Why Every Deal Is Winnable: The Math Behind Solvable Solitaire
If you've played a lot of Klondike, you've almost certainly hit a deal where no move on the board could possibly lead to a win. It's not a skill issue — some deals are mathematically unsolvable from the first card. For a long time this was just part of the game. Now it doesn't have to be.
How many Klondike deals are winnable?
Exact numbers vary by study and by whether you allow looking at face-down cards, but the commonly cited figure is that around 79-82% of random Klondike (Draw-1) deals are theoretically winnable with perfect play and full information. In Draw-3 mode the number drops — typically quoted between 35% and 50% depending on methodology — because restricted stock access closes off lines that Draw-1 leaves open.
These percentages come from solvers: programs that exhaustively search the game tree to determine whether a given deal has at least one path to a complete foundation. The brute-force search space is enormous, so real solvers use pruning heuristics, transposition tables, and iterative deepening to keep the work tractable.
What "guaranteed winnable" actually means
When a site like SuperSolitaire promises every deal is winnable, here's roughly what's happening:
- A random shuffle is generated.
- A solver tests the deal against the rules you're playing (Draw-1 or Draw-3).
- If the solver finds at least one winning line, the deal is dealt to the player. If not, it's discarded and a new shuffle is generated.
Critically, winnable does not mean easy. It means a win exists. Finding that win is still on you. Some winnable deals have exactly one solution and require near-perfect play; others have many paths and forgive small mistakes.
Why solvers are harder than they look
Klondike looks simple, but its state space is huge. The number of distinct reachable positions from a single deal can run into the millions. Worse, cycles are possible — the same position can be reached through different move sequences — so naive depth-first search explodes combinatorially.
Good solvers handle this with a few tricks:
- Transposition tables — remember every position you've already seen, skip if you revisit it.
- Move ordering — try foundation moves and face-down reveals first; they prune the tree fastest.
- Dead-state detection — some positions are provably lost (e.g., foundations can't receive any more cards and the tableau is locked). Prune those early.
What this changes for the player
Winnable-only Klondike is a subtly different game from random Klondike. In random mode, giving up on a stuck board is rational — it might be unwinnable. In winnable-only mode, giving up is always a strategic mistake, because you know a solution exists.
This changes how you should play. When you're stuck:
- Look harder for a non-obvious move.
- Consider moves that temporarily make your position worse but unlock new cards.
- Use the hint system — on a winnable deal, it will always have something to offer.
Should the solver always find it?
In theory, yes — the solver guarantees a line exists. In practice, hint systems usually use fast heuristics rather than running a full solve on every click, so occasionally the suggested move isn't the globally optimal one. On truly difficult deals you may need to trust your own planning more than the hint.
The philosophical bit
Winnable-only mode turns Klondike from a game of luck and skill into a game ofpure skill. Some players love this — every loss is a lesson. Others miss the occasional "this deal was impossible" release valve. Both are valid.
What's harder to argue with: if you want to actually improve at Klondike, playing deals you know are winnable is the faster path. Every loss is now your fault — and that's what makes you better.