Can Every Game of Solitaire Be Won? The Real Odds

How many Klondike, FreeCell, Pyramid, TriPeaks, Golf, Scorpion and Mahjong deals can really be won — and why losing a game of solitaire often isn't your fault.

Sunroom Games · Updated October 10, 2026

You've played carefully, turned over every card you could, and now nothing will move. Did you make a mistake somewhere — or was that deal never winnable?

Often, it's the deal. Different solitaire games have very different odds, and in some of them a fair share of deals simply can't be won. Here's what researchers and puzzle fans have worked out, game by game — and, for the games where we couldn't find good published figures, what our own computer searches found.

The short answer

Those "if you could see" notes matter, so let's take each game in turn.

Klondike: luck plays a real part

Klondike is the hardest of these games to pin down, because you can't see most of the cards. For a long time nobody knew what share of deals could be won. In research first released in 2019, the computer scientists Charlie Blake and Ian Gent tackled a simpler question: how many deals could be won by a player who could see where every card was?

Their answer was about 90% of Draw 1 deals and about 82% of Draw 3 deals, with unlimited passes through the stock. So even with perfect knowledge, roughly one Draw 1 deal in ten — and nearly one Draw 3 deal in five — can't be won at all. A few deals are hopeless from the very start: in about 1 in 400 Draw 3 deals, there isn't a single move to make.

What about real games, with hidden cards?

Real Klondike is harder than that, because you have to make choices without knowing what's under the face-down cards. Sometimes two moves look equally good and only one of them works — and nothing on the table can tell you which. Even the researchers say the true odds for normal play aren't known with any precision.

You'll see all sorts of win rates quoted online for ordinary players, but none of them rests on solid evidence. What we do know is that computer programs built by researchers to play by the normal rules, without peeking, win only some of their games — one well-known program won a little over a third.

So if you win around a third of your Klondike games without using Undo, you're doing well. (Undo lets you peek at a hidden card and then change your mind, which makes the game quite a bit easier.) If you lose one after careful play, it may simply have been a deal that couldn't be won from the moment it was dealt.

FreeCell: nearly always winnable

FreeCell is the opposite of Klondike. Every card is dealt face up, so there's nothing hidden and almost no luck involved.

When FreeCell came to Windows, its deals were numbered from 1 to 32,000. In 1994 and 1995, a group of more than a hundred volunteers organized online as the Internet FreeCell Project set out to solve every one of them. They managed all but one: deal #11982, which has been shown to be impossible.

Puzzle fans later used computers to check the first million deals. Only eight of them can't be won: #11982, #146692, #186216, #455889, #495505, #512118, #517776 and #781948.

On Sunroom Games, the random and daily FreeCell games skip all eight, so you'll always get a deal that can be won. If you'd like to try the famous one anyway, you can play deal #11982.

Pyramid, TriPeaks and Golf: what our own tests found

We couldn't find trustworthy published figures for these three games, so we measured them ourselves. For each game we wrote a program that tries every possible line of play for a deal, knowing where every card is, and ran it on hundreds of random deals. The figures below are what we found — careful estimates, not the final word.

Pyramid: about one deal in three can't be won

Of 1,000 random Pyramid deals, 684 could be cleared — about two in three — even allowing as many passes through the stock as you like. The rest are lost from the start, usually because a card is buried under the only partners that could ever clear it. With just three passes through the stock, the odds are lower still.

TriPeaks: nearly always winnable, if you could see

TriPeaks did much better: 292 of 300 deals could be cleared by a player who knew where every face-down card was — about 97%. Real games are harder, because the face-down cards are a surprise until they turn over, so you won't always find the winning path even when there is one.

Golf: a little under half

Golf is tougher. By the classic rules, where Kings and Aces don't go on each other, a little under half of all deals can be cleared — 1,000 out of the 2,199 we checked. Letting Kings and Aces go on each other makes a big difference: then about 92% of deals can be cleared.

What we do with these numbers

On Sunroom Games, every new and daily game of Pyramid, TriPeaks and Golf is one we've checked can be won, so you'll never be dealt a game that was impossible from the start. Our strategy guides for Pyramid, TriPeaks and Golf will help you find the way through.

Spider: hard to measure

We haven't found reliable published figures for Spider, and it's a hard game to measure: with two decks and hidden cards there are far too many possibilities for a quick check like the ones above. What's clear from playing is how much the number of suits matters. With one suit, most games can be won with patience; with four, Spider is one of the hardest solitaire games there is. Our Spider strategy guide has tips for all three versions.

Yukon: planning beats luck

In Yukon every card is dealt at the start and there's no stock, so luck plays a smaller part than in Klondike, but some cards still start face down. We haven't found a published figure for how many deals can be won. In our tests, simply following our Hint button, which looks up to six moves ahead using only the cards you can see, won about half of 1,000 deals, and looking further ahead won more. Our Yukon strategy guide explains how to plan.

Scorpion: one of the hardest

Scorpion deals twelve cards face down, keeps just three in reserve and lets only Kings into an empty column, so it's a hard game to win. We ran our program, which knows where every card is, on 300 random deals. By the classic rules it found a win in 36 of them (about 1 in 8) and proved 155 impossible (about half); the other 109 were too big to settle within the search we allowed, so the true share of winnable deals lies somewhere between those two figures. With the easier Wasp rule, where any card can go into an empty column, it found a win in 260 of the 300 (87%).

Knowing a deal can be won isn't the same as winning it, because the face-down cards are a surprise. Simply following our Hint button, which only uses the cards you can see, won 17 of those 300 classic deals, and every one of the 17 was among the 36 the program could win. That's why, on Sunroom Games, every new and daily game of Scorpion is one the program has won. On those deals, following the hints alone wins about half the games.

Forty Thieves: about one deal in ten

Forty Thieves is famously hard: the usual estimate is that about 1 deal in 10 can be won. Every card in the layout is face up, so the only surprise is the order of the stock, but you go through it just once, and two decks give you a lot to sort out. Our program, which knows the order of the stock, found a win in about 1 in 13 random deals (1,000 of the 12,605 we tried) within the search we allowed. It couldn't settle most of the others either way, so that's a lower limit rather than the true share.

A second pass through the stock changes the game. With it, our program won 175 of 200 random deals (about 9 in 10), and simply following our Hint button, which never looks at the stock, won 162 of the 200. On Sunroom Games, every new and daily game of Forty Thieves is one our program has won by the classic rules, and its Easier setting gives you that second pass. On those deals, following the hints alone wins about 4 games in 10 by the classic rules, and 96 in 100 on Easier.

Canfield: winnable, but rarely won

Canfield is a puzzle of a different kind: most deals can be won, but few are. One computer study, reported on Wikipedia, found about 71% of 50,000 random deals winnable, while the usual estimate for people playing three cards at a time is about 1 game in 30. Our own program, which knows where every card is, found a win in about half the random deals it tried (1,000 of 1,946) within the search we allowed. The gap is the hidden cards: the reserve and the stock keep their secrets until you play them. Turning one card at a time instead of three makes a great difference: on the deals we hand out, following our Hint button won 795 of 1,000 games turning one card at a time, against 130 turning three. On Sunroom Games, every new and daily game of Canfield is one our program has won turning three at a time.

Mahjong solitaire: depends how the tiles are dealt

Most computer versions of Mahjong solitaire deal the tiles at random. A 2012 study of ten million random deals of the classic Turtle layout found that about 3% couldn't be cleared even by a player who could see under every tile. In real play, where you can't see what's underneath, more games are lost along the way: two identical tiles can end up stacked on top of each other with no third copy left to free them.

Our Mahjong works differently. Each game is built by placing the tiles in pairs along a planned path to a clear board, so every game we deal can be cleared. You can still get stuck by choosing the wrong pair along the way — that's the puzzle — but the Undo and Shuffle buttons will always get you back on track. Our Mahjong solitaire tips explain how to avoid the most common traps.

Sudoku: always solvable — if it's a proper puzzle

A proper Sudoku has exactly one solution, and you can reach it by logic alone. Every puzzle on Sunroom Games is checked for both: it has one solution, and it can be solved with the techniques for its level, without guessing.

Puzzles printed elsewhere aren't always so careful. If a Sudoku has two possible solutions, you'll eventually reach a point where you must guess — which feels unfair, because it is. Mathematicians have also worked out that a Sudoku needs at least 17 numbers filled in to have only one solution. Puzzles with fewer starting numbers always have more than one answer.

Minesweeper: the guessing problem

In the classic game, only the first square you click is sure to be safe, and many boards reach a point where no reasoning can tell you what's under the next square. In our tests, the standard reasoning (one number at a time, comparing two numbers, and counting the mines left) could finish only about 28 in 100 random Easy boards, 15 Medium and fewer than 2 Hard boards. Every other board needed a guess somewhere.

On Sunroom Games, every Minesweeper board starts with an open area and is solved by our computer, using only that reasoning, before you see it, so you never have to guess. Our Minesweeper strategy guide teaches the patterns.

Word search: winnable every time

Every word in our word searches is really there, and each one appears exactly once. Our puzzles never hide words backwards, so if you can't find one, check the diagonals running down to the right.

What this means when you play

• Losing isn't failing. In Klondike especially, some deals were lost before you made your first move.
• Use Undo to explore. In Klondike and Mahjong, trying a different choice is the best way to learn which decisions matter.
• Choose the game for your mood. Klondike for a relaxed game where luck plays its part, FreeCell for a puzzle you know can be solved, Pyramid, TriPeaks or Golf for a quick game that's always winnable here, and Sudoku for pure logic.

Our Klondike strategy guide and FreeCell strategy guide will help you win more of the games that can be won.

Play Klondike Solitaire