Klondike Solitaire for real people: what to expect when you just deal and play

Most people don't sit down to Solitaire with a strategy. They deal, flip a card, drop an Ace up top and keep going until the moves run out. So we asked a simple question: if you play that way — one move at a time, no peeking ahead, never taking a move back — how often do you actually win?

To find out, we built a computer player that plays exactly like that and let it loose on 10,000 shuffled decks of the classic Turn 1 game you see on this page. It won 42.3% of them. That is your realistic starting line. Everything on this page is about what pushes you above it: a little planning, the Undo button, and knowing which deals are worth fighting for.

The game is free, it runs right in your browser, and there is nothing to install or sign up for. Deal a hand, and keep reading while the cards are out.

Your odds on this page, in one table

Look at the first and last rows together. Our no-look-ahead bot wins 42.3%; real players on solitairex.io won 44.8% across more than 900,000 games. The typical human, in other words, plays only a little better than "make the obvious move." The gap between that and the solver's figure is the room you have to improve.

How Klondike works here

You get one standard 52-card deck, laid out like this:

• Seven columns (the tableau). The first has one card, the seventh has seven. Just the last card in every column is dealt face up.
• The stock: the 24 cards left over, face down in the corner. Click it to turn one card at a time onto the waste pile.
• Four foundations, one per suit, which you fill from Ace up to King.

In the columns you build downward, alternating red and black: a black 9 on a red 10, a red 4 on a black 5. You can move a whole face-up run as a unit. When a column empties, only a King (or a run headed by a King) can move into it. When the stock runs out, you can turn the waste back over and go through again — as many times as you like in this version.

You win when all 52 cards are on the foundations.

Playing on autopilot vs. playing with a plan

Our bot is a fair picture of autopilot play. It follows a fixed checklist: put any card on a foundation, then make any move that turns over a face-down card, then place a King in an empty column, then play from the waste, and only then draw. It never looks two moves ahead, and it never undoes.

That checklist is not bad — it is roughly what a casual player does — and it still wins more than four games in ten. But it leaves a lot on the table. When our solver was allowed to see every card, it proved at least 76.6% of deals winnable, and couldn't finish the search on another 21.3%. Only 2.1% were proven impossible. Most of the games the bot loses were not lost at the shuffle.

A real player can't see face-down cards, so you won't reach the solver's number. But you have two things the bot doesn't: you can think ahead, and you can press Undo.

Habits that lift you above the 42% line

1. Undo is not cheating — it's scouting

The bot never takes a move back, and that alone costs it games. When you reach a fork — two columns you could dig into — try one. If it turns over nothing useful, undo and try the other. There is no penalty for it here and no score to protect.

2. Dig where the most cards are hidden

The seventh column hides six cards; the second hides one. When you have a choice, free cards from the taller columns first. That's where the stuck Aces and low cards usually are.

3. Don't worry too much about a buried Ace

In our 10,000 deals, 88% started with at least one Ace hidden under the tableau, and on average 1.6 of the four Aces were face down. It barely mattered: the bot won 41.6% of deals where two or more Aces were buried and 44.5% where none were. A bad-looking opening is usually still a live game.

4. Keep an empty column until a King needs it

Only a King can go into an empty column. Clearing one just to clear it is fine, but don't move a King into it unless doing so frees something — otherwise you may block a better King later.

5. Use the stock more than once

Winning games in our test went through the stock about 2.3 times on average. If you've gone through once and feel stuck, flip again and watch for the one card that unlocks a column.

Where casual games usually go wrong

• Sending cards to the foundation too early. A red 6 sitting on the foundation can't catch a black 5 later. If you still have low cards of the other color buried, hold back.
• Playing from the waste first. Moves that uncover a face-down card are worth more than moves that simply use up the waste card.
• Walking away at the first dead end. Undo a few moves and look for the branch you skipped.
• Treating a no-move opening as a lost game. In 8.6% of our deals, none of the seven face-up cards could move anywhere at the start. That just means you begin with the stock.

Features you get on this page

This game is completely free and runs in any modern browser on a computer, tablet or phone. There's nothing to download and no account, email or sign-up. The cards are large and easy to read, with a zoom control if you want the board bigger. You get a Hint button, unlimited Undo, Restart for the same deal, and Auto Play to finish a game once it's solved. You can drag cards or just tap one and then tap where it goes. The game tracks wins, time and moves for the current game, you can switch the sound off, and there's fullscreen mode and a choice of table colors. No gambling, nothing to buy, no timer pushing you.

When you want a change, the Turn 3 version is the tougher cousin, FreeCell shows every card from the start, and Spider is the one to try next if you like long runs. The full game list has the rest.

The research behind these numbers, step by step

The play-solitaire-free-online.us team at Solitairen OOD ran this study in September 2026. We wanted numbers for the game exactly as it plays on this site, not for Klondike in general, so here is precisely what we did.

1. Read the rules from our own game code. The rules above — one card per draw, unlimited trips through the stock, only Kings into empty columns, whole face-up runs can move, and one card allowed back off a foundation — were taken from the game's JavaScript, not from a rulebook.
2. Wrote the game again as software. We wrote Klondike again from scratch and checked over 2,000 test games (1,000 at Turn 1 and 1,000 at Turn 3) that the bot only ever made legal moves.
3. Dealt numbered decks. Every deal comes from a numbered seed (1 to 10,000), so exactly the same shuffles can be played again. These are our own deal numbers, not the numbers from Windows Solitaire or from this site's game.
4. Chose two very different players. The bot sees only what you see and never undoes, so its result is a floor. The solver sees every face-down card, which no human can, so its result is a ceiling. Real skill lands in between.
5. Ran the games. Every one of the 10,000 deals was played through by the bot. The solver worked through 1,000 deals, stopping each search after 150,000 positions; a deal it hadn't finished by then is counted as undecided rather than guessed.
6. Compared with published research. A peer-reviewed study by Blake & Gent (Journal of Artificial Intelligence Research, 2026) found 90.48% of Turn 1 deals winnable when every card is known. Our solver's range, 76.6% to 97.9%, contains that figure.
7. Put error bars on every percentage. A "95% range" is the band the true value almost certainly sits in. For the bot, 42.3% could honestly be anywhere from 41.3% to 43.3%.
8. Wrote down what the numbers can't tell you. The bot is simple, so a thoughtful player with Undo should beat it comfortably. The solver's wide range comes from the 21.3% of deals it left undecided. And neither says anything about your personal luck on the next deal.

The code behind the test

All of the simulations are written in Python 3.11 using only its standard library — no outside packages. Klondike lives in its own module, klondike.py, which contains both the checklist bot and the solver. A shared script, runner.py, hands the deals out to 19 worker processes running side by side on a 20-core Windows machine. That's why the 10,000 bot games took about 14 seconds, while the far heavier solver run on 1,000 deals took about 17 minutes. The solver is a depth-first search that remembers positions it has already visited, so it never explores the same one twice, and it only makes automatic foundation moves that can never hurt. The 95% ranges use the Wilson score method, a standard way to put error bars on a percentage.

Parts of the simulation code, and a first draft of this page, were written with help from Claude, an AI assistant made by Anthropic, which also reviewed the code. The AI didn't supply any of the figures. Every number came out of running the code, and we checked the results against the published study before putting them here.

By Admin, play-solitaire-free-online.us (Solitairen OOD). Research dated September 18, 2026.

Questions new players ask

Is a 42% win rate good?

It's roughly what you get from playing the obvious move every time. Real players on solitairex.io won 44.8% of their games, so if you're winning around four or five in ten, you're right in line with most people. Using Undo well is the quickest way to climb.

Why did I lose a game that looked easy?

Usually a card got locked in the wrong place — often an early foundation move or a King that went into the wrong empty column. Only 2.1% of deals were proven impossible by our solver, so most losses can be turned around by replaying the deal with Restart.

Should I start with Turn 1 or Turn 3?

Turn 1, the version on this page. Our bot won 42.3% at Turn 1 but only 10.6% at Turn 3. Move up when Turn 1 starts to feel routine.

Does the Hint button make it too easy?

It points out a legal move, not the best plan. Think of it as a second pair of eyes when you've missed something, not an autopilot.

Does it work in a phone browser?

Yes. Everything works in a phone browser: tap a card, then tap where you want it, and use zoom if the cards feel small.