← Blog · 6 min read · Updated May 2026

Why FreeCell Has a 99.999% Win Rate

FreeCell is the rare solitaire variant where, given perfect play, nearly every deal is winnable. The original Microsoft FreeCell shipped with 32,000 numbered deals; exactly one — deal #11982 — was eventually proven unsolvable. The other 31,999 each have at least one winning move sequence. That's a solvability rate of 99.997%. When the analysis was extended to all 8 × 10⁶⁷ possible FreeCell deals, the estimate landed at about 99.999%. Why? And what does this mean for you sitting in front of an actual game?

What Makes FreeCell Different

Two structural features explain almost everything:

In contrast, Klondike Turn 3 hides 21 cards in the stock and gives you no buffers. Spider gives you two decks of complications. FreeCell's combination of full information and flexible buffers is what makes 99.999% achievable.

The Math Behind 99.999%

Researchers Don Woods and Adrian Ettlinger ran systematic solver attempts on the original 32,000 Microsoft deals in the 1990s. Their solvers (using depth-limited search with various pruning heuristics) eventually found wins for 31,999 deals. The remaining one — #11982 — resisted every approach. A 2019 mathematical proof confirmed that #11982 has no winning move sequence under FreeCell's rules. It's the one provably impossible deal in the original set.

Subsequent work has analyzed random deals beyond the Microsoft 32,000. Estimates converge on a per-deal unsolvability rate of around 1 in 100,000, producing the famous "99.999%" figure. It's not exactly 100% — some deals really are dead — but it's close enough that calling FreeCell "nearly always solvable" is accurate.

The Legend of Deal #11982

Deal #11982 is the most famous specific deal in solitaire history. When Microsoft FreeCell shipped in 1992, the company offered no hint that any deal was unwinnable. Players who hit #11982 spent hours trying every combination and convinced themselves they were missing something obvious. Web forums in the late 1990s collected hundreds of reports from frustrated players before Woods and Ettlinger's solver formally confirmed it was dead.

Microsoft eventually updated the game to include a note acknowledging the existence of unsolvable deals, with #11982 listed by number. It became a small badge of honor for FreeCell veterans to identify the deal and quit immediately without wasting time.

What This Means for Your Win Rate

If 99.999% of deals are solvable, why don't real players win 99.999% of hands? Because solvability is theoretical, and play is human.

Typical real-world win rates:

The gap between you and 99.999% is mostly impatience. FreeCell rewards careful planning; most losses are unforced errors — a move made too quickly, a free cell wasted, an empty column filled with the wrong card. Slow down. The puzzle is solvable.

How To Capture the 99% Win Rate

Three rules unlock most of the gap:

  1. Plan a winning path before move one. All cards are visible. Trace the sequence: which Aces go up first? Which 2s? Which columns get cleared? This sounds tedious; it takes 30 seconds and triples your win rate.
  2. Free cells are temporary, never strategic. A card in a free cell is one card that can't reorder itself. Park briefly, retrieve fast. If three of your four free cells are full and you're not on the cusp of solving, you've miscounted.
  3. An empty column is worth more than three free cells. The number of cards you can move as a group is (empty cells + 1) × 2^(empty columns). Each empty column literally doubles your move power. Clear column 1 (or any column with few cards) as early as possible.

The Bigger Lesson

FreeCell is the solitaire that proves the deck isn't your enemy. Every loss is almost certainly a move you made wrong. That mindset transfers to other solitaire variants: even in Klondike Turn 3, where dead deals are common, many "losses" are still skill errors disguised as dealer's luck.

Try the math yourself on the FreeCell page. Play 50 hands deliberately. Your win rate will climb faster than you expect.

Related