Golf Solitaire: the game where planning pays off most
If you want one game to show what thinking ahead is worth, Golf is it. Every one of the 35 table cards is face up from the start, dealt in seven columns of five. You play the bottom card of any column onto a single foundation pile, one rank up or down, and turn a card from the 16-card stock when nothing fits. Clear all seven columns and you win.
We played the same deals two ways. A greedy program that grabs the best-looking play it can see won 22.4% of 10,000 deals (95% range 21.6–23.2%). An exhaustive solver that plans every line, and also knows the stock order, found a win in 92.7% of the same 10,000 deals (92.2–93.2%). That gap of roughly 70 points is the space a thoughtful player can work in. You will not see the stock like the solver does, but you can plan much further than the greedy program.
At a glance: greedy play vs full planning
| Rule set | Greedy program won | Solver found a win | Proven impossible | Deals |
|---|---|---|---|---|
| King–Ace wrap (this site) | 22.4% (21.6–23.2%) | 92.7% (92.2–93.2%) | 7.3% (6.8–7.8%) | 10,000 |
| No wrap (comparison only) | 3.2% (2.7–3.9%) | 44.0% (42.2–45.8%) | 56.0% (54.2–57.8%) | 3,000 |
The solver settled every single deal, so its figures are exact for these deals: each one is either won or proven lost. Because it knows the order of the stock, 92.7% is a ceiling. The greedy program is a floor. A real player lands somewhere between, and how far up depends almost entirely on planning.
Why one extra rule multiplies what planning can do
On spidersolitaireonline.us, Golf lets a King and an Ace connect in both directions. Without that wrap, a run dies at every King. With it, a chain can climb through the top and carry on from the Ace.
The wrap helps the planner and the greedy player very differently. The solver's winnable share goes from 44.0% to 92.7%, roughly double. The greedy program's win rate goes from 3.2% to 22.4%, about seven times as high. Our no-wrap 44.0% agrees with the 45.1% published by Blake and Gent for the same game, which gave us a useful check on the code.
Golf rules on spidersolitaireonline.us
- Table: 35 cards in seven face-up columns, five cards each.
- Foundation: one pile, begun with a single card.
- Stock: 16 cards, turned one at a time onto the foundation, with no second trip.
- Moves: the bottom card of a column may go on the foundation if it is one rank higher or lower. Suit and color are ignored.
- Wrap: Ace and King count as neighbors.
- Win: empty all seven columns. Unused stock cards do not matter.
How to play more like the planner
Count the chain before you take the first card
Our greedy program already picks the card that starts the longest visible chain, and it still won only 22.4%. The planner goes further: it asks which chain leaves the next chain open. Before a play, look at what the new column bottoms will be.
Treat the stock as a budget
In the deals the solver won, its first solution left on average only 1.08 stock cards unused (95% range 1.05–1.10, over 9,272 winning deals). There is almost no slack. Turning a stock card while a useful column play remains is the most common way to waste a winnable deal.
A blank start is normal
In 27.4% of deals (26.5–28.3%, 10,000 deals) the opening foundation card had no match on the table. That is not a bad deal; it is just the moment to turn a stock card.
Route runs through the wrap
A run heading toward a King can keep going into Aces and 2s. Planners look for these bridges; greedy play only finds them by accident.
Replay the same cards
With 92.7% of deals winnable, a loss usually means a better order exists. Restart deals the same cards again, and unlimited Undo lets you test a line and back out.
Playing Golf here
Golf runs free in any modern browser on a computer, tablet or phone. No download, account or email needed. You get large cards with zoom, Hint, unlimited Undo, Restart, Auto Play, and drag-and-drop or tap. The game tracks wins, time and moves; sound can be muted, and fullscreen and table colors are available.
How we measured it
The spidersolitaireonline.us team ran these tests in September 2026, using the rules of the game on this site.
- Read the rules from our game code: seven columns of five, one foundation, a 16-card stock with a single pass, plays one rank up or down, and the King–Ace wrap.
- Rebuilt Golf in Python, with a switch to turn the wrap off for comparison.
- Dealt numbered hands from fixed seeds: 1 to 10,000 with the wrap, 1 to 3,000 without, so every deal can be replayed exactly.
- Played each deal with two players. The greedy program sees only what you see and draws only when stuck. The exhaustive solver knows the stock order and tries every line.
- Checked against published work. Blake and Gent (Journal of Artificial Intelligence Research, 2026) report 45.1% for no-wrap Golf; ours is 44.0%, within the margin of error. They do not cover the wrap rule.
- Computed 95% ranges with the Wilson score method.
The code is Python 3.11 with only the standard library. The solver is a depth-first search that remembers every dead end so it never repeats one, which is why it could settle all 13,000 deals. We wrote and reviewed the code and drafted this page with help from an AI assistant, Claude, made by Anthropic. Every figure comes from running the code, and the team checked the results before publishing.
Questions about strategy in Golf
How much does thinking ahead really help?
On the same deals, greedy play won 22.4% and full planning found wins in 92.7%. No other quick game on this site shows a wider gap.
Is every deal winnable?
No. 7.3% are proven impossible even for a solver that knows the stock.
Can I play a King on an Ace?
Yes, here the ranks wrap both ways.
Where else does planning matter this much?
Pyramid shows a similar gap between greedy and planned play. For the Spider version of the story, see Spider Solitaire. All games are on the games page.
By Admin, spidersolitaireonline.us. Research run September 2026.