Solitaire Turn 3: the harder draw, measured
Turn 3 is Klondike with one rule changed: instead of turning a single card from the stock, you turn three and may only play the top one. Everything else is identical to the Turn 1 game on our home page. It is noticeably tougher.
How much tougher? That question has a proper scientific answer. Blake and Gent at the University of St Andrews calculated that, with every card known, 81.945% of Turn 3 deals can be won. We ran our own simulation of this site's Turn 3 game and compared. Our results agree with theirs on how many deals are winnable, and they add something the published paper does not measure: how much harder Turn 3 is for a player who cannot see the hidden cards.
The headline: our no-peeking computer player won 10.6% of 10,000 Turn 3 deals, against 42.3% at Turn 1. One rule, a fourfold drop.
Side by side: our study and St Andrews
| Measure | Our result | Published result (Blake & Gent, 2026) | Verdict |
|---|---|---|---|
| Winnable with every card visible | At least 66.2% proven; at most 90.5% (1,000 deals) | 81.945% (±0.084%) | Agrees: inside our range |
| Proven impossible | 9.5% proven, with 24.3% undecided | 180,472 of 1,000,000 deals, about 18.0% | Agrees: inside 9.5–33.8% |
| Won by a player who cannot see hidden cards | 10.6% (95% range 10.0–11.2%, 10,000 deals) | Not measured | New figure from our study |
| Same player at Turn 1 | 42.3% (10,000 deals) | Not measured | — |
The published study also shows how winnability falls as you turn more cards at a time, always assuming every card is known: 90.480% when turning one, 88.620% turning two, 81.945% turning three, 69.337% turning four and 53.434% turning five. Going from one card to three costs less than nine points of winnability for a player who can see everything. For our player, who cannot, it cost more than thirty points. Turn 3 is harder mostly because it hides options from you, not because it makes deals impossible.
The rules of Turn 3 here
- Seven columns of one to seven cards, bottom card face up, and a 24-card stock.
- Build down in alternating colours in the columns; build up by suit from Ace on the foundations.
- Any ordered face-up run may be moved as a group.
- Only a King, or a run starting with a King, may go into an empty column.
- Click the stock to turn three cards onto the waste. Only the top one is available; play it to reach the next.
- When the stock is empty, turn the waste back over and go again. There is no limit on passes.
- A card may be brought back down from a foundation when you need it.
Because the stock is dealt in threes, the same cards tend to come up in the same positions each pass. Playing one card from the waste shifts the whole sequence, and that shift is the key to the game.
Playing Turn 3 better, based on the numbers
Treat the waste as a timetable
With Turn 1, every stock card is eventually reachable. With Turn 3, only every third card is, until you play one and the rhythm changes. Before a second pass, note which cards you need and whether they are sitting in a reachable slot. Playing a single waste card early can bring a buried one within reach on the next pass.
Plan for several passes
Our player's winning Turn 3 games used 3.8 passes through the stock on average, against 2.3 at Turn 1. Winning at Turn 3 usually means patient cycling, not a quick finish.
Don't read too much into hidden Aces
Our player won 10.1% of deals where two or more Aces started face down and 12.3% where none did. The draw matters far more than where the Aces lie.
Use Undo as the solver would
The St Andrews solver wins 82% because it can see every card. You cannot, but unlimited Undo lets you explore: turn through the stock, see what comes up, and step back if a line goes nowhere. That is the closest a human gets to perfect information.
Do not panic at a thin start
Turn 3 punishes early commitment. If the opening offers little, cycle the stock once before making tableau moves that you cannot take back without Undo.
Playing on solitaire-free.uk
Turn 3 is free here with no download, account or sign-up, and plays in any up-to-date browser on computer, tablet or phone. You get big readable cards with zoom, Hint, unlimited Undo, Restart, Auto Play, drag-and-drop or tap-to-move, and a record of wins, time and moves for the game in hand. Sound, fullscreen and table colour are yours to set.
How we ran the research, step by step
This comparison was carried out by the solitaire-free.uk team in September 2026, using the Turn 3 rules from this site's own game code.
- Rules from the code. Three cards turned per click, unlimited passes, group moves of ordered runs, Kings only into empty columns, and one card allowed back off a foundation.
- A program version of the game. We rebuilt Turn 3 in Python, modelling exactly which stock cards can be reached as the three-card rhythm shifts. The bot's moves were checked against the rules across 1,000 Turn 3 games.
- Numbered deals. Deals 1 to 10,000, each shuffled from its own seed, so the same set can be dealt again. The first 1,000 went to the solver.
- Two players. The bot plays by fixed priorities using only visible cards and never undoes, which gives a floor. The solver sees every hidden card and the stock order, which gives a ceiling, and was allowed up to 150,000 positions per deal.
- Running it. 10,000 deals for the bot, 1,000 for the solver.
- Checking against St Andrews. The published 81.945% lies between our proven 66.2% and our upper bound of 90.5%, and the published impossible share of about 18% lies within ours. Agreement, with our range wider because our search was smaller.
- The margin of error. "95% range" means the true value very probably lies in that band. For our bot's 10.6%, the band is 10.0% to 11.2%.
- What the figures don't show. The bot is a floor for a thoughtful human, not a typical score. A quarter of solver deals were undecided, and undecided does not mean lost.
Behind the numbers: the code
Everything is Python 3.11 with no extra libraries. klondike.py handles both Turn 1 and Turn 3, and a shared runner spreads deals across 19 worker processes on a 20-core machine. For Turn 3 the bot may only touch the top of each three-card group, so its choices mirror yours. The solver is a depth-first search over every legal move, remembering positions it has already visited so it never searches the same position twice, and treating the stock as a sequence whose reachable cards change as you play from it. An exhausted search is a proof of impossibility, because the only shortcuts it takes are provably safe. The bot's run finished in seconds; the solver's took about seventeen minutes. Percentage ranges use the Wilson score interval.
We used Claude, an AI assistant from Anthropic, while writing and checking this code and drafting the page. The figures themselves come only from running the code, and each one was checked against the published research before this page went live.
Frequently asked questions
How many Turn 3 games can be won in theory?
About 82% with perfect knowledge of the cards, according to Blake and Gent's study of a million deals. Our smaller study agrees: between 66.2% and 90.5%.
So why do I win so few?
Because you cannot see the hidden cards. Our computer player, in the same position as you, won 10.6%. A careful human with Undo should do better, but Turn 3 is genuinely hard.
Is Turn 3 the "real" rule?
Both are traditional. Turn 1 is kinder; Turn 3 is the sterner test. Pick whichever you enjoy.
How many passes should I expect?
Our winning games averaged 3.8 passes through the stock, so four or so is normal.
What should I play if Turn 3 is too frustrating?
Try Turn 1, or Golf, where our solver found 92.7% of deals winnable. For an all-visible puzzle, FreeCell is the natural next step. Background reading lives on the blog.
By Admin, solitaire-free.uk. Study date: 18 September 2026.