Solitaire Turn 3: one rule, four times harder
Turn 3 is classic Klondike with a single change: each click on the stock turns over three cards instead of one, and only the top card of the three can be played. Everything else stays the same. That one change is enough to move the game several rungs up our difficulty ladder.
Our computer player, which sees only the face-up cards and never undoes a move, won 10.6% of 10,000 Turn 3 deals. With one card at a time it won 42.3%. So the three-card draw cuts its success roughly fourfold, and leaves Turn 3 sitting just above two-suit Spider (5.7% for our Spider player) on the ladder. For a game that looks identical to the easy version, that is quite a jump.
This page shows where Turn 3 ranks, explains the rules on this site, gives advice rooted in what we measured, and walks through how the study was done.
Where Turn 3 ranks against Spider
| Game | Our computer player won | Deals played | 95% range |
|---|---|---|---|
| Spider, 1 suit | 52.8% | 4,000 | 51.2–54.3% |
| Solitaire, Turn 1 | 42.3% | 10,000 | 41.3–43.3% |
| Solitaire, Turn 3 (this page) | 10.6% | 10,000 | 10.0–11.2% |
| Spider, 2 suits | 5.7% | 4,000 | 5.0–6.5% |
| Spider, 4 suits | 0.2% | 2,000 | 0.08–0.5% |
The Turn 1 and Turn 3 rows use the very same player on the very same deal numbers, so the gap between them is a clean measure of what the draw rule does. Comparisons with Spider are rougher, because the Spider player is a different program.
Winnable, but less often
Our solver, which is allowed to see every hidden card, proved 66.2% of 1,000 Turn 3 deals winnable and 9.5% impossible, with 24.3% undecided when its search budget ran out. That places the true share of winnable deals somewhere between 66.2% and 90.5%. Blake and Gent's published figure for this rule set is 81.95%, inside our range. Compare the impossible share with Turn 1, where the solver proved just 2.1% of deals unwinnable: under Turn 3, more than four times as many deals are simply lost from the start.
Turn 3 rules on spidersolitaire.us
- Layout: seven columns of one to seven cards, top card face up; 24 cards in the stock; four foundations built by suit from Ace to King.
- The draw: each click turns three cards onto the waste. Only the top one is playable. When it goes, the one beneath becomes available.
- Columns: build down in alternating colors. Any face-up run can move together if its top card fits.
- Empty columns: only a King, or a run starting with a King, may fill one.
- Stock passes: unlimited. When the stock runs out, turn the waste back over and go again.
- Foundations: one card can come back down if you need it.
How to climb the Turn 3 rung
Know that the order of the stock changes
Under Turn 3, a card you take from the waste shifts which cards will be on top next time round. Taking one card early can put a different card within reach on the next pass. That is why Turn 3 rewards patience with the stock, and why our winning games went through it 3.8 times on average, against 2.3 at Turn 1.
Don't be put off by a poor-looking deal
We compared deals by how many Aces started hidden. Our player won 10.1% of deals with two or more Aces face down and 12.3% of deals with none face down. The difference is real but small: the draw rule matters far more than the opening layout.
Uncover first, sort later
Every hidden card you expose is a new option, and under Turn 3 options are scarce. Prefer moves that flip a face-down card, starting with the tallest column.
Lean on Undo while you learn
Our player never used Undo, and it still won about one deal in ten. You have unlimited Undo here. Use it to test which waste card to take, and you will do better than the program.
Treat it as practice for Spider
The skill Turn 3 trains, planning with limited access to the cards you need, is exactly what two-suit Spider asks of you. If you can win Turn 3 regularly, Spider with two suits is the natural next step.
Features on spidersolitaire.us
The game is free and runs in any modern browser on a computer, tablet or phone, with nothing to install and no sign-up, account or email address. Cards are large and clear, with a zoom control for a bigger board. There is a Hint button, unlimited Undo, Restart, Auto Play to finish a solved game, and drag-and-drop or tap-to-move controls. Wins, time and moves are tracked for the game in progress. You can turn the sound off, go fullscreen and pick a table color.
Behind the numbers: how we ran the study
The spidersolitaire.us team ran this research in September 2026 to see exactly how much harder the three-card draw makes Klondike on this site. These were the steps.
- Took the rules from our own game code. For Turn 3 that means three cards per click with only the top one in play, unlimited passes through the stock, Kings only in empty columns, and one card allowed back from a foundation.
- Rebuilt the game as a program. The same Klondike program serves both draw settings. We checked the player's moves against the rules over 1,000 Turn 3 test games (and 1,000 at Turn 1).
- Dealt numbered hands. Deals 1 to 10,000 for the player and 1 to 1,000 for the solver, each shuffled from its seed, so anyone repeating the test gets identical cards. They are the same deal numbers used for our Turn 1 test.
- Used two kinds of player. A player that sees only face-up cards and never undoes a move sets the floor. A solver that sees every card sets the ceiling.
- Ran every deal. The player finished all 10,000. The solver had 150,000 positions per deal before a deal was marked undecided.
- Compared with published research. Blake and Gent (Journal of Artificial Intelligence Research, 2026) give 81.95% for this game with every card known. Our range, 66.2% to 90.5%, includes it.
- Measured the margin of error. A "95% range" means the true value is almost certainly inside it. For our player at Turn 3, that is 10.0% to 11.2%.
- Stated the limits. The player is a floor, not a typical human; it never undoes. About a quarter of solver deals stayed undecided, which is why we quote a range.
The code behind the ladder
The code is Python 3.11 with nothing beyond the standard library, and Turn 3 runs through the same klondike.py module as Turn 1. For the draw, the solver works out precisely which stock cards can be reached from each position under the three-card rule, rather than treating the stock as open. A shared runner spread the work across 19 processes on a 20-core machine: about 8 seconds for the player's 10,000 deals and roughly 17 minutes for the solver's 1,000.
Percent ranges are Wilson score intervals. Claude, an AI assistant made by Anthropic, helped us write and review the code and draft this text; the figures themselves come only from running the program, and we checked them against the published study before publishing.
Ladder questions, answered with data
Why is Turn 3 so much harder when the layout is the same?
Because two thirds of the stock is out of reach at any moment. Cards you need can sit in the wrong position of a group of three pass after pass.
Is Turn 3 harder than Spider?
Harder than one-suit Spider, easier than two-suit Spider, going by our players: 10.6% here, 52.8% and 5.7% for Spider.
Can every Turn 3 deal be won?
No. Our solver proved 9.5% of deals impossible even with every card visible.
Is it cheating to use Hint?
Not at all. It is a single-player game, and Hint is there to help you learn the patterns.
Where do I go after Turn 3?
Go back to Turn 1 for a lighter game, or try Yukon, which has no stock at all. Each game we tested has its own page, listed on the games page.
By Admin, spidersolitaire.us. Research run September 2026.