FreeCell: the game at the opposite end from Spider

If four-suit Spider is the hardest game on this site, FreeCell is its mirror image. One deck, 52 cards, and every single one dealt face up from the start. There is no stock, no hidden card and no luck after the shuffle. Four free cells give you somewhere to park cards while you work, and the goal is to send all four suits home from Ace to King.

Because nothing is hidden, FreeCell is the game on our ladder where the cards are most reliably on your side. Our solver won 99.68% of 10,000 deals and did not find a single deal it could prove impossible. The remaining 0.32% simply needed more search time than we gave it. So when you lose a game of FreeCell here, the deal was almost certainly winnable.

This page shows how FreeCell ranks against Spider and the other games, how each free cell changes the odds, the rules we use, and a full walk-through of our method.

Where FreeCell sits on the ladder

Most games on this site were played by a program that only sees the face-up cards. FreeCell has no hidden cards, so the only thing worth measuring is how many deals can be won at all. On that measure it leads the site:

Note the difference in kind. FreeCell's figure needs no special knowledge, because all 52 cards are visible to you too. The four-suit Spider figure assumes the solver can see face-down cards, something no player can do.

Take away a free cell, and this happens

We also ran 3,000 deals with fewer free cells to see how much each one matters. For each setting, the range runs from the deals our solver solved to solved-plus-undecided:

Every published value falls inside our range. The table is also the best strategy lesson in this page: losing the fourth cell barely matters, losing the third costs about a fifth of all deals, and losing the second wipes out most of what is left. When you are down to one or two free cells in a real game, you are playing a much harder game than the one you started.

How FreeCell works here

• Layout: eight columns, four of seven cards and four of six, all face up.
• Free cells: four, each holding one card.
• Foundations: four, built up by suit from Ace to King.
• Columns: build down in alternating colors, a red 5 on a black 6.
• Empty columns: any card can go in.
• Moving runs: you can drag a sorted run in one go. The size you can move is the number of empty free cells plus empty columns, plus one; it is simply a shortcut for moving the cards one at a time.

Playing FreeCell with the numbers in mind

Keep one cell in reserve

From the table above: with three cells, most deals still work; with two, one in five is gone. Keeping one cell free as a reserve means that even if you misjudge, you are still in "three-cell" territory.

Don't expect an Ace on move one

Only 49.0% of our deals had an Ace ready to play on the very first move. Half the time you begin by digging, and that is normal.

Plan for a game of around 60 moves

The first solution our solver found for each deal averaged 62 moves. These are not the shortest solutions, but they give a sense of how much work a typical deal needs.

Treat a loss as your mistake, not the deal's

Zero impossible deals in 10,000 is a strong statement. If a game is stuck, undo and look for a different order.

Features on spidersolitaire.us

FreeCell is free on this site and runs in any modern browser on a computer, tablet or phone. Nothing to download, and no account, sign-up or email. Big clear cards with a zoom control, a Hint button, unlimited Undo, Restart, Auto Play to finish once you are clear, and drag-and-drop or tap. Wins, time and moves are tracked for the current game, sound can be switched off, and there is fullscreen and a choice of table colors.

Behind the numbers: how we ran the study

The spidersolitaire.us team ran this research in September 2026. For FreeCell, the question was how winnable deals are on our rules and how much each free cell contributes. Here is how we answered it.

1. Started from the code that runs our game. Eight columns of seven and six, all face up; four free cells; alternating-color builds; any card to an empty column; multi-card moves limited by free space. The site also lets a card come back down from a foundation; our solver never uses that move, which can only make our numbers slightly lower than the truth.
2. Rebuilt FreeCell as a program. Baker's Solitaire shares the same code, with its own rules switched on.
3. Dealt numbered hands. Deals 1 to 10,000 for the four-cell test and 1 to 3,000 for each reduced setting. They are our own seeds, not the classic Windows deal numbers.
4. Chose the solver. Because FreeCell hides nothing, a solver sees exactly what you see. We did not need a separate "human-like" player.
5. Ran it twice. Every deal got a search budget of 200,000 positions. That solved 99.13% outright. The deals still open got a second run with 1,000,000 positions, which brought the total to 99.68%, with 0.32% still undecided and none proven impossible.
6. Checked against published research. Blake and Gent (Journal of Artificial Intelligence Research, 2026) report 99.9989% for four cells and give figures for zero to three cells. Every published value falls within our ranges.
7. Measured the margin of error. For 99.68%, the 95% range is 99.55% to 99.77%.
8. Double-checked "impossible" verdicts. A separate, simpler brute-force program re-tested a sample of deals (1 to 5 with one free cell, plus deals 739 and 1053, which our solver had called impossible with three cells) and agreed on every one. We also note one small tension: at three cells our share of proven-impossible deals (0.97%) is a little above what the published figure implies, which we put down to sampling.

The code behind the ladder

The simulation is Python 3.11, standard library only, in the freecell.py module. The solver is a best-first search: it always expands the position that looks closest to a win, judged by how many cards are not yet home and how many are sitting on a lower card. It stores every position it has seen in a canonical form so it never searches the same one twice, and it makes automatic foundation moves only when they are provably safe. The brute-force cross-check lives in a separate script and moves one card at a time with no shortcuts. A shared runner spread the deals over 19 processes on a 20-core machine; the four-cell run took about 18 minutes, and the reduced-cell runs between one and 25 minutes each.

Percent ranges use Wilson score intervals. We used an AI assistant, Claude from Anthropic, to help write and review this code and draft the page. The AI did not generate the results; the numbers come straight from running the program, and the team compared them with the published figures before this page went live.

Ladder questions, answered with data

Is any FreeCell deal impossible?

A tiny number, according to published research: roughly one deal in 90,000. We found none in our 10,000.

Why is FreeCell easier than Spider?

You can see every card, and four free cells give you room to work. Spider hides most cards and only moves single-suit runs.

Could I win with only two free cells?

Often, but not always: with two cells, between 74.0% and 79.8% of our deals could be solved. With three it is 96.0–99.0%, and with one, fewer than one in five.

Is there a harder version?

Yes. Baker's Solitaire uses the same layout with stricter rules, and only 31.8% of its deals can be won here.

Which game pairs well with FreeCell?

For another open-layout game, try Yukon. To start the Spider ladder, go to one-suit Spider. Everything is on the games page, and deeper write-ups are on the blog.

By Admin, spidersolitaire.us. Research run September 2026.