FreeCell: almost every deal can be won, and we checked

FreeCell deals all 52 cards face up into eight columns and gives you four "free cells", each able to hold one card. There is no stock and nothing hidden. That makes it the ideal game for a scientific question: if you could play perfectly, how many deals would you win?

The best answer comes from Blake and Gent at the University of St Andrews, who ran around ten million deals through their solver and found that 99.998881% can be won. Only 110 of their deals could not. We built our own solver for this site's FreeCell, dealt 10,000 games, and then took away free cells one at a time to see how quickly the game becomes impossible. Every one of our ranges contains the published figure.

What we found, beside what St Andrews found

Our range runs from the share our solver proved winnable to that share plus the deals it could not decide in time. The one small tension is at three cells, where we proved 0.97% of deals impossible (95% range 0.67–1.39%), slightly more than the published figure implies. We re-checked two of those deals with a completely separate brute-force search, and both really are impossible, so we put the difference down to the luck of which 3,000 deals we happened to sample.

The table is also the clearest lesson FreeCell has to teach. Losing one cell barely matters. Losing two turns a near-certain win into roughly four in five. Losing three leaves you with about one deal in five.

The FreeCell rules on this site

• All 52 cards face up in eight columns, the first four holding seven cards and the rest six.
• Four free cells, top left, each holding one card.
• Four foundations, top right, built up by suit from Ace to King.
• In the columns, build down in alternating colours.
• An emptied column accepts any card or run.
• You can move a run of several cards at once if there is enough empty space to do it one card at a time: the limit is the number of empty columns plus empty cells, plus one.

What the science means for your play

A loss is nearly always a mistake

Neither our solver nor the St Andrews one found more than a sliver of impossible deals with four cells. If a game beats you, the deal was almost certainly winnable. Undo back to the last point you had a choice and try the other line.

Play as if you had one cell fewer

The table shows why experienced players protect their cells. With three cells, the game is still almost always winnable. With two, one deal in five is gone. Keep at least one cell free for as long as you can, and treat the last empty cell as an emergency reserve.

Expect a first move that is not an Ace

In only 49% of our deals was an Ace playable on the very first move. Most games open by shifting a card or two to reach one.

Solutions are not short

The first solution our solver found averaged 62 moves across 9,913 solved deals. Those were not the shortest possible routes, but they show that a win typically takes several dozen careful steps.

Empty columns are worth more than cells

An empty column can hold a whole run of any length, while a cell holds one card, and it adds to the number of cards you can move in one go. Clearing a short column early is often the move that unlocks the game.

FreeCell on solitaire-free.uk

Free to play in any current browser on desktop, tablet or phone, with no downloads, no sign-up and no email address. Cards are large and clear, and you can zoom. There is Hint, unlimited Undo, Restart, Auto Play, drag and drop or tap to move, plus a running count of wins, time and moves. Sound can be muted, and you can go fullscreen or change the table colour.

Mistakes the solver never makes

• Sending cards home too early. A card on the foundation cannot catch the lower card of the opposite colour. Our solver only makes automatic foundation moves when they are provably safe; copy that caution.
• Parking cards in cells without a plan. Each filled cell lowers the number of cards you can move in one go. Know how each card will leave its cell before it goes in.
• Using an empty column as a dumping ground. It is worth more than any cell. Put a long run there, not a single stray card.

How we ran the research, step by step

The solitaire-free.uk team carried out this study in September 2026 to see whether this site's FreeCell matches the published science.

1. Taking the rules from our code. Eight columns of 7, 7, 7, 7, 6, 6, 6 and 6 cards, four cells, alternating-colour building, any card into an empty column, and the usual multi-card move limit.
2. Rebuilding the game. We wrote FreeCell in Python and built a solver for it. The solver's claims of impossibility were cross-checked on sample deals by a second, completely independent brute-force search: five deals with one cell and two deals with three cells. Both methods agreed every time.
3. Numbered deals. Deals 1 to 10,000 for the normal game and 3,000 deals each with 0, 1, 2 and 3 cells, all shuffled from numbered seeds so they can be dealt again. These are not the old Windows deal numbers.
4. The player. FreeCell has no hidden cards, so a solver sees exactly what you see. It searched up to 200,000 positions per deal.
5. A second pass. Deals left undecided with four cells were run again with up to 1,000,000 positions. After that, 32 of the 10,000 were still undecided and none had been proven impossible.
6. Checking against St Andrews. All five of our ranges contain the published figures, as the table shows.
7. The margin of error. A 95% range is the band the true answer very probably lies in. For four cells, our proven share of 99.68% has a range of 99.55% to 99.77%.
8. What the figures can't tell you. The site lets you move a card back down from a foundation; our solver never does. That can only make our figures slightly low. And a solver's "undecided" is not the same as "impossible".

Behind the numbers: the code

Our solver is written in Python 3.11 and needs nothing beyond the standard library. FreeCell shares a module, freecell.py, with Baker's Solitaire; a separate script, retry.py, ran the second pass, and verify_freecell_bruteforce.py is the independent cross-check. The solver is a best-first search: it always explores the most promising position next, judged by how many cards are not yet home and how many sit on a lower card. It keeps a table of positions it has already seen, stored in a standard form so that the same position is never searched twice, and it makes only foundation moves that are provably safe. A shared runner spread the work across 19 processes on a 20-core machine; the four-cell run took about eighteen minutes, and the reduced-cell runs between one and twenty-five minutes each. Ranges use the Wilson score interval.

Claude, Anthropic's AI assistant, helped us write and review the solver and put together a first draft of this page. The AI supplied no figures. Each one comes from running the solver, and we compared them with the published study before publishing.

Frequently asked questions

Is every FreeCell deal winnable?

Almost. The St Andrews study found 110 impossible deals in roughly ten million. We found none in 10,000, which is exactly what you would expect at that rate.

How much does each free cell matter?

Very little for the first, a great deal after that. Our solver and the published study agree: about 99% winnable with three cells, about 80% with two and about 19% with one.

Why is your four-cell figure 99.68% rather than 99.999%?

Because 0.32% of our deals were undecided when the solver hit its search limit. None was proven impossible, so our result is consistent with the published one, just less exact.

Is FreeCell more skill than luck?

Yes. With nothing hidden and nearly every deal winnable, the result depends on your planning.

What is similar but harder?

Baker's Solitaire uses the same layout with stricter rules, and on this site it is dramatically harder. Yukon and Alaska also deal everything at the start. See all twelve games on the games page.

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