Spider Winnability: What the Numbers Say

Spider Solitaire is not a single game with a single win rate. It is three games — 1-suit, 2-suit, 4-suit — and each has its own winnability distribution. A player who wins 90% of 1-suit games might struggle to win 20% of 4-suit games on the same skill. At the Research Desk we care about those numbers because they shape how players approach the game, and because the public numbers circulating on the web are frequently wrong by large margins. This page is our attempt to summarize what the data actually says, with methodology notes so you can decide which numbers to trust.

A note on what this page is and is not. We are summarizing the range of winnability estimates that the public community has converged on, supplemented by our own internal play logs and spot checks against open-source Spider solvers. We are not publishing fresh solver runs on this page; when we do publish our own Spider solver data, it will be documented here with the full method. Until then, think of these figures as informed estimates with confidence ranges, not authoritative constants.

In 1-suit Spider, all 104 cards are the same suit, which means every descending run is movable as a block. Human win rates cluster between 85 and 92 percent for players who have played more than fifty hands and learned the basic column-clearing framework. Beginners win 1-suit games in the 60 to 75 percent range as they learn the mechanics; experienced players rarely dip below 85 except on deals with unusually bad initial distributions.

The reason 1-suit is so high is structural. With no suit constraint, the only way a 1-suit deal becomes unwinnable is if the face-down distribution traps low cards under too many high cards to recover. That happens occasionally — roughly 3 to 8 percent of 1-suit deals are estimated to be unwinnable even with perfect play — but it is rare. Most 1-suit losses are player mistakes, not structural dead ends. That fact makes 1-suit valuable as a diagnostic mode: when you lose, you can almost always identify the move that did it.

The 85 to 92 percent range we cite aggregates our own play logs, community-reported win rates from solitaire forums, and data from other Spider implementations that publish aggregate stats. The numbers converge across sources, which gives us confidence the range is roughly correct. The spread within the range (7 percentage points) mostly reflects differences in player care: a player who pays attention wins closer to 92; a player who clicks quickly wins closer to 85.

Beginners should not be discouraged by the 60 to 75 percent beginner range. Almost everyone climbs out of it within the first thirty hands of deliberate 1-suit play, because the game mainly rewards a small number of teachable habits: expose face-down cards early, pace stock deals, protect empty columns. The leap from 75 to 85 percent is a function of internalizing those habits. After that, the remaining seven points of improvement come from attention to detail on specific deals.

In 2-suit Spider, the deck is split between two suits (typically Spades and Hearts). Group moves now require same-suit runs, which is the constraint that turns Spider into a planning game. Human win rates typically sit between 60 and 70 percent for intermediate players and between 70 and 80 percent for skilled players. The range is wider than 1-suit because 2-suit is where player skill begins to matter.

2-suit is the middle-difficulty sweet spot. It is hard enough to reward planning but forgiving enough that careful play wins most of the time. That combination makes 2-suit the version where most players spend the bulk of their hours, and where most skill-building happens. It is also the version where the win-rate-per-unit-time-invested is highest — an hour of deliberate 2-suit practice produces more learning than an hour of either 1-suit or 4-suit for most players.

The structural unwinnability rate for 2-suit is higher than for 1-suit. We estimate 8 to 15 percent of 2-suit deals are unwinnable even with perfect play, though the precise number depends on how the specific implementation shuffles and deals. That estimate gives a theoretical ceiling of around 85 to 92 percent win rate with a perfect solver, which matches the observed upper bound for the strongest human players.

One caveat: 2-suit win rates are especially sensitive to tempo. Players who move fast win noticeably less than players who pause to plan. The range we cite assumes unhurried play. Blitz-style 2-suit (fast clicks, minimal planning) bottoms out around 45 to 55 percent, which is effectively luck.

The reason 2-suit is the sweet spot deserves a closer look. At 1-suit, the game rewards correct habits and punishes sloppy ones, but the ceiling is low because there is not much complexity to reward. At 4-suit, the game is so punishing that even correct play fails half the time. In between, 2-suit is the mode where strategic thinking translates most directly into win-rate gains per unit of attention. Every small improvement in suit discipline, empty-column management, or stock timing shows up cleanly in the win column. That is why 2-suit is also the mode where tracking your own win rate over time is most informative — the signal is cleaner than at either extreme.

In 4-suit Spider, all four suits are present in equal distribution (two of each suit), which means only a quarter of random descending placements match suits. The constraint is severe, and human win rates reflect that severity. Most players who claim to play 4-suit seriously win somewhere between 5 and 15 percent of hands. Strong players push that to 25 to 35 percent. World-class players can reach 40 to 50 percent under ideal conditions, but those ceilings are rare.

The 5 to 15 percent range is where most casual players live. It is discouraging if you expect Spider to be a high-win-rate game, but it is the honest number. The reason the range is so wide is that 4-suit rewards specific expert habits — suit tracking, lookahead, empty-column discipline — and players with those habits live in the upper half of the range while players without them live in the lower half.

Structural unwinnability is a larger factor in 4-suit. We estimate 40 to 55 percent of 4-suit deals are unwinnable even with perfect play, which puts a theoretical ceiling of 45 to 60 percent on any player. The gap between the theoretical ceiling and observed expert play (around 40 to 50 percent) is small, which suggests top players are approaching the limits of what the game allows.

That theoretical ceiling estimate is contested. Community solvers disagree by five to ten percentage points, depending on solver design, look-ahead depth, and treatment of the stock. The number we cite reflects our best reading of the consensus range, but it is an area where numbers should be treated as estimates rather than firm facts.

What makes 4-suit so hard, mechanically, is the interaction of suit matching with face-down cards. With 44 face-downs at deal time and four suits, the number of possible future states the player must reason over is dramatically larger than in 1-suit or 2-suit. Every face-down card is uncertainty, and every uncertainty compounds when combined with the suit constraint.

The gap between casual and strong 4-suit players is the single largest skill gap in solitaire. At the low end, players are effectively playing randomly, their wins determined by deal quality. At the high end, players recognize cascades, track suits mentally, plan five-to-seven moves ahead, and maintain two empty columns through most of the mid-game. The strategic distance between those two profiles is enormous, and it is all compressed into the same mode with the same rules. No other solitaire variant has a comparable skill-expression range.

Spider solvers exist, but they are harder to build than FreeCell solvers. The reasons are hidden information (44 face-down cards at deal time) and the branching factor from stock-deal timing. A FreeCell solver can brute-force the move graph because every card is visible. A Spider solver must either reveal the random deal in advance (cheating from a player perspective but fine for theoretical winnability) or run a stochastic search.

Academic and community Spider solvers have published theoretical maximum win rates in the following ranges: roughly 92 to 98 percent for 1-suit, 82 to 90 percent for 2-suit, and 45 to 60 percent for 4-suit. These are ceilings — what a perfect player with full knowledge could theoretically win. Real human win rates sit below these ceilings because humans cannot see face-down cards and cannot perfectly plan lookahead.

The solvers we track include several open-source projects maintained by hobbyist programmers and a handful of academic papers that analyze Spider as a game-theoretic problem. We do not currently run our own Spider solver at the Research Desk — we reference published results and cross-check against aggregate community data. When we eventually publish our own solver runs, we will document the methodology here.

One detail that matters when comparing solver results: the definition of “perfect play” varies across solvers. Some solvers assume the player can see face-down cards (omniscient play); others do not. A solver that cheats on hidden information reports a higher theoretical win rate than one that plays the actual game. When you read a Spider winnability figure online, check which definition the solver used. Many widely-cited numbers come from omniscient solvers and therefore overstate what a human can achieve.

Another detail: treatment of the stock deal. Some solvers evaluate only the winnability of the initial tableau and ignore the stock; others simulate the entire 50-card stock through all five deals. The latter is more realistic but more expensive to compute. Full-stock solvers produce lower theoretical win rates than tableau-only solvers, because the stock randomness is itself a source of unwinnability.

Spider deals are not all winnable. Unlike FreeCell, which is solvable 99.999 percent of the time, Spider has significant rates of structurally unwinnable deals in every difficulty mode. The percentages above (3 to 8 for 1-suit, 8 to 15 for 2-suit, 40 to 55 for 4-suit) represent the proportion of random deals that no player, however skilled, can win.

The consequence is that Spider players are always playing against two things: the game itself and the random draw. In 4-suit especially, a player can execute perfect strategy and still lose half the time because the underlying deal was not solvable. That is a fundamental difference from FreeCell, where losing almost always means the player made a mistake.

The practical implication is that Spider players should expect losses that are not their fault. Restarting a deal that looks broken is usually the right move — the deal probably is broken, or at least sufficiently hard that the effort is better spent on a fresh deal. This is not an excuse to abandon hands that are merely challenging; it is recognition that Spider's random component is real and that fighting a genuinely unwinnable deal is lost time.

How do you tell an unwinnable deal from a hard one? The honest answer is that you often cannot. A conservative heuristic: if you have played twenty moves, made no progress on face-down cards, and cannot see a sequence that leads to an empty column, the deal may be structurally broken. A more forgiving heuristic: give every deal at least to the third stock deal before declaring defeat, because early Spider positions often look worse than they are. Most players under-fight early and over-fight late; the right balance is the reverse.

Spider has clear skill curves, though they differ by mode. In 1-suit, most players reach their ceiling within twenty to fifty hands, because the game does not reward complex strategy once the basics are learned. In 2-suit, improvement continues for two to three hundred hands, because suit discipline is a learnable skill with many edge cases. In 4-suit, improvement is slower and longer — strong players report continued learning after a thousand or more hands.

A rough curve for a typical new player moving from complete beginner to skilled competence, across all modes: 50% win rate in 1-suit improving to 85% within thirty hands; 35% win rate in 2-suit improving to 65% within a hundred hands; 5% win rate in 4-suit improving to 20% within two hundred hands. Those numbers vary widely with practice habits and strategic awareness, but they give a sense of how the difficulty modes differ in how fast skill accumulates.

Players who learn from losses, not just wins, climb faster. After a lost hand, naming the move that broke it — or the earliest move you would have played differently — converts the loss into a lesson. Players who replay the last few moves mentally after each loss improve roughly twice as quickly as players who move immediately to the next deal, based on our own practice logs. The mechanism is simple: most losses trace to a handful of repeated errors, and naming those errors lets you catch them earlier next time.

The inflection points are informative. In 1-suit, the big jump happens around hand ten, when the player internalizes the column-clearing framework. In 2-suit, the big jump happens around hand fifty, when suit discipline stops requiring conscious effort. In 4-suit, there is no single big jump — it is more like a steady accumulation of specific cascade recognition patterns.

One implication of these curves is that switching modes during practice is a legitimate training strategy. Players who only grind one mode plateau; players who rotate modes build different skills at different rates and then combine them. A reasonable rotation for someone trying to raise their 4-suit win rate is: play five hands of 4-suit to see current position, play twenty hands of 2-suit to drill suit discipline on a cleaner feedback loop, play ten hands of 4-suit to test whether the drilled skills translate. Most improvement comes from that kind of deliberate cross-mode practice rather than from sheer volume in one mode.

The numbers on this page come from three sources: internal play logs maintained by the Strategy and Research Desks, aggregated community data from public solitaire forums and Spider implementations that publish aggregate stats, and published solver results from hobbyist programmers and academic papers. Where sources disagree, we cite the convergent range rather than a single figure.

Confidence in the numbers varies by mode. The 1-suit and 2-suit ranges are well-established across sources. The 4-suit ranges are less settled because solver design choices and player skill vary widely. Treat the 4-suit figures as estimates with confidence intervals of five to ten percentage points. Improvements in solver technology may shift these figures over time.

The sample sizes behind these numbers also vary. Our 1-suit figures are backed by thousands of human hands across many players. Our 4-suit figures are backed by thousands of hands too, but the wider variance inside the data means the confidence intervals are wider in percentage-point terms. This is the same pattern you see with other chance-heavy games: more data is needed to pin down the signal, because the noise floor is higher.

For a fuller account of how the Research Desk runs simulations, cites sources, and reports confidence intervals, see our Solitaire Research Methodology page.

A few specific sourcing notes. Human win-rate figures draw primarily on our own play logs, cross-checked against community forums (notably the Solitaire subreddit and Pagat community discussions) and against aggregate stats from other solitaire implementations that publish them. Solver figures draw on published open-source Spider solvers and academic papers that analyze Spider as a partially-observable Markov decision process. Unwinnability rate estimates are the area where sources disagree most; we cite ranges that span the center of the disagreement rather than picking a single source to privilege.

When the Research Desk eventually runs its own Spider solver, we will replace these estimates with figures we can fully document. For now, we want the numbers on this page to be useful without overclaiming. The ranges are wide because the underlying data is wide; do not take a single figure as the truth.

FreeCell is famous for being solvable: 99.999 percent of FreeCell deals can be won with perfect play, and only a handful of specific deals (the 8 identified in the Microsoft 32,000-game numbered set) are proven unwinnable. Spider is the opposite. Even in 1-suit, a small fraction of deals are unwinnable; in 4-suit, it might be a majority.

The structural reason is information. In FreeCell, every card is visible at the start, so the game is a pure deterministic puzzle. Every unwinnable FreeCell deal is a quirk of the specific card distribution. In Spider, 44 cards are hidden at the start, and 50 more enter the game through the stock. That randomness is the reason Spider has significant unwinnable-deal rates — you simply cannot plan perfectly when you cannot see half the cards.

That difference shapes how players should approach the two games emotionally. In FreeCell, losing almost always means you could have played better. In Spider, losing sometimes means the deal was never going to work. Internalizing that difference is important for sustained Spider practice — it prevents the discouragement that comes from treating every loss as a personal failing.

The structural contrast also changes what a win-rate number means. A 90 percent FreeCell win rate is a judgment of player skill almost entirely. A 90 percent 1-suit Spider win rate is similar. But a 40 percent 4-suit Spider win rate contains two signals: player skill and deal luck. Comparing win rates across games without that caveat is misleading. A player who wins 85 percent of Klondike hands and 30 percent of 4-suit Spider hands is not necessarily a better Klondike player than Spider player — Klondike has fewer unwinnable deals.

Put differently: FreeCell is closer to chess in its solvability profile (almost every position is theoretically drawable or winnable), while Spider is closer to bridge (a significant fraction of hands are simply losing hands, regardless of play quality). That analogy is not precise, but it captures the right intuition for someone coming from FreeCell to Spider: expect the random component to be larger.