There is no single answer to "What are the odds of winning solitaire?" Solitaire is a family of games, and small rule changes can move the theoretical winnability rate by many percentage points. The best-supported figures describe whether a winning path exists, not how often a person will find it.
For rules close to those used on Cards4.net, current published evidence supports these reference points:
| Game and rules | Estimated deal winnability | What the number means |
|---|---|---|
| Klondike Turn 1, thoughtful, unlimited redeals, no foundation-to-tableau moves | about 90.2% | Solver estimate for random deals when hidden information can be learned through backtracking |
| Klondike Turn 3, same assumptions | about 81.5% | Solver estimate; the three-card draw order lowers access to the stock |
| Microsoft FreeCell deals 1-32,000 | 31,999 of 32,000 | Every deal except #11982 has a solution |
| Microsoft FreeCell deals 1-1,000,000 | 999,992 of 1,000,000 | Eight specific deals are known to be unsolvable |
| Classic 4-suit Spider, thoughtful | near 99.99% in a 32,000-deal study | 31,998 solved, two unresolved; not a human win rate |
The Klondike and Spider figures come from Blake and Gent's extended 2024 paper, The Winnability of Klondike Solitaire and Many Other Patience Games. The paper defines its exact rules, reports confidence intervals, and separates solved, unsolved, and unresolved deals. That methodology is why its numbers are more useful than uncited win-rate claims copied between game sites.
Winnability Is Not the Same as Win Rate
Three different measurements are often collapsed into one:
- Deal winnability: whether at least one legal sequence reaches a win.
- Solver success rate: how often a particular program finds or proves an answer within its resource limits.
- Player win rate: the share of started games a particular person or population completes.
A solver can establish that a deal has a winning path without suggesting a human will find it. Conversely, a solver timeout does not prove a deal is impossible. Good research reports unresolved cases separately instead of counting them as losses.
Human win rates are especially difficult to compare. A platform may allow unlimited undo, filter out unwinnable deals, count abandoned games differently, or offer hints and autocomplete. Skill levels also vary. Without a published dataset and ruleset, a statement such as "experienced players win 60%" is not a reliable benchmark.
Klondike: Why the Rules Change the Answer
Klondike starts with hidden tableau cards, so the strongest published estimates usually analyze thoughtful Klondike. In this model, the player knows the hidden cards or can learn them and return to an earlier position. Blake and Gent note that unlimited undo in a computer implementation approximates this condition.
Under a standard alternating-color, Kings-only-space ruleset with unlimited redeals and no moves back from the foundations, their estimates are approximately:
- Turn 1 / Draw 1: 90.204%
- Turn 3 / Draw 3: 81.524%
Allowing cards to move back from foundations raises those figures slightly, to about 90.480% and 81.945%. The difference matters because some digital versions allow that move and others, including Cards4.net, do not.
These results correct two persistent online claims: 79% is not the best-supported Turn 1 figure for this ruleset, and Turn 3 is not limited to roughly one game in five when unlimited redeals and thoughtful play are allowed. Restricting redeals or preventing the player from learning hidden cards would lower the achievable rate, but the exact real-play probability remains a different and harder question.
FreeCell: Almost Every Deal, Not Every Deal
FreeCell exposes every card at the start and provides four temporary cells, so luck after the deal is minimal. Its theoretical winnability is exceptionally high.
The original Microsoft set contains 32,000 numbered deals. The Internet FreeCell Project found solutions for every one except #11982, which later solvers confirmed is impossible. In the expanded range from 1 through 1,000,000, the eight known impossible deals are:
#11982, #146692, #186216, #455889, #495505, #512118, #517776, and #781948.
That is eight unsolvable deals in one million, or 0.0008% of that numbered set. It does not prove that exactly the same proportion applies to every possible 52-card FreeCell arrangement, because the Microsoft generator samples a specific sequence of deals. It does support the practical conclusion that a randomly encountered FreeCell loss is far more likely to be a missed line than an impossible layout.
See FreeCell Microsoft Deal Numbers Explained for the generator, the numbered set, and how to load a specific deal on Cards4.net. The list of eight is also documented by the FreeCell Help Center.
Spider Solitaire: Difficult Does Not Mean Unsolvable
Spider is a useful warning against inferring winnability from human difficulty: four-suit Spider is hard to play well because mixed-suit sequences cannot move as units, yet a classic 32,000-deal analysis solved 31,998 and left only two unresolved, corresponding to a literature estimate near 99.99%. The Spider Solitaire winnability guide covers that study in full, including why the two remaining deals are unresolved rather than proved impossible and why the classic figure cannot be split into separate percentages for 1-suit, 2-suit, and 4-suit play.
What About Yukon, Pyramid, and TriPeaks?
Published figures for these names are often not comparable because the rules vary between implementations. Pyramid may allow one pass or several through the stock; TriPeaks may allow King-to-Ace wrapping; Yukon implementations differ on foundation moves and sequence handling. Those changes alter the game tree.
Without a source that states both the rules and the sampling method, a precise percentage creates false confidence. The useful player-facing facts are simpler:
- Yukon exposes much of the tableau but can still lock key low cards.
- Pyramid depends heavily on whether removable pairs become accessible in the right order.
- TriPeaks combines short-term choice with a stock order the player cannot control.
Cards4.net can offer a solvable-mode deal for a supported variant without claiming that a universal percentage exists for every version of that game.
How Cards4.net Changes Your Odds
Cards4.net has two different deal paths:
- Random mode samples a normal seeded shuffle. No winnability promise is made.
- Solvable mode and daily challenges select from pre-validated seed libraries. The offline solver found a winning line before a seed entered the library.
The second path changes the deal distribution. It removes deals the solver did not validate, so statistics for an unrestricted random shuffle no longer describe the set you are playing. A guaranteed-solvable deal can still be difficult; the guarantee says a path exists, not where it is.
FAQ
What percentage of Klondike games are winnable?
For thoughtful Klondike with unlimited redeals, alternating-color builds, Kings-only empty columns, and no moves back from foundations, solver research estimates about 90.2% for Turn 1 and 81.5% for Turn 3. Different rules produce different answers.
Is every FreeCell game winnable?
No. One of the original 32,000 Microsoft deals is impossible, and eight are known to be impossible in the first one million numbered deals. Almost every deal in those sets is solvable.
Why is Spider hard if almost every studied deal is solvable?
Winnability asks whether a path exists. Difficulty asks whether a player can find that path. Spider's large search space and suit restrictions can make a solvable deal extremely hard to navigate.
Does undo change the theoretical odds?
Undo does not change the legal starting position, but it changes what a person can learn and retry. For games with hidden cards, unlimited backtracking approximates the "thoughtful" model used in solver research.
What should I use as a personal benchmark?
Track your own rate by game, variant, and settings. Record whether solvable mode, hints, and undo were enabled. That produces a meaningful comparison; a generic percentage from a different ruleset does not.
Sources
- Charlie Blake and Ian P. Gent, The Winnability of Klondike Solitaire and Many Other Patience Games, extended August 2024 version.
- Michael Keller and the Internet FreeCell Project history summarized in Blake and Gent's paper, especially the original 32,000 Microsoft deals and #11982.
- FreeCell Help Center: known impossible numbered deals.
Research and Cards4.net rule mappings were rechecked on July 16, 2026.