ResearchSolver study
How many solitaire games can be won? A solver played 10,000 deals (2026)
Is every game of solitaire winnable? No. I ran the solver that checks our Daily on the same 5,000 random deals at Draw 1 and at Draw 3, then set the results beside published research and our players’ own win rates.
Key findings
- Our solver won 79.9% of Draw 1 deals and 74.7% of Draw 3 deals. It played the same 5,000 random deals at each draw, seeing every card and stopping after 100,000 positions.
- Many wins are found almost at once. 39.9% of its Draw 1 wins and 49.7% of its Draw 3 wins took fewer than 1,000 positions. Searching ten times longer, from 10,000 to 100,000 positions, added 17.2 points at Draw 1 and 11.8 at Draw 3.
- With every card visible, about 9 in 10 Draw 1 deals can be won. Exhaustive research puts the share at about 90.5% for Draw 1 and 81.9% for Draw 3. Our faster solver skips some moves, so it finds fewer.
- Hidden cards cut the odds sharply. Playing as you do, without seeing face-down cards, a published computer strategy wins more than 35% of Draw 3 games. On VSolitaire, players won 29.2% of Draw 1 and 34.1% of Draw 3 games (1,233 server-verified games, 21–30 September 2026).
- The Daily now uses this solver. Since 19 November 2026, each Daily (Draw 3) is the first of up to 20 random deals the solver wins within 10,000 positions. Between 13 October and 11 November, when it was any random deal, nobody won it on 7 of 30 days.
Draw 3: four answers to “how many can be won?”
Draw 3 is the classic Windows rule and our Daily’s. Each bar is a share of deals won.
Draw 1: easier, but not always
Turning one card at a time lets you reach every card in the stock, so more deals can be won.
How long the solver searched
A position is one arrangement of the cards the solver looks at. A won game took a median of 127 moves at Draw 1 and 113 at Draw 3, counting every card turned from the stock and every redeal.
| Search limit | Draw 1 | Draw 3 |
|---|---|---|
| 1,000 positions | 31.9% | 37.1% |
| 10,000 positions | 62.7% | 62.9% |
| 100,000 positions | 79.9% | 74.7% |
Method
- 5,000 random deals, each played at Draw 1 and at Draw 3 (10,000 games). The shuffle gives every order of the deck the same chance.
- Standard Klondike rules: unlimited passes through the stock, and only kings on empty columns. At Draw 3 the three cards are turned over together, so the third one is playable.
- The solver sees every card, face-down ones included. It is a weighted A* search, the one that checks our Daily, stopped after 100,000 positions per game with no time limit, so the run gives the same result on any computer.
- Every deal can be rebuilt from its seed: the CSV names the shuffle and the public study key.
Limits
- A deal the solver did not win is not proof it cannot be won. To stay fast, it never moves a card back from a foundation and tries few partial-column moves.
- It won 298 deals at Draw 3 but not at Draw 1. Any Draw 3 win can be replayed at Draw 1 by turning one card three times, so those are deals where its Draw 1 search ran out first.
- The solver sees every card; a player does not. Its rates are not what a person can expect.
- VSolitaire’s own Draw 3 currently differs from the standard rule in one detail: it keeps the three cards in stock order, so after each redeal a different card of every three comes up. Under that rule the solver won 83.1% of the same deals, which is why this study uses the standard rule.
- The players’ rates come from 1,233 solo games on our site over ten days; a game counts once a card is moved, and restarting or leaving counts as not won.
Published research
The figures we compare with, as each paper states them. VSolitaire is not connected with these studies.
- Blake C, Gent IP (2026). The winnability of Klondike Solitaire and many other patience games. Journal of Artificial Intelligence Research 85. Every card visible: 90.480% ± 0.116% of Draw 1 and 81.945% ± 0.084% of Draw 3 deals can be won.doi.org/10.1613/jair.1.17167
- Dang N, Gent IP, Nightingale P, Ulrich-Oltean F, Waller J (2025). Constraint models for Klondike. CP 2025, LIPIcs 340, 9:1–9:20. Refines the estimates to 90.454% ± 0.090% (Draw 1) and 81.942% ± 0.080% (Draw 3).doi.org/10.4230/LIPIcs.CP.2025.9
- Bjarnason R, Fern A, Tadepalli P (2009). Lower bounding Klondike Solitaire with Monte-Carlo planning. ICAPS 19. With cards hidden, a strategy wins more than 35% of Draw 3 games.doi.org/10.1609/icaps.v19i1.13363
- Yan X, Diaconis P, Rusmevichientong P, Van Roy B (2004). Solitaire: man versus machine. Advances in Neural Information Processing Systems 17. An expert who could see every card won 36.6% of 2,000 games under stricter rules.papers.nips.cc/paper_files/paper/2004/hash/48c3ec5c3a93a9e294a8a6392ccedeb4-Abstract.html
Credit
Data and analysis by VSolitaire. Free to reuse with a link to this page.
Data
Cite this page
VSolitaire. “How many solitaire games can be won? A solver played 10,000 deals (2026).” Solver run of 19 November 2026. https://vsolitaire.com/en/research/winnable-solitaire-2026
Play today’s solver-checked Daily
One deal for everyone, with a known win. No ads, no account.
More from VSolitaire
Frequently asked questions
Is every game of solitaire winnable?
No. Even a player who could see every card could win only about 90.5% of Draw 1 and 81.9% of Draw 3 Klondike deals (Blake & Gent 2026). With the cards hidden, a published computer strategy wins a little over a third at Draw 3.
Is the VSolitaire Daily always winnable?
Since 19 November 2026 the Daily is the first of up to 20 random deals our solver wins, so it has a known win. If the solver won none of the 20, the first would be used.
