Golf Solitaire: we matched the published result, then found a rule that doubles it
Golf is a quick, tidy patience. Thirty-five cards are dealt face up in seven columns of five, one card starts a single foundation pile, and sixteen cards wait in the stock. You clear the table by moving cards onto the foundation one rank up or one rank down, turning a stock card when you are stuck. Clear all seven columns and you win.
Golf is one of the games in the 2026 study by Blake and Gent at the University of St Andrews, which found that 45.109% of deals can be won with perfect knowledge of the stock. That gave us a perfect test. We built an exact solver for Golf, ran it under the classic rules, and got 44.0%, in line with theirs. Then we ran it under the rules this site actually uses, where a King and an Ace count as neighbours. The answer jumped to 92.7%.
Our figures against St Andrews
| Version | Our exact solver | Blake & Gent (2026) | Verdict |
|---|---|---|---|
| Classic: no wrapping between King and Ace | 44.0% winnable (95% range 42.2–45.8%, 3,000 deals) | 45.109% (±0.032%), from 10,000,000 deals | Agrees |
| This site: King and Ace wrap round | 92.7% winnable (95% range 92.2–93.2%, 10,000 deals) | Not studied | New figure |
Both figures assume you know the order of the stock, as the published study also does. Our solver checked every possibility for every deal, so there is no "undecided" category: each deal is either winnable or proven impossible. Of St Andrews' ten million deals, 4,510,859 were winnable and 5,489,141 were not.
What a player without foresight wins
| Player | With wrap (this site) | Without wrap |
|---|---|---|
| Exact solver, knows the stock order | 92.7% | 44.0% |
| Simple player, doesn't know the stock order | 22.4% (10,000 deals) | 3.2% (3,000 deals) |
| Deals where the first foundation card has no play on the table | 27.4% | 31.2% |
Nine out of ten deals here can be won with perfect play, but a player who simply grabs the most promising-looking card wins barely one in five. Golf rewards looking ahead far more than its simple rules suggest.
The rules on solitaire-free.uk
- Thirty-five cards laid out face up, five to a column across seven columns.
- One card is turned to start the foundation; 16 cards form the stock.
- Move the top card of any column onto the foundation if it is one rank higher or lower, regardless of suit.
- King and Ace wrap round: a King may go on an Ace and an Ace on a King.
- When no card fits, turn the next stock card onto the foundation. There is no redeal.
- Clear all 35 table cards to win; cards left in the stock don't matter.
Turning the data into better play
Look for the longest chain, not the first card
Our simple player picks the card with the longest visible follow-on chain, and even so wins only 22.4%. The solver's secret is looking further: which choice leaves the next several cards playable, including the ones under what you remove.
Treat the wrap as a tool
The King–Ace wrap is the rule that lifts winnability from 44.0% to 92.7%. Treat Kings and Aces as bridges. A run like 2, A, K, Q is legal here, and spotting those runs is often what clears the last column.
Save stock cards
Every stock card you turn is a fresh start for the foundation, and there are only 16. The solver's winning lines left about one stock card unused on average, so the margin is tight. Turn one only when the table truly offers nothing.
A dead first card is common
In 27.4% of deals the opening foundation card has no match on the table. That just means your first move is to turn a stock card; it is not a bad sign.
Undo is the solver's advantage, borrowed
The solver wins by trying everything. With Undo, you can test a run and step back. Golf games are short, so exploring costs little.
Golf here
Free to play, with no download, no account and no email address, in any current browser on computer, tablet or phone. Large, legible cards and zoom; Hint, unlimited Undo, Restart and Auto Play; drag and drop or tap to move; and wins, time and moves for the game in progress. Sound, fullscreen and table colour are adjustable.
Enjoy quick games? Pyramid is another fast, foundation-focused game. For a longer puzzle, try FreeCell.
Mistakes that turn a winnable Golf deal into a loss
- Playing the first card that fits. Often two or three cards fit. The right one is the one that leaves the longest run to follow.
- Forgetting the wrap. On this site a King goes on an Ace and an Ace on a King. Missing those links is the quickest way to throw away the rule that doubles your chances.
- Turning a stock card too soon. With only sixteen, each one should be turned only when nothing on the table plays.
- Ignoring the cards underneath. Every column is face up, so you can see what each play will uncover. Use that to plan two or three moves ahead.
How we ran the research, step by step
The solitaire-free.uk team ran this in September 2026. Golf was the ideal game to check our methods against published science, because it can be solved exactly.
- Rules from our code. Seven columns of five, 16-card stock with no redeal, plays one rank up or down, and the King–Ace wrap, which is the only Golf variant on this site. You win by clearing the table.
- A program version. We rebuilt Golf in Python with an exact solver and a simple player.
- Numbered deals. 10,000 deals with the wrap and 3,000 without, each shuffled from a numbered seed.
- Two players. The exact solver knows the stock order, as in the published study, and examines every possibility, so each deal ends as winnable or proven impossible. The simple player does not know the stock order; it picks the card with the longest visible chain and turns a stock card only when stuck.
- Running it. All 13,000 solver deals were fully decided.
- Comparing with St Andrews. Without the wrap, our 44.0% matches their 45.109% within our margin of error. That confirms the solver; the wrap result, 92.7%, is new.
- The margin of error. A 95% range shows where the true figure very probably lies: for 92.7%, between 92.2% and 93.2%.
- What the figures can't tell you. Knowing the stock order is something no player can do, so 92.7% is a ceiling. The simple player's 22.4% is a floor. Your own rate sits in between.
Behind the numbers: the code
Everything was written in Python 3.11 using only the standard library, with Golf in its own module, golf.py. The solver is an exhaustive depth-first search over the column heights, the position in the stock and the current foundation card; it remembers positions it has already found to be dead ends, so it never explores them twice, which is what lets it settle every deal. A shared runner spread the deals over 19 processes on a 20-core machine: about twelve minutes for the 10,000 wrap deals and about seven for the 3,000 without. Ranges use the Wilson score interval.
Claude, an AI assistant built by Anthropic, helped us write and review the solver and draft this page. It generated none of the results: every figure was produced by running the code, and our no-wrap result was checked against the published study before we went live.
Frequently asked questions
How many Golf deals can be won?
On this site, with the King–Ace wrap, 92.7% with perfect knowledge of the stock. Without the wrap, about 44–45%, which both we and the St Andrews study found.
Why does the wrap matter so much?
Without it, a King or an Ace on the foundation is a dead end. With it, every card always has two possible neighbours, so runs continue far more often.
Does suit matter in Golf?
No. Only rank counts: any 6 plays on any 5 or 7.
Why do I win less than 92.7%?
Because you do not know the order of the stock, and the solver does. A simple player without that knowledge won 22.4% in our study; planning ahead and Undo lift you above that.
Where can I find more?
The blog goes deeper into our research, and the games page lists all twelve games.
By Admin, solitaire-free.uk. Study date: 18 September 2026.