Klondike Turn 3: the same game, four times harder

Turn 3 is classic Klondike with one change: each time you click the stock, three cards turn over and only the top one is yours to play. The cards underneath stay locked until the one above them moves. It sounds like a small tweak. Our numbers say it is anything but.

We dealt 10,000 games of Turn 3 to a computer player that sees only what you see, and it won just 10.6%. On the Turn 1 version of the same study it won 42.3%. That is roughly a fourfold drop from one rule. A solver allowed to see every hidden card still proved at least 66.2% of deals winnable, so Turn 3 is far from hopeless. It simply punishes careless play much harder.

If a long, quiet evening calls for a proper challenge, this is it. Here is what we found, and what it means for how you play.

At a glance: Turn 3 in numbers

Our perfect-knowledge solver ran out of budget on 24.3% of Turn 3 deals, so our honest range is 66.2% to 90.5% winnable. The published 81.9% falls inside it.

Why three cards changes everything

At Turn 1 every card in the stock is available on every pass. At Turn 3 only every third card is reachable on a given pass, and which cards those are depends on what you have already played from the waste. Two things follow:

• Some cards you need are simply out of reach until you play something from the waste and shift the pattern.
• More deals are genuinely lost. Even seeing every card, our solver proved 9.5% of Turn 3 deals impossible, against 2.1% at Turn 1.

The upshot: at Turn 3 the stock is a puzzle in its own right, not just a supply of cards.

Rules of Turn 3 on this site

• Seven tableau columns of one to seven cards, bottom card face up; 24 cards in the stock; four empty foundations.
• Build the tableau down in alternating colours. Face-up runs can move together.
• Only a King, or a run headed by a King, may fill an empty column.
• Clicking the stock turns three cards onto the waste. Only the top waste card can be played.
• When the stock is empty, turn the waste over and deal again. There is no limit on passes here.
• Foundations build up by suit from Ace to King, and a card can come back down from a foundation if you need it.

Playing Turn 3 better: what the data suggests

Expect to lose more, and don't read too much into it

If you are used to winning four in ten at Turn 1, a win rate nearer one in ten at Turn 3 is not you getting worse. That is the game. Our computer player dropped by about three quarters when we changed nothing but the draw.

Learn the stock on your first pass

Winning Turn 3 games took our player 3.8 trips through the stock on average, against 2.3 at Turn 1. Since passes are unlimited here, the first one is cheap. Use it to see which cards come up, and hold off on waste plays that would change the order until you know what you are chasing.

A waste play shifts every later card

Each card you take from the waste changes which cards appear on top from then on. Sometimes playing one waste card you do not really need is the key to reaching the card you do. Strong Turn 3 players think about the stock this way, and our perfect-knowledge solver finds many of its wins by doing exactly that.

Uncover the tableau before anything else

The tableau is the one place where your choices are not filtered through the three-card rhythm. Our computer player's priority after foundation moves is turning over hidden cards, largest column first. In Klondike deals, 88% start with at least one Ace face down (1.6 of the four on average), and every card you free from the tableau is one you don't need to fish out of the stock.

Buried Aces cost less than you think

At Turn 3 our player won 10.1% of deals with two or more Aces hidden and 12.3% with none hidden. The difference is real but modest. A bad-looking start is not a reason to redeal.

Use Undo as a map

Our computer player never undoes a move, which is partly why it scores so low. You can. When the stock stops producing anything useful, step back a few moves and try playing a different waste card. That changes the three-card pattern and may bring your missing card to the top.

Turn 3 or Turn 1: which should you play?

It is the obvious question once you have seen the numbers, and the answer depends on what you want from a game.

• Want to finish most of what you start? Play Turn 1. Our no-peek player won 42.3% there, and at least 76.6% of deals were proven winnable with perfect knowledge.
• Want every win to feel earned? Play Turn 3. Around one game in ten fell to our computer player, and 9.5% of deals are impossible even for a solver that sees every card.
• Want to improve at both? Turn 3 is the better teacher. Because the stock is so restrictive, it forces you to plan ahead and to think about the order in which cards appear. Those habits carry straight back to Turn 1.

Keep your expectations in line with the numbers either way. A losing streak of five or six games at Turn 3 is entirely normal and says little about your skill. Over a long run of games, careful stock management and patient use of Undo are what lift your results above the floor our computer player set.

Playing here

Turn 3 on free-solitaire.nz is free, with nothing to download or sign up for. There is no account and no email. It works in any modern browser on a phone, tablet or computer.

• Large, clear cards with a zoom control.
• Hint, unlimited Undo and Restart.
• Auto Play once the game is won.
• Drag and drop, or tap to move.
• Wins, time and moves tracked for your current game.
• Sound off, fullscreen mode and a choice of table colours.

If Turn 3 is feeling a bit brutal tonight, Klondike Turn 1 is the same game with four times the win rate in our tests.

How we did the research, step by step

The free-solitaire.nz team ran this study in September 2026 to find out how often Turn 3, as it is played on this page, can really be won, and how much harder it is than Turn 1.

1. Reading the rules from this site’s own game code. The layout is classic Klondike: seven columns, a 24-card stock, building down in alternating colours, only Kings into empty columns. The difference is that the stock turns three cards at a time and only the top one is playable. Redeals are unlimited.
2. Rebuilding the game as a program. We rewrote the game in Python and checked that our computer player only made legal moves over 2,000 test games, 1,000 of them at Turn 3.
3. Dealing the cards. Deals 1 to 10,000, each from a numbered seed so the shuffle can be repeated exactly. They are the same deals as our Turn 1 test, so the only thing that changes is the draw.
4. Choosing two kinds of player. A bot that sees only what you see and never undoes a move gives a floor. A solver that sees every card, including exactly what each three-card turn will show, gives a ceiling. The solver was allowed up to 150,000 positions per deal.
5. Running it. The bot won 10.6% of 10,000 deals, against 42.3% at Turn 1. The solver took deals 1 to 1,000: it found a win in 66.2%, proved 9.5% impossible and left 24.3% undecided, so between 66.2% and 90.5% can be won with perfect knowledge.
6. Checking against published research. Blake & Gent, Journal of Artificial Intelligence Research, 2026 give 81.945% for this game with every card known. That falls inside our range.
7. Working out the margin of error. A “95% range” is the band the true figure falls inside 95 times out of 100 if you ran the same test again on fresh deals. The bot’s 10.6% came with a range of 10.0–11.2%.
8. What the numbers cannot tell you. The bot is a simple rule-follower and a lower bound; a thoughtful player with Undo wins more. The solver left about a quarter of deals undecided, so treat its result as a range.

Every deal is seeded, so the whole test can be repeated exactly.

The technical side

Everything runs in Python 3.11 with the standard library only. Turn 3 shares klondike.py with the Turn 1 game on our home page; a single setting switches the draw from one card to three. A shared runner.py spreads the deals across 19 worker processes using Python's multiprocessing, and each deal is shuffled from a numbered seed, so a given deal number always produces the same cards.

• The bot uses the same priority list as at Turn 1 (foundations, then uncovering hidden cards, then Kings, waste and draws), but it can only reach the top card of each group of three. The 10,000 deals ran in about 8 seconds.
• The solver sees every card and searches depth-first, keeping a record of visited positions so it never repeats work. For Turn 3 it models exactly which stock cards can be reached as the three-card groups shift between passes. The budget is 150,000 positions per deal, and the 1,000 deals took about 17 minutes.
• The 95% range comes from the Wilson score interval, a standard way to put error bars on a percentage.

The bot's moves were checked against the rules over 2,000 games (1,000 at each draw setting) before the main runs. We used an AI assistant, Claude from Anthropic, to help write and review the simulation code and to draft this page. The figures themselves come only from running that code, and the team checked them against published research.

By Admin, free-solitaire.nz. Study date: 18 September 2026.

Frequently asked questions

How many Turn 3 games can be won?

Published research puts it at about 81.9% for a player who could see every card. Our own solver proved at least 66.2%. A computer player limited to what you can see won 10.6%.

Why is Turn 3 so much harder than Turn 1?

Because only every third stock card is reachable on each pass, and some cards you need stay trapped. In our tests the no-peek win rate fell from 42.3% to 10.6%.

Is there a limit on how many times I can go through the stock?

Not on this site. You can redeal as often as you like.

Is Turn 3 the "real" Klondike?

Both are standard ways to play Klondike. Turn 3 is the harder of the two by a wide margin, so choose it when you want a challenge and Turn 1 when you want to win more often.

Where can I read more?

See our FreeCell page for a game where skill decides almost everything, the games page for every variant, and the blog for more of our findings.