Klondike Turn 3: the game where planning the stock matters most
Turn 3 keeps every Klondike rule but one: the stock deals three cards at a time, and only the top card of each group can be played. That single change turns the draw pile from a simple queue into a puzzle of its own, and it is where the gap between playing move by move and thinking ahead is widest of any Klondike game we tested.
A bot that sees only face-up cards, takes waste cards by rote and never undoes won 10.6% of 10,000 three-card deals (95% range 10.0% to 11.2%). A planning search on the same rules proved 66.2% of 1,000 deals winnable (95% range 63.2% to 69.1%). The bot is left at least 55 points short, and most of that shortfall is the stock: cards the bot never managed to bring to the top.
Turn 3 at a glance
| What we measured | Result | Deals | 95% range |
|---|---|---|---|
| No-plan bot, face-up cards only, no Undo | 10.6% won | 10,000 | 10.0–11.2% |
| Stock passes in the bot's wins | 3.8 passes | 1,063 wins | 3.7–3.9 |
| Planning search (knows hidden cards): proven winnable | 66.2% | 1,000 | 63.2–69.1% |
| Planning search: undecided within budget | 24.3% | 1,000 | 21.7–27.1% |
| Planning search: proven impossible | 9.5% | 1,000 | 7.8–11.5% |
The bot's figure is a floor for a human, not a typical result. The planning search knows the face-down cards, so it marks a ceiling. The practical question is how far up from the floor you can climb, and in Turn 3 the answer is mostly about the stock.
What the three-card draw does to your options
After the deal, 24 cards wait in the stock. Each click flips three onto the waste; you can see all three, but only the top one is live. If you take nothing from the waste, a full pass shows you only every third card — positions 3, 6, 9 and so on up to 24 — and the next pass shows exactly the same eight cards again. Sixteen cards stay locked away.
The only thing that changes this is you. Take one card from the waste and every card after it moves up one place for the next pass, so different cards land on top. A bot that plays waste cards only when its priority list says so shifts the stock by accident. A player who shifts it on purpose reaches cards the bot never sees.
Planning the stock, step by step
Track what sits under the top card
When a group of three shows a card you need in second or third place, note it. That card is one waste play away from being reachable next pass.
Make the shifting play deliberately
Look for any waste card you can use before the one you want. Playing it, even onto a modest spot in the tableau, pulls every later card forward by one and can put your target on top.
Test it with Undo
Take the waste card, run through the next pass and see what comes up. If the shift did not help, undo and try a different waste card. This is the look-ahead the bot never does.
Expect several passes
The bot's wins took 3.8 passes through the stock on average, against 2.3 at one-card draw. Three-card games are won by patient cycling, not by a single sweep.
Don't rush waste cards home
Every card you take from the waste changes the stock order for good. Send one to a foundation when that also helps your next pass, not just because it can go.
Late in the game, the stock opens up
As the stock shrinks, fewer cards hide inside each group. A game that looks stuck early often loosens once only a handful of cards remain in stock and waste.
A worked example
- A red 6 sits on three face-down cards. You need a black 7 to move it, and a red 8 elsewhere could hold that 7.
- On this pass you spot a black 7 as the tenth stock card, under two others in the fourth group.
- A red 9 is on top of the second group, the sixth card, and a black 10 in the tableau can take it. You play it.
- Next pass, the black 7 has moved up to ninth place, the top of the third group. You play it on the red 8, move the red 6 across and turn over a hidden card.
- If the shift had not worked, one tap of Undo would have put the 9 back.
The Turn 3 toolkit
Turn 3 is free here in any modern browser on computer, tablet or phone, with no installer and no account. Unlimited Undo is the tool that matters most for stock planning; you also get Hint, Restart, Auto Play, a zoom control, drag-and-drop or two-tap moves, and settings for sound, fullscreen and table color.
How the study was run
The spidersolitaireonline.us team measured Turn 3 in September 2026, on the rules this site's own game code uses.
- Checked the engine. This page and Solitaire Turn 3 run the same three-card game with unlimited passes.
- Rebuilt it in Python, letting only the top waste card be played, as on screen.
- Dealt numbered hands from seeds so every deal can be replayed exactly: 1 to 10,000 for the bot, 1 to 1,000 for the search.
- Used two players: a bot that sees only face-up cards, follows a priority list and never undoes; and a depth-first search that knows the hidden cards and works out exactly which stock cards it can reach under the three-card rule.
- Capped the search at 150,000 positions per deal; deals it could not finish count as undecided.
- Compared with published work. Blake and Gent (2026) estimate 81.945% of Turn 3 deals winnable with perfect knowledge, inside our range of 66.2% to 90.5%.
- Stated the limits. The bot never plans its waste plays, which is exactly the skill on this page, so a careful player should beat it.
The code uses only the Python standard library. Claude, an AI assistant made by Anthropic, helped us write and review the code and draft this text; every figure comes from running the code, not from the AI.
Turn 3 questions
Why do I see the same cards every pass?
Because nothing has been taken from the waste. Play any waste card and the groups shift.
Is every Turn 3 deal winnable?
No. Even knowing every card, our search proved 9.5% of deals impossible.
Does the stock ever get reshuffled?
No. Its order stays fixed; only your waste plays change how it is grouped.
Should I learn on Turn 1 first?
It helps. The Klondike page covers the one-card game, and the full Turn 3 numbers are on Solitaire Turn 3. More games are on the games page.
By Admin, spidersolitaireonline.us (Solitairen OOD). Research run September 2026.

