How many solitaire games can be won?
Not every deal of solitaire has a way through. Before a game reaches you here, our solver has played it and won it. So far it has played 360,000 deals of eleven games. This is what it found.
| Game | Won |
|---|---|
| FreeCell 999,992 of 1,000,000 deals | |
| Eight Off 49,929 of 50,000 deals | |
| TriPeaks 96,989 of 100,000 deals | |
| Spider 9,285 of 10,000 deals | |
| Seahaven Towers 17,698 of 20,000 deals | |
| Klondike 4,261 of 5,000 deals | |
| Spider 2 Suits 8,451 of 10,000 deals | |
| Baker's Game 74,931 of 100,000 deals | |
| Klondike Turn 3 14,811 of 20,000 deals | |
| Pyramid 13,294 of 20,000 deals | |
| Spider 4 Suits 2,324 of 5,000 deals | |
| Golf 5,224 of 20,000 deals |
* FreeCell is the one game we did not solve ourselves: other solvers checked the first million Microsoft FreeCell deals years ago and found only eight that cannot be won.
Same cards, different rules
FreeCell, Baker's Game, Eight Off and Seahaven Towers all deal the 52 cards face up, and only their rules differ. The rules make all the difference.
- In FreeCell you build down in alternating colours and have four free cells. Only eight of the first million games cannot be won.
- Baker's Game has the same layout, but you build by suit: the 7 of hearts can only go on the 8 of hearts. That one change makes about one game in four impossible. Our solver won 74.9% of the first 100,000 deals, and a published study of random deals found 75.1%.
- Eight Off also builds by suit, but it gives you eight free cells instead of four, and almost every game comes out again: 99.9% (the study: 99.9%).
- Seahaven Towers builds by suit too, spreads the cards over ten columns and lets only a King into an empty column: 88.5% (the study: 89.3%).
The lesson is that free space matters more than anything else. Every free cell is room to untangle a column, and with eight of them even the strict rule of building by suit rarely gets stuck.
Golf, Pyramid and TriPeaks: the deal decides
In these three you clear cards from a layout onto a single pile, one rank up or down in Golf and TriPeaks, or in pairs that add up to 13 in Pyramid. How the cards fall decides a lot.
- Golf: only 26.1% of deals can be cleared, about one in four, with the usual rule that nothing can be played on a King. Three deals in four cannot be won however well you play, which is why a fair deal matters most in Golf.
- Pyramid: 66.5%, about two games in three.
- TriPeaks: 97.0%. Kings and Aces join up here, so an Ace can go on a King and a King on an Ace, and a chain rarely breaks for good.
Golf and TriPeaks have few enough positions for our solver to try every possible order of play. It settled every one of the 100,000 TriPeaks deals and the 20,000 Golf deals, either winning each one or proving it cannot be won, so these two numbers are exact. Pyramid has more ways to play, because the stock goes round three times, and our solver gives up on 51 deals; a second check with the whole stock open shows that at most 66.7% can be won, so its number is a floor, but a close one. The ones below are floors too.
Klondike and Spider: hard for a solver too
Klondike and Spider have face-down cards and so many possible positions that our solver cannot try them all in a reasonable time. Ours follows the most promising moves and gives up after a set number of positions. A win proves a game can be won; giving up proves nothing. The same solver plays the FreeCell family, which is why the Seahaven Towers number sits a little below the published one. For Baker's Game and Eight Off we went further and tried every position each lost deal can reach: there the lists are exact, and the deals not on them cannot be won.
- Klondike: our solver won 85.2% of games turning one card and 74.1% turning three. The published study, with a solver that keeps going much longer, found 90.5% and 81.9%. So some winnable Klondike games are not dealt here yet; we only deal the ones our solver has won.
- Spider: the share falls from 92.8% with one suit to 84.5% with two and 46.5% with four. That drop is our solver, not the game: research suggests almost every Spider game can be won, even with four suits. With more suits there are fewer safe moves, and the solver gives up more often.
Can be won is not the same as easy
In Klondike, Spider, Golf, Pyramid and TriPeaks some cards are hidden from you: face-down cards, or the order of the stock. The solver sees them all. So "can be won" means a way through exists from the first card, and you can still miss it with an unlucky guess. That is why players win far fewer games than the numbers above. When it happens here, Undo and Restart are there, and so is the way through.
In FreeCell, Baker's Game, Eight Off and Seahaven Towers every card is face up from the start. What the solver sees, you see.
How a win is checked
When the solver wins a game, it writes down every move. The game you play then replays that list one move at a time, with the same rules that check your own moves, and only when every card reaches the foundations does the game's number join the list we deal from. New game picks a random number from that list, and if you type a number that is not on it, you get a winnable game instead. Read more about why we made Fair Solitaire.
Read more
Which solitaire game should you play? compares the twelve games on these numbers. What the first cards tell you gathers what the opening layout predicts in each game, Winnable is not the same as easy explains what the promise covers, and The impossible deals looks at the games nobody can win.
Sources
The published figures come from Charlie Blake and Ian Gent, The Winnability of Klondike Solitaire and Many Other Patience Games. All the other numbers on this page are our own.