No—Solitaire is not always winnable. But a useful percentage depends on which game and rules you mean. The strongest current Klondike result is for a computer solving a “thoughtful” version in which every card is known. It is not an ordinary player’s win rate, and it is not a documented win rate for Google Solitaire.
A 2026 peer-reviewed study by Charlie Blake and Ian P. Gent reports that thoughtful Klondike is winnable in 81.945% ± 0.084% of deals. The authors used the Solvitaire search program and define the tested condition as one where the rank and suit of every card are known. Hidden-card play, Draw 1 versus Draw 3, redeal limits, and software-specific “winnable deal” settings can all change the question. See the February 2026 Journal of Artificial Intelligence Research paper.
What does “winnable” mean?
A deal is winnable if at least one legal sequence of moves reaches the game’s winning condition. That is different from the percentage of games a person wins. A player can lose a winnable deal by choosing the wrong move, and can make the best available choices yet still lose a deal that has no solution under the selected rules.
| Term | What it measures | What it does not tell you |
|---|---|---|
| Deal winnability | Whether any winning line exists under one defined ruleset | Whether a particular player will find that line |
| Solver result | What a specified program found under documented assumptions | A universal human win rate |
| Personal win percentage | Wins divided by completed games in a defined record | The mathematical solvability of all possible deals |
| Guaranteed-winnable mode | A provider’s selected or generated deal pool, if documented | The winnability of random physical-card deals |
What the 81.945% Klondike result actually shows
Blake and Gent used a general-purpose program to study 73 variants of 35 patience games. For Klondike, the headline estimate applies to thoughtful play: the solver knows all cards, including cards that would normally be face down. The reported 95% confidence interval is ±0.084 percentage points around 81.945%.
The result supports a bounded conclusion: even with complete information and computer search, a meaningful share of the tested Klondike deals had no winning line. It does not justify writing “you should win about 82% of your games.” A normal player lacks complete information, may use different stock rules, and is not performing exhaustive search.
Five rule choices that change the answer
1. Klondike is not the whole Solitaire family
“Solitaire” names a large family of single-player card games. Klondike, Spider, FreeCell, Pyramid, and Accordion have different layouts, legal moves, information, and winning conditions. A percentage for one cannot be transferred to another.
2. Draw 1 and Draw 3 are different games
In the Bicycle Klondike rules, stock cards turn up in groups of three. Many digital versions also offer Draw 1. How cards become available from the stock changes the set and order of possible moves.
3. Redeal limits matter
A version may permit unlimited stock passes, a fixed number, or none. If one ruleset gives you another pass and another ends the game, they do not have the same solution space.
4. Hidden information matters
In ordinary Klondike, many cards begin face down. You must act without knowing every future reveal. Thoughtful Klondike removes that uncertainty for analysis, which makes it useful for studying whether a solution exists but different from normal play.
5. Digital assistance changes play, not necessarily the deal
Undo, hints, auto-finish, and replay can help a player revisit decisions. A “winnable deals only” option goes further by changing which deals enter the game. These controls should be recorded separately when comparing results.
An assumption matrix for any win-rate claim
Before accepting or repeating a Solitaire percentage, fill in this matrix. A missing answer means the number is not ready to compare with another result.
- Game: Klondike, Spider, FreeCell, Pyramid, or another named variant?
- Stock rule: Draw 1 or Draw 3, and how many passes?
- Information: normal hidden cards or thoughtful/all-known cards?
- Actor: human player, a specific solver, or a provider’s preselected deal generator?
- Assistance: undo, hints, replay, auto-finish, or guaranteed-winnable mode?
- Outcome: mathematical solution found, player win, or game completed?
- Sample: how many deals, selected how, and during what period?
Worked example: why two percentages may not conflict
Suppose a solver checks 100,000 randomly generated thoughtful Draw 3 Klondike deals and reports the share with a solution. Separately, a player records 200 Draw 1 games in an app that offers unlimited undo. The two percentages answer different questions. One estimates solvable deals under complete information; the other describes one person’s results with different stock and assistance rules.
This is a hypothetical comparison, not a new experiment. Its purpose is to show which assumptions must match before two numbers can be compared.
Can strategy still improve your results?
Yes, within the opportunities a deal provides. Strategy cannot create a solution for an unwinnable deal, but it can reduce avoidable losses on winnable ones. In Klondike, useful decision checks include exposing face-down tableau cards, avoiding a premature foundation move that blocks tableau access, and comparing alternative moves before committing. These are decision heuristics, not guarantees.
Start with the site’s defined Klondike rules and worked example, then use the advanced strategy guide to examine option-preserving moves. If you want to compare how information and objectives change across games, see the Solitaire variant chooser.
The practical answer
Some Solitaire deals are unwinnable, and a loss is not proof that you missed a winning move. For Klondike, use 81.945% ± 0.084% only with its full label: the 2026 Solvitaire estimate for the thoughtful, all-cards-known variant. For your own records, name the game, rules, assistance settings, and number of completed games. That produces a useful personal measure without turning a research result into a promise.
Related reading
Source and scope note: This article is an independent explanation of the cited rules and research, updated August 13, 2026. No first-hand solver experiment or Google Solitaire win-rate test was performed for this article. Jack is the accountable editor. Corrections can be sent to blog@solitairegoogle.com; see the editorial policy.
Related practical guides: how Draw 1 and Draw 3 change stock access.