each one matters. Because some deals ran out of search budget, each result is a range: the lower figure is "proven solvable", the upper adds the undecided deals.

Losing the fourth cell barely matters. Losing the third costs about a fifth of all games. With one cell, four in five deals are lost, and with none the game is almost unplayable. That curve is the best strategy lesson we can offer: every free cell you fill during play puts you further down this table.

FreeCell rules on this site

• Eight columns: four of seven cards and four of six, all face up.
• Four free cells, each holding one card.
• Four foundations, built up by suit from Ace to King.
• Build down the columns in alternating colours: a red 9 on a black 10.
• Any card can go into an empty column.
• You can move several cards at once. The game allows (empty free cells + empty columns + 1) cards in one move. This is only a shortcut: the same move could always be made one card at a time.

Note that our sister game Baker's Solitaire looks almost identical but builds in suit and only accepts Kings in empty columns. Our data shows that turns a 99.7% game into a 31.8% one.

Strategy, backed by the numbers

Treat free cells as a budget, not storage

The table above is really a table of how much room for error you have. Park a card in a cell and, for as long as it sits there, you are playing a harder game. Before you use a cell, know what move will empty it again.

Half the time you start without an Ace

Only 49% of deals offer an Ace on the first move. If yours isn't one of them, don't grab the first move you see. Find the shallowest Ace (the one with the fewest cards on top), count how many cells and empty columns you need to reach it, and work towards that.

An empty column beats a free cell

On this site an empty column adds to how many cards you can move at once, just as a cell does, and it can also take a whole run or a King. Clearing a short column early is usually worth more than keeping a cell spare.

If you lose, look for your mistake, not bad luck

We found zero impossible deals in 10,000. That does not mean they don't exist; the published estimate is about 11 in a million. But it does mean the chance your particular loss was unwinnable is tiny. Unlimited Undo lets you replay the critical moment and find the better line.

Solutions are shorter than they feel

Our solver's first-found solutions averaged about 62 moves, with automatic foundation moves included. A game that drags well past that is often a sign of shuffling cards between cells without a plan.

A worked example of the move limit

The multi-card move rule is where many players leave wins on the table, so it's worth a closer look. On this site you can move (empty free cells + empty columns + 1) cards at once.

Two practical points follow. First, when the game refuses to move a long run, it's telling you that you've spent too much of your space. Free a cell or clear a column and try again. Second, the limit is only a convenience. As our solver's method notes explain, any group move can also be made one card at a time, so a run that "won't move" can often still be shifted in stages if you have somewhere to put the cards temporarily.

This is also why our reduced-cell results fall so steeply. Fewer cells mean smaller group moves, and smaller group moves mean long runs get stuck where they are. At one free cell only 17–20% of our deals were winnable at all.

Playing here

FreeCell is a good match for a rainy weekend or a quiet half hour after work. On free-solitaire.nz it is free, with nothing to download, install or sign up for, and no account or email needed. It runs in any modern browser.

• Large, clear cards with zoom.
• Hint, unlimited Undo and Restart.
• Auto Play once the game is won.
• Drag and drop, or tap a card and then where it should go.
• Wins, time and moves tracked for your current game.
• Sound off, fullscreen and table colour options.

How we did the research, step by step

The free-solitaire.nz team ran this study in September 2026. FreeCell is famous for being almost always winnable, and we wanted to check that for the version on this page, and to measure what each free cell is worth.

1. Reading the rules from this site’s own game code. Eight columns (four of seven cards, four of six), all face up; four free cells; build down in alternating colours; any card may go into an empty column. The game lets you drag a run of several cards, but that never changes whether a deal can be won, because the same move can always be made one card at a time.
2. Rebuilding the game as a program. We rewrote FreeCell in Python, with the same moves and the same foundations.
3. Dealing the cards. Deals 1 to 10,000, each from a numbered seed. These are our own numbers, not the old Windows deal numbers, so our deal 1 is not the Windows deal 1.
4. Choosing the player. Every card in FreeCell is face up, so one method is enough: a solver that always explores the most promising position next and only makes automatic moves to the foundations when they are provably safe. Each deal got up to 200,000 positions.
5. Running it. Deals still undecided after the first pass were run again with up to 1,000,000 positions. In the end 99.68% were solved, none were proven impossible, and 32 of the 10,000 were still undecided. We then ran 3,000 deals each with 3, 2, 1 and 0 free cells.
6. Checking the solver twice. When the solver said a deal with fewer cells was impossible, we re-checked a sample with a completely separate brute-force program. From the 3-cell test we re-checked 2 deals, and both programs agreed neither can be won.
7. Checking against published research. Blake & Gent, Journal of Artificial Intelligence Research, 2026 report 99.998881% for four cells, and 99.3583%, 79.544%, 19.348% and 0.2137% for three, two, one and no cells. Every one of those published figures sits inside our ranges.
8. Working out the margin of error. A “95% range” is the band the true figure falls inside 95 times out of 100 if you ran the same test again on fresh deals. For four cells, 99.68% solved came with a range of 99.55–99.77%.
9. What the numbers cannot tell you. The site lets a card come back down from a foundation; our solver never uses that move, so if anything our figures are slightly low. And “undecided” means the search ran out of room, not that the deal is lost.

Every deal is seeded, so the whole test can be repeated exactly.

The technical side

The code is Python 3.11, standard library only. FreeCell is handled by freecell.py, the same module that plays Baker's Solitaire. A shared runner.py farms deals out to 19 worker processes with Python's multiprocessing, and every deal is shuffled from its own numbered seed.

• The solver is a best-first search. At every step it expands the most promising position first, scored by how many cards are not yet home plus how many are resting on a lower card. It stores each position in a table in a standard order, so the same layout reached two ways is only searched once. Safe automatic moves to the foundations cut out pointless branches.
• Budget and re-runs. Each deal got 200,000 positions. The 10,000 four-cell deals took about 18 minutes; deals still undecided went through retry.py with 1,000,000 positions each. The 0- to 3-cell runs of 3,000 deals took between 1 and 25 minutes each.
• The 95% range is a Wilson score interval.

As a cross-check, a separate brute-force program (verify_freecell_bruteforce.py) that makes single-card moves only re-tested deals the main solver had called impossible: seeds 1 to 5 with one free cell, and seeds 739 and 1053 with three. It agreed every time. An AI assistant (Claude, by Anthropic) helped write and review this code and draft the page. It did not supply any results: each figure comes from running the code, and the team checked them against the published research above.

By Admin, free-solitaire.nz. Study date: 18 September 2026.

Frequently asked questions

Can every FreeCell game be won?

Almost. Our solver solved 99.68% of 10,000 deals and proved none impossible. Published research estimates that about 99.9989% are winnable, so a tiny number are not.

How many free cells do I really need?

Four gives you almost every deal; three still wins the large majority; with two, roughly one deal in four or five is lost; with one, most deals are lost.

Is FreeCell easier than Klondike?

Easier to win, harder to play well. In Klondike hidden cards and the stock decide a lot; in FreeCell almost every loss is a decision.

How long is a typical game?

Our solver's solutions averaged about 62 moves. Most people take a little longer while they think things through, and that's the fun of it.

Does the multi-card move limit make some deals impossible?

No. The limit only decides how many cards move in one drag. Any group move can be made one card at a time using free cells and empty columns, so the limit changes how quickly you play, not which deals can be won. That is why our solver could ignore it.

Are these the same deal numbers as the old Windows FreeCell?

No. Our deals come from our own numbered seeds, so "deal 11982" in our study is not the famous Windows deal. The figures above describe random deals in general.

What should I play after FreeCell?

Try Baker's Solitaire if you want the same layout with a much steeper challenge, or browse the games page. More results are on the blog.