Home › Is every solitaire deal winnable? We checked 60,000 of them

Is every solitaire deal winnable? We checked 60,000 of them

By Gleb Tkatchouk · last updated 2026-09-18

The honest answer to "is this solitaire winnable?" is usually "nobody checked". Most apps shuffle a deck and hand it over; if the deal was hopeless from the first card, you find out after twenty minutes. This page gives numbers instead — measured, level by level, on the deals that ship inside Decks — and explains what the numbers do and do not mean.

Play it on iPhone — no ads, no account, every deal proven winnable.

Scan with your phone to install

How we measured

Every game in Decks is a document the engine reads, so every game can be played by a machine as well as a person. Before a deal reaches the app it is played to the end by a bot, or searched by a solver that must find a winning line; the line is then replayed move by move by the engine to confirm it is legal. A deal counts as proven winnable only if that check passed.

Two things follow. First, the percentages below are floors, not the true winnable share: a deal the bot lost may still be winnable by a better player. Second, the pool the app deals from contains only proven deals. When you lose in Decks, the deal you lost was winnable — and the app tells you so on the loss screen, with the deal code, so you can try it again.

Klondike, by level

Klondike in Decks has five levels that differ in the number of columns; all deal one card at a time with five passes through the deck.

Level	Columns	Deals measured	Proven winnable by the bot
Very easy	5	8,000	6.8% under a stricter rule — winnable in a single pass of the deck
Easier	6	8,000	70.5%
Classic	7	10,000	40.1%
Hard	8	10,000	4.2%
Brutal	9, two decks	9,000	1.6% — measured on 5 September on an earlier configuration of the level; a re-measurement on the shipped configuration is pending

The "Very easy" figure needs a word: that level is held to a harder standard on purpose — a deal enters its pool only if it can be laid out in one pass of the deck, without ever recycling the waste. That is what makes the level feel easy in the hand, and it is why its proven share looks low.

The Classic number is the one to compare with folklore. Estimates in the literature for standard draw-one Klondike run around 80–90% theoretically winnable; our greedy bot, which never plans ahead, wins 40%. The gap between those numbers is what a human does with foresight and undo.

Spider

Suits	Deals measured	Proven winnable by the solver
One suit	60	98.3%
Two suits	100	66.0%
Four suits	50	0% — the solver cannot prove four-suit deals within a practical budget

For one and two suits, Decks deals only the proven deals. Four-suit Spider is dealt unchecked, and a loss is marked "solvability unknown" rather than "unwinnable": a search that ran out of budget proves nothing. See the Spider page for what changes between suit counts.

FreeCell, Pyramid, TriPeaks, Spanish Solitaire

Game	Level	Deals measured	Proven winnable
FreeCell	—	252	19.8% with a look-ahead bot
Pyramid	Easier (four passes)	180	26.1%
Pyramid	Classic (two passes)	225	20.4%
Pyramid	Hard (one pass)	690	8.3%
TriPeaks	—	2,000	51.1%
Spanish Solitaire	Easy / Normal / Hard	2,000 each	67.7% / 39.1% / 13.9%

FreeCell is famous for being almost always winnable — of the original 32,000 Microsoft deals, all but one are — and our 19.8% says more about the bot than the game: FreeCell rewards exactly the planning a simple bot lacks. What matters for the app is that the pool it deals from is proven.

Why deal only proven deals?

Because "you lost" and "it could not be won" feel the same from the inside, and only one of them is fair. A solitaire app that hands out impossible deals is not harder; it is dishonest about where the difficulty came from. Dealing from a proven pool means a loss in Decks is always something you could have done differently — and the deal code lets you find out what.

It also lets the app say something most cannot: on the loss screen, "this deal is winnable". Never "unwinnable" — a failed search is not a proof.

Frequently asked questions

What percentage of solitaire games are winnable? It depends entirely on the variant and the rules. For classic draw-one Klondike, published estimates land around 80–90%; for FreeCell, nearly all; for four-suit Spider, nobody has a reliable number. Our measured floors are in the tables above.

Is every game of Klondike winnable? No. Some deals are impossible from the start, which is exactly why Decks checks deals before dealing them.

Are unwinnable FreeCell games real? Yes, but rare: deal 11,982 of the original Microsoft set is the famous one. Decks deals FreeCell only from its proven pool.

How is "proven winnable" different from "winnable"? Proven means a machine actually won it and the win was replayed and checked. Winnable in the abstract may be true of more deals than anyone has proven.

Does this make the game easier? No — the deals are the same shuffles; the losing ones are just filtered out. What changes is that a loss is always your own.