FreeCell: where thinking ahead decides almost everything
FreeCell is the purest test of planning on spidersolitaireonline.us. All 52 cards are dealt face up into eight columns, there is no stock and no hidden card, and four free cells give you room to maneuver. Once the cards are dealt, luck has finished its work. What happens next depends on how far ahead you look.
Our study asked a simple question: how much does extra thinking buy you in FreeCell? We let a search program play 10,000 deals on this site's rules. With a modest thinking budget it solved 99.1% of them (95% range 98.9%–99.3%). When we let it think five times longer on the deals it had not cracked, the total rose to 99.7% (9,968 of 10,000; 95% range 99.5%–99.8%), and not one deal was proven impossible. The lesson is plain: in FreeCell, a stuck game is nearly always a game that needed more planning, not a bad deal.
We did not build a "greedy" player for FreeCell, so this page cannot show a greedy-versus-planner gap in the way our Spider pages do. Instead it shows the two things that planning controls most here: how much search a deal needs, and how much every free cell you keep open is worth.
What more thinking buys in FreeCell
| Setting | Deals solved | 95% range | Deals |
|---|---|---|---|
| Search budget of 200,000 positions | 99.1% | 98.9%–99.3% | 10,000 |
| Hard deals re-run with 1,000,000 positions | 99.7% | 99.5%–99.8% | 10,000 |
| Proven impossible (after both runs) | 0.0% | 0.0%–0.04% | 10,000 |
| Still undecided | 0.3% | 0.2%–0.5% | 10,000 |
The deals that needed the second, longer search are the ones that feel impossible to a person. They are not. They simply reward a longer plan.
Every free cell is thinking space
A free cell is where you park a card while a plan unfolds. We ran 3,000 deals each with fewer cells to measure what that space is worth. Figures show deals solved; the published values are from Blake and Gent (2026).
| Free cells | Solved (our solver) | 95% range | Proven impossible | Published |
|---|---|---|---|---|
| 4 (as played here) | 99.7% | 99.5%–99.8% | 0.0% | 99.9989% |
| 3 | 96.0% | 95.3%–96.7% | 1.0% | 99.36% |
| 2 | 74.0% | 72.4%–75.5% | 20.2% | 79.54% |
| 1 | 17.4% | 16.1%–18.8% | 79.6% | 19.35% |
| 0 | 0.2% | 0.1%–0.4% | 99.7% | 0.21% |
Read this table as a strategy chart. Every time you fill a cell with no plan to empty it, you are playing one row further down. Filling one cell costs little; filling two turns a near-certain win into a game where one deal in five cannot be won from that point; filling three leaves fewer than one deal in five alive.
The rules of FreeCell on this site
- Deal: 52 cards, all face up, in eight columns: four of seven cards and four of six.
- Free cells: four, one card each.
- Foundations: four, one per suit, built up from Ace to King.
- Tableau: build down in alternating colors, such as a black 8 on a red 9.
- Empty columns: accept any card.
- Group moves: an ordered run can be moved at once if it fits: empty free cells plus empty columns, plus one. This is a shortcut for moving the cards one by one.
Playing like the planner, not the grabber
Look before the first move
Only 49.0% of our deals (95% range 48.0%–49.9%, 10,000 deals) had an Ace free to play at once. In the other half, the first job is choosing which column to dig into. Pick the one where the Ace or 2 sits under the fewest cards, and count how many cells the dig will use before you start.
Count cells like money
The table above says a spare cell is worth more than most tableau moves. A grabbing player fills cells to make a move look possible. A planning player asks: which card will take this parked card back out, and how soon?
An empty column is worth more than a cell
Because group moves count empty columns, one empty column roughly doubles the length of run you can shift. Aim your early plan at emptying a short column, then keep it empty until a long run needs it.
Expect a plan of about 60 moves
The first solution our solver found averaged 62.2 moves (95% range 61.9–62.4, over 9,913 solved deals). These are not the shortest lines, but they show a real FreeCell win is a sequence, not a handful of lucky moves.
When stuck, rewind further
Our solver needed a much deeper search for the last few hundred deals. You have unlimited Undo here; use it to go back ten or twenty moves, not two.
Playing FreeCell on spidersolitaireonline.us
FreeCell is free and plays in any modern browser on a computer, tablet or phone, with no download and no account. You get large cards with zoom, Hint, unlimited Undo, Restart, Auto Play to finish, and drag-and-drop or tap to move. Wins, time and moves are counted for the current game, and sound, fullscreen and table color are yours to set.
How we ran the FreeCell study
The spidersolitaireonline.us team ran these simulations in September 2026, on this site's own FreeCell rules. The programs are written in Python, and an AI assistant, Claude by Anthropic, helped write and review the code and draft this page. Every figure above came from running that code.
- Rules from the game itself. We copied the rules from the code that runs FreeCell here: the 7-and-6 layout, four cells, alternating colors, any card in a gap, group moves sized by free space. The site also allows a card back down from a foundation; the solver never uses that, so if anything our figures are slightly low.
- Numbered deals. Deals 1 to 10,000 for the normal game and 1 to 3,000 for each reduced-cell version, from fixed seeds so anyone with the code can replay them. They are not the old Windows deal numbers.
- A solver, not a bot. Nothing is hidden in FreeCell, so a search program sees what you see. It always explores the position that looks closest to a win and never visits the same position twice.
- Two thinking budgets. First up to 200,000 positions per deal, then up to 1,000,000 for deals left open. That is the "more thinking" comparison in the first table.
- Published check. Blake and Gent (Journal of Artificial Intelligence Research, 2026) give values for zero to four cells; all sit close to ours. At three cells our proven-impossible share is a little higher than theirs implies, which we put down to sampling.
- Margins of error. Every range is a 95% Wilson score interval.
FreeCell questions
Can every FreeCell deal be won?
Almost. We found no impossible deal in 10,000, and published research puts impossible deals at roughly one in 90,000.
How much does keeping a cell free really help?
A lot. With three cells our solver solved 96.0% of deals; with two, 74.0%; with one, 17.4%.
Why do some deals feel unwinnable?
Because they need a long plan. Our solver needed five times more search for its hardest deals, and still found wins.
What should I play next?
Eight Off doubles the cells but asks for same-suit builds; Baker's Solitaire is the strict version where planning matters even more. For hidden cards and real greedy-versus-planner gaps, try Spider. See all games on the home page.
By Admin, spidersolitaireonline.us (Solitairen OOD). Research run September 2026.