Short answer: if you play Klondike one card at a time and simply follow the rules as you go, expect to win somewhere around 4 games in 10. Our own testing puts the floor at 42.3% of deals, and the best public data from real people playing online sits at 44.8%. So if you win a little under half your games, you are normal. You are not doing anything wrong.
That number surprises people. Solitaire has a reputation as a gentle time-passer, so a loss can feel like a personal failing. It usually isn't. Some deals simply cannot be won, and many more can only be won if you happen to make the right choice at two or three quiet moments early on. Below is where the numbers come from, what the gap between a casual player and a perfect player really means, and three habits that move your win rate up without turning the game into homework.
How we got our figures
In September 2026 our team ran two simulations on this site's own Klondike rules: a rule-following bot that sees only the face-up cards and never undoes a move played 10,000 deals in Turn 1 mode, which gives us a floor, and a solver that can see every card in the deal analyzed 1,000 deals, which gives us a ceiling.
Those two words matter. A bot that cannot see hidden cards and never takes a move back is a rough stand-in for a relaxed human player, not a measure of how winnable the game is. The solver, which knows where every card sits, tells us how many deals are winnable in principle. Both are useful. Confusing them is how bad statistics get repeated for years.
The floor: 42.3% of 10,000 deals
Our bot won 42.3% of 10,000 Klondike Turn 1 deals, with a 95% range of 41.3% to 43.3%. That range is the honest version of the result: run the same test again and you should land inside it. Please don't quote it back as a single tidy number.
Real people: 44.8% of 913,106 games
The public site solitairex.io published a month of its own play data: 913,106 games of classic Solitaire with 409,374 wins in the month before 16 July 2025, a win rate of 44.8%. Think about what that says. A whole month of human players, with all their patience and pattern-spotting, finished barely two percentage points above a bot that never looks back.
The ceiling: what's actually winnable
Our solver proved that at least 76.6% of 1,000 deals were winnable. It proved 2.1% genuinely impossible and ran out of time on 21.3%, so the true figure is higher than 76.6%, we just can't claim credit for what we didn't finish computing.
Academic work goes further. Charlie Blake and Ian Gent, writing in the Journal of Artificial Intelligence Research in February 2026, report the winnability of Klondike as 81.945% with a 95% confidence interval of ±0.084% for the classic three-card draw in the "thoughtful" version, where the player knows the rank and suit of every card. Their figure for the one-card draw is higher still, at 90.48%. In other words: turn the cards face up, hand the deal to a perfect player, and about nine deals in ten are beatable.
The gap is the whole story
Put the three numbers side by side:
| Who is playing | Win rate | What it measures |
|---|---|---|
| Our bot, face-up cards only, no undo (10,000 deals) | 42.3% (95% range 41.3-43.3%) | A floor for casual play |
| Real players, solitairex.io (913,106 games) | 44.8% | What people actually score |
| Perfect play, every card visible (Blake & Gent, 2026) | 90.48% (one-card draw) | A ceiling: how winnable the game is |
So roughly half of all Klondike deals are winnable but get lost anyway. That lost half is not bad luck. It is hidden information plus a handful of decisions that quietly close doors. You will never reach 90%, nobody can see through the face-down cards, but the space between 45% and, say, 55% is genuinely available to a casual player who picks up a few habits.
Three habits that move your win rate up
1. Dig where the pile is deepest
Every face-down card is a card you cannot plan around. Early on, the most valuable move is almost always the one that flips a new card, and the most valuable flip is in the tallest untouched column. Clearing a column with five hidden cards buys you five pieces of information; clearing one with a single hidden card buys you one.
When two moves look equally good, ask: which one shows me something new? Do that and take it.
2. Don't send cards to the foundations too early
Playing an ace or a two up top feels like progress, and it is. A four or a five can be a mistake. Low cards left in the tableau are the hooks that hold the black and red sequences together. Send a red 5 to the foundation and the black 6 you needed to park somewhere has nowhere to go.
A simple rule of thumb: keep a card in the tableau if either of the next lower cards of the opposite color is still hidden or still buried in the stock. Move it up when it can no longer help you.
3. Treat kings and empty columns as decisions, not rewards
An empty column is the strongest thing you own, and the moment you drop a king into it, it's gone. Before you fill it, look for the queen of the opposite color. A king with a queen ready to follow starts a working pile. A lone king with no queen in sight is a wall.
Likewise, don't move a king into a space just because the space exists. If you can hold the column open for another two or three moves to unstack a column you're digging into, that is usually the better trade.
One extra habit, free of charge: on this site, use Undo deliberately. Our bot never undoes a move, and that single restriction is a large part of why it sits at the floor. When you take a move back and try the other branch, you are giving yourself a little of what the solver has, a second look at the same position.
Turn 1 or Turn 3? It changes the benchmark
All the numbers above are for the one-card draw unless stated. Drawing three cards at a time is harder, because you can only reach every third card in the stock. Blake and Gent's own figures show the drop clearly: 90.48% winnable with a one-card draw against 81.945% ±0.084% with a three-card draw, even for a player who can see everything.
If you want an easier ride, play Klondike Turn 1. If you want the version most Americans grew up with on an office computer, play Klondike Turn 3 or Solitaire Draw 3, and judge yourself against a lower bar.
Other games sit in completely different places. FreeCell is famously kind: all 52 cards are visible from the start, and of the 32,000 numbered deals in the classic Microsoft set, only one has ever been shown to be unsolvable, a result long established by the volunteer FreeCell Project. Spider Solitaire with four suits, at the other end, will humble anyone. Comparing your Spider score to your Klondike score tells you nothing useful.
How to play here, with nothing in the way
You can test all of this on this site right now. Everything is completely free, there are no pop-ups and no paywalls, and nothing is required to start: no download, no install, no sign-up, no account, no email. Open the home page, pick a game, and deal.
- Large, clear cards and readable text, with a zoom control to make the board bigger, comfortable if your eyes get tired.
- Hint when a position looks stuck, and unlimited Undo so you can explore a line and step back.
- Restart to replay the same deal, and Auto Play to finish a game that's already solved.
- Drag and drop or simply tap to move cards, on desktop, tablet or phone in any modern browser.
- Wins, time and moves tracked for your current game, so you can watch your own rate instead of guessing.
- Sound can be switched off, plus fullscreen mode and table color settings.
Every feature is available to everyone. There is no gambling, nothing to buy, and no timer pushing you along, the site exists for the enjoyment of the game. Browse the full list on the games page, from Pyramid to Yukon.
Frequently asked questions
Is a 45% win rate good at Solitaire?
It is completely normal. Real players across 913,106 games on solitairex.io won 44.8%, and our own rule-following bot won 42.3% of 10,000 Turn 1 deals (95% range 41.3-43.3%). If you're near that, you're playing the game the way most people play it. Sixty percent and up on the one-card draw is strong casual play.
Are some Solitaire deals impossible to win?
Yes. Our solver proved 2.1% of 1,000 deals genuinely impossible and could not decide 21.3% of them within its time limit. Blake and Gent's published work implies a small but real share of deals are unwinnable even with every card visible. So a few of your losses were never yours to win.
Does using Hint or Undo count as cheating?
Only if you decide it does. They are learning tools, and both are free here with no limits. Our bot's inability to undo is one reason it sits at the floor of casual performance, taking a move back is how you discover why a line failed.
Which version should a beginner start with?
Start with Klondike Turn 1: it is the most winnable form of the classic game, so you see more finished boards while you learn. Move to Turn 3 once the habits above feel automatic.
Why does my win rate swing so much week to week?
Small samples. In a run of 20 games, winning 6 or winning 13 are both ordinary results at a true rate near 45%. That is exactly why we report a range, 41.3% to 43.3%, rather than a single number. Judge yourself over a hundred games, not ten.
Do I need an account to keep my stats?
No account, no email, nothing to sign. The site tracks wins, time and moves for your current game so you can see how you're doing, and that's it. More strategy pieces are on the blog.
Written by Admin for play-solitaire-free-online.us, operated by Solitairen OOD. Simulation figures come from tests our team ran in September 2026 on this site's own rules.
Sources
- Blake, C. & Gent, I. (2026). The Winnability of Klondike Solitaire and Many Other Patience Games. Journal of Artificial Intelligence Research, 85.
- The Winnability of Klondike Solitaire and Many Other Patience Games. JAIR article page
- Data Insights and Strategies to Improve Your Win Rate, solitairex.io