What percentage of solitaire games are winnable?

Roughly 79% of classic Klondike (draw-1) deals are winnable in principle — that's the figure published solver studies converge on, with estimates for "thoughtful" play running as high as about 82%. Draw-3 is harder, and real players win far, far fewer. Here are the honest numbers and what they mean for your game.

Where do these numbers come from?

For most of solitaire's two-century history, nobody knew the answer. The mathematician Persi Diaconis and his co-authors famously called the unknown probability of winning solitaire "one of the embarrassments of applied probability" in their 2005 paper Solitaire: Man Versus Machine (Yan, Diaconis, Rusmevichientong & Van Roy). That same line of research started producing real answers: their rollout technique won about 70% of games and established that the true winnable rate was well above what anyone guessed.

The strongest results come from studying "Thoughtful Solitaire" — Klondike played with every card's location known, so a computer can search the whole game. Research on thoughtful play (Bjarnason, Fern & Tadepalli's work on Klondike, and later the large-scale Solvitaire solver by Blake & Gent at the University of St Andrews, 2019) puts the winnable fraction of draw-1 deals at about 79–82%, with Solvitaire's estimate near 82% after solving millions of deals.

Draw-3 is meaningfully harder. Because only every third stock card is playable on each pass, published estimates for draw-3 winnability are lower and vary more with the exact rules (especially how many passes through the stock you allow) — broadly in the mid-60s to low-80s percent range rather than a single tidy number. We'd rather give you an honest range than fake precision.

Winnable is not the same as won

Here's the gap that should change how you play: those ~79–82% figures assume perfect play with full information — the solver sees every face-down card. Actual human win rates, as reported by the big solitaire sites and app publishers from their own player data, typically land somewhere between about 10% and 30%, depending on skill, mode, and how freely people use undo.

Read those two numbers together: the deal you just lost was probably winnable. Roughly four deals in five are. The difference between 80% and 15% isn't luck — it's decisions made blind. Ordinary Klondike keeps 21 cards hidden and forces you to commit before you know what a move uncovers, so even good players burn winnable games on reasonable-looking guesses.

Why can't every game be won?

Some deals are simply dead on arrival — the cards you need are stacked in orders no sequence of legal moves can untangle. A common culprit: both black 6s buried under both red 5s across the wrong columns, or an Ace trapped beneath cards that can only move onto cards trapped beneath it. No skill fixes those deals; that missing ~20% is why even a perfect player re-deals sometimes. (Solvers exist precisely because spotting a dead deal by eye is nearly impossible.)

How do you close the gap?

Two ways. The first is classic strategy — flip the big stacks first, keep Kings ready for empty columns, don't rush cards to the foundations. Our guide to winning more often covers the habits that reliably move a win rate.

The second is information. Remember what the solver studies actually measured: Klondike becomes ~80% winnable when you can see the cards. That's the entire idea behind the Peek button — our patent-pending exclusive that lets you press and hold to see face-down cards before you commit a move. It's the same informed game the "thoughtful" studies describe, in your hands instead of a computer's. You still make every decision; you just stop making them blind. (It's better brain training, too — planning beats guessing.)

Does using undo change the odds?

Meaningfully, yes. Undo lets you take back a move after seeing what it revealed, which quietly converts hidden-information Klondike into something closer to the open game the solvers study — one reason player-reported win rates vary so wildly. There's no shame in it (ours is unlimited and never triggers ads), but it's worth naming what the undo loop really is: discovering information by trial and error, one regret at a time. Peeking first is the same information, minus the regret :)

Test the theory yourself

Play a free deal right now, and when you hit a genuinely ambiguous choice, hold Peek before you commit. You'll feel the win-rate gap close in real time :)