Golf Solitaire: why 92.7% of our deals can be won
Golf Solitaire is the quickest game on the site: seven columns of face-up cards, one pile to play onto, and a simple rule — play a card one rank higher or lower. A game can be over in two minutes. But how often can it actually be won?
We tested every deal exhaustively. Across 10,000 random deals using the rules on this site, 92.7% can be won with perfect play. That is roughly double the figure you will read elsewhere, and the reason is one small rule: on our version, Kings and Aces connect.
Golf at a glance
| What we measured | Our version (K↔A wrap) | Classic rules (no wrap) |
|---|---|---|
| Deals winnable with perfect play | 92.7% (10,000 deals) | 44.0% (3,000 deals) |
| Deals proven impossible | 7.3% | 56.0% |
| Win rate of a simple “no peeking” bot | 22.4% | 3.2% |
| Deals where the first card gives no tableau play | 27.4% | 31.2% |
Our no-wrap result of 44.0% lines up with the 45.1% found in a peer-reviewed study by Blake and Gent of the University of St Andrews (Journal of Artificial Intelligence Research, 2026). That agreement gives us confidence that our solver works — and that the jump to 92.7% really is down to the wrap rule.
How to play Golf Solitaire
The layout
- Seven columns of five cards, all face up (35 cards).
- One foundation pile, which starts with a single card.
- A stock of 16 cards, dealt one at a time onto the foundation. There is no redeal.
The goal
Clear all 35 cards from the tableau onto the foundation pile.
How cards move
- Only the bottom card of each column can be played.
- It may go onto the foundation if it is one rank higher or lower than the top card, regardless of suit. A 7 plays on a 6 or an 8.
- On this site, Kings and Aces wrap around: a King plays on an Ace and an Ace plays on a King. (Classic Golf treats the King as a dead end.)
- When nothing plays, turn the next stock card onto the foundation.
You win when the tableau is clear — stock cards may be left over. In the solvable deals, our solver's first solution left about one stock card unused on average.
Why the K↔A wrap matters so much
In classic Golf, a King on the foundation stops a chain dead: nothing can go on it. With seven columns, four Kings and only 16 stock cards, that dead end costs you a precious stock card every time. Wrapping King to Ace removes the dead end, so chains can run on — J, Q, K, A, 2 — and far more deals become winnable. Our exhaustive solver found that the wrap turns a 44% game into a 92.7% game.
Strategy that follows from the numbers
1. Look for the longest chain, not the first move
Our simple bot (which cannot see the stock) did best when it played the card that opened the longest visible run of follow-up plays. You can do the same: before playing, trace how far each option lets you go. A 6 that leads to 5, 4, 5, 6, 7 is worth far more than a 6 that leads nowhere.
2. Use both directions
Up-and-down play means you can zig-zag: 8, 7, 8, 9. Chains that turn back on themselves are how you clear several columns in one go.
3. Save stock cards
You have only 16. Every card you turn is a card less for later. Only draw when you are truly stuck. Our results show how precious they are: in the solvable deals, perfect play usually needed all but about one of them.
4. Dig out the deep columns
All columns start at five cards, but you will clear some faster than others. Keep an eye on the columns you are neglecting — a buried card you need late on is the most common reason a Golf game fails.
5. Don't be put off by a slow start
In 27.4% of deals on our version, the opening foundation card has no play at all, so your first move is to turn the stock. That is normal, not a sign the deal is lost.
Common mistakes we see
- Playing the first card that fits. Several columns often offer a legal play. The wrong one can end a chain two cards early, while the right one clears half the board.
- Drawing too soon. Scan all seven bottom cards — and remember the King–Ace wrap — before you turn the stock. It is easy to miss a play that crosses from King to Ace.
- Ignoring the stubborn column. If one column is full of ranks far from the ones you keep playing, start working on it while you still have stock cards left to bridge the gap.
- Giving up at the first dead end. Undo is unlimited here. Step back a few plays and try the other branch; in Golf a different order often changes the whole game.
A good habit is to count before you commit: how many cards will this chain clear, and which card will be left on the foundation at the end of it? The best chain is not always the longest — it is the one that leaves the next chain open. Ending on a middle rank such as a 7 keeps both directions alive, while ending on a rank that is already mostly played out can strand you.
Perfect play versus real play
The 92.7% figure assumes perfect knowledge of the stock order, which no player has. Our “no peeking” greedy bot, which plays only what it can see, won 22.4%. A thoughtful human with Undo sits somewhere between the two — and with Undo on this site you can in effect peek at the stock, then step back and choose a better line. That is how dedicated players push their Golf win rate well above the bot's.
Playing Golf on free-solitaire.nz
Golf is free here, and there is nothing to download, install or sign up for. It runs in any modern browser, on phone, tablet or computer — perfect for a quick round on the bus into Auckland or a tea break.
- Large, clear cards and a zoom control.
- Hint, unlimited Undo, Restart and Auto Play.
- Tap or drag and drop to play a card.
- Wins, time and moves tracked for your current game.
- Sound off, fullscreen and table colour options.
No gambling, nothing to buy, and no timer pushing you.
How we did the research, step by step
The free-solitaire.nz team ran this study in September 2026. These are the steps we took, in order.
- Read the rules from our own game. From this site's game code: seven columns of five face-up cards, one card to start the foundation, a 16-card stock with no redeal, and cards playing one rank up or down. This site's only Golf version lets Kings and Aces wrap (K on A, A on K).
- Rebuilt the game in code. We wrote a program that plays Golf by exactly those rules, with and without the wrap.
- Generated numbered deals. Every deal comes from a fixed seed, so every figure can be reproduced.
- Ran an exhaustive solver. It tries every line of play and remembers dead positions, so every deal ends as winnable or impossible, with no “undecided” group. It knows the stock order, so it shows what perfect play could achieve.
- Measured the site's game. With the wrap, 92.7% of 10,000 deals are winnable (95% range 92.2–93.2%), and 7.3% are impossible.
- Checked the solver against published research. Without the wrap, the same solver found 44.0% of 3,000 deals winnable (42.2–45.8%). Blake and Gent, Journal of Artificial Intelligence Research, 2026, report 45.1% for that version. The two agree, which gives us confidence in the solver.
- Ran a player that cannot see ahead. A greedy bot that does not know the stock plays the tableau card with the longest visible follow-up chain and draws only when stuck. It won 22.4% with the wrap and 3.2% without.
- Inspected the opening deals. In 27.4% of deals the first foundation card has no tableau play at all (31.2% without the wrap).
What this means: most Golf deals on this site can be won, and the gap between 92.7% and the bot's 22.4% is almost all in the choices. Ranges are 95% confidence intervals.
The technical side
Golf is simulated in golf.py, written in Python 3.11 with only the standard library. A shared runner.py plays 19 deals at once across processor cores using Python's multiprocessing, and each deal is shuffled from a numbered seed.
- The solver is an exhaustive depth-first search. A Golf position is small (how many cards are left in each column, how far through the stock you are, and the card on the foundation), so the solver can try everything. It remembers every position that led nowhere and never tries it again. That makes it exact: every deal is either solved or proven impossible. It knows the stock order, which a player does not.
- The bot does not know the stock. It plays the tableau card with the longest visible follow-up chain and draws only when stuck.
- Scale. The 10,000 deals with the King–Ace wrap took about 12 minutes; the 3,000 without it about 7.
- The 95% range is a Wilson score interval.
Our best check is the no-wrap run: its result lines up with the published figure for the classic game, which tells us the solver is sound. An AI assistant (Claude, made by Anthropic) helped write and review the code and draft this page. It produced none of the figures: each comes from running the code, and the team checked them against published research.
Frequently asked questions
Can every Golf game be won?
No. Even with perfect play, 7.3% of deals on our version are impossible. Under classic rules without the wrap, more than half are.
Why is Golf so often called hard?
Most figures quoted online are for classic rules without the King–Ace wrap, where only about 45% of deals are winnable even with perfect knowledge. Our version is kinder.
How long does a game take?
Usually just a few minutes — there are only 51 cards to play.
Is Golf related to Pyramid or TriPeaks?
They share the idea of clearing cards to a single pile. If you enjoy Golf, try Pyramid Solitaire, where you pair cards that add up to 13.
Do I need an account?
No sign-up, no email, nothing to install.
What else can I play?
For a longer, deeper game try FreeCell or classic Klondike. The full list is on our games page, with more research on the blog.