FreeCell Two Decks: what changes when you shuffle in a second deck?
Put two decks on the table instead of one and a FreeCell win stops being a 60-move sprint. It becomes a 150-move project. That is the first thing our test showed. The first winning line our solver found averaged 148.0 moves in FreeCell Two Decks (95% range 145.9–150.1, over 352 solved deals), against 62.2 moves in one-deck FreeCell (61.9–62.4, over 9,913 solved deals). The game you are about to play on this page is the long one.
The second finding is what did not change: fairness. Of 400 two-deck deals, the solver finished 88.0% (95% range 84.4%–90.8%) and proved not one impossible (0.0%, range 0.0%–1.0%). The other 12.0% (9.2%–15.6%) were still open when it hit its limit of 150,000 positions per deal. One-deck FreeCell, given a bigger search, came out at 99.7% solved of 10,000 deals (99.5%–99.8%), also with none proven impossible. A second deck makes the road longer. On our evidence, it does not put a wall across it.
freecell-two-decks.com is built around that question: what does doubling the deck actually do? Below, every game on the site is sorted into one-deck and two-deck, with the figures we measured for each and a plain note where we have none.
One deck against two, side by side
| What we measured | One deck (52 cards) | Two decks (104 cards) |
|---|---|---|
| FreeCell: length of the solver's first winning line | 62.2 moves (61.9–62.4), 9,913 solved deals | 148.0 moves (145.9–150.1), 352 solved deals |
| FreeCell: deals solved by the solver | 99.7% (99.5–99.8%), n = 10,000, up to 1,000,000 positions | 88.0% (84.4–90.8%), n = 400, up to 150,000 positions |
| FreeCell: deals proven impossible | 0.0% (0.0–0.04%), n = 10,000 | 0.0% (0.0–1.0%), n = 400 |
| FreeCell: an Ace playable on move one | 49.0% (48.0–49.9%), n = 10,000 | 55.0% (50.1–59.8%), n = 400 |
| Klondike vs Solitaire Double: simple bot's win rate (Turn 1) | 42.3% (41.3–43.3%), n = 10,000 | 81.5% (80.7–82.2%), n = 10,000 |
| Klondike vs Solitaire Double: no move among the face-up cards at the start | 8.6% (8.1–9.2%), n = 10,000 | 2.7% (2.4–3.0%), n = 10,000 |
Two cautions. The two FreeCell searches had different budgets and very different sample sizes, so the solved shares are not a like-for-like contest; the fair reading is "both near the top, none proven lost". And the Klondike figures come from a simple bot that never plans or undoes, so they are floors, not human win rates. Even so, the pattern is clear. A second deck gives every card a twin, and twins mean more landing spots.
Which games here use two decks
Two decks, 104 cards
- FreeCell Two Decks (this page, and its full guide at /freecell-twodecks): 88.0% solved, 12.0% undecided, 0 proven impossible, n = 400.
- Solitaire Double: simple bot won 81.5%, n = 10,000. No solver was run, so we have no winnable share.
- Spider at two and four suits: look-ahead bot won 52.8% (1 suit, n = 4,000), 5.7% (2 suits, n = 4,000) and 0.2% (4 suits, n = 2,000). There is no one-deck Spider on this site to compare with.
- Forty Thieves: look-ahead bot won 5.8% (5.0–6.7%), n = 3,000. No one-deck version was measured.
- Diplomat: look-ahead bot won 47.9% (45.7–50.0%), n = 2,000. No one-deck version was measured.
One deck, 52 cards
FreeCell, Klondike and Klondike Turn 3, Eight Off, Baker's Solitaire, Yukon, Alaska, East Haven, Flower Garden, Golf, Pyramid, Scorpion and Simple Simon. Each page says whether a two-deck cousin exists here and what, if anything, we can compare.
Playing a 104-card game well
Budget for length, not for luck
With every card face up there is no luck left after the deal. What a second deck adds is distance. Treat the game as three stages: open a column, get the low cards moving, then build long runs.
Use the twins
Each card has an identical partner somewhere on the table. When a red 6 could go on either black 7, choose the 7 whose column you want to keep, and leave the other free.
Eight cells are not "plenty"
Twice the cells also means twice the cards that want them. In one-deck FreeCell, dropping from four usable cells to two cut the solved share from 99.7% to 74.0% (72.4–75.5%, n = 3,000). Guard your cells the same way here.
Rewind rather than redeal
None of our 400 two-deck deals was proven impossible. When you jam, Undo back to the last position you liked and try another line.
What you get when you play here
- Free play in any current browser on computer, tablet or phone. Nothing to install, no account, no sign-up.
- Zoom and fullscreen, which help a lot with ten columns and 104 cards.
- Hint, unlimited Undo, Restart and Auto Play to finish a won game.
- Drag-and-drop or tap-to-move.
- A count of wins, time and moves, sound on or off, and a choice of table colors.
How we got these numbers
The freecell-two-decks.com team ran these simulations in September 2026 on the rules read from the code of the games on this site.
- Rebuilt each game in Python (3.11, standard library only), move rules copied from our own game code.
- Dealt numbered shuffles so every deal can be replayed. FreeCell Two Decks used deals 101 to 500, because deals 1 to 8 were used to tune one scoring setting.
- Used a solver where every card is visible (both FreeCells) and a simple bot elsewhere (Klondike, Solitaire Double, Spider, Forty Thieves, Diplomat).
- Sorted solver deals into solved, proven impossible and undecided.
- Put a 95% range on every figure (Wilson score intervals for percentages, normal intervals for averages) and checked against Blake and Gent (Journal of Artificial Intelligence Research, 2026) where they cover a game. They do not cover FreeCell Two Decks, Solitaire Double, Forty Thieves or Diplomat.
An AI assistant, Claude by Anthropic, helped write and review the code and draft this text. It produced no figures: every number comes from running the code or from the linked study, and the team checked them before publishing.
Questions about two decks
Is FreeCell Two Decks harder than FreeCell?
Longer, certainly: 148.0 moves against 62.2 for the solver's first wins. We found no sign that it is less fair, since no deal in 400 was proven impossible.
Does a second deck always make solitaire easier?
Not always, and we cannot say for every game. In Klondike it helped a simple bot a lot (42.3% to 81.5%). For Spider, Forty Thieves and Diplomat we have no one-deck version to compare.
Are these my odds?
No. Solver figures show what is possible with perfect play; bot figures show what a player with no plan reaches. You will land somewhere between.
Where can I read more?
Every game has its own page, the longer write-ups are on the blog, and the full list is on the games page.
By Admin, freecell-two-decks.com (Forty Media). Research run September 2026.