Spider Solitaire, four suits: the top of the ladder
Four-suit Spider is the summit of this site. It is the full game: two decks, every suit in play, 104 cards across ten columns, and eight King-to-Ace runs to complete, each in a single suit. Of all twelve games we tested, it is the one our computer players found hardest by far.
Our Spider program, which sees only face-up cards and never takes a move back, won 4 of 2,000 four-suit deals, which is 0.2%. And yet published research finds that about 98.5% of four-suit deals can be won by a solver that knows where every card is. Those two numbers together tell the whole story of this game: the cards are almost always on your side, and the difficulty is entirely in the planning.
This page shows how four suits compare with everything else on the ladder, what the rules are here, how people actually climb this rung, and exactly how we produced the numbers.
The ladder from the top
| Game | Our computer player won | Runs completed per game | Deals |
|---|---|---|---|
| Spider, 1 suit | 52.8% | 4.96 of 8 | 4,000 |
| Spider, 2 suits | 5.7% | 0.83 of 8 | 4,000 |
| Yukon | 4.5% | – | 10,000 |
| Alaska | 1.5% | – | 10,000 |
| Spider, 4 suits (this page) | 0.2% (0.08–0.5%) | 0.05 of 8 | 2,000 |
Only 0.2% of our player's games got as far as four completed runs. On average it finished one run in every twenty games. Every row uses a different program except the three Spider rows, which use the same one, so the Spider curve is the most reliable part of this table.
Three numbers that belong together
- 98.5% (±1.5%): four-suit deals a solver seeing every card could win, from Blake and Gent, Journal of Artificial Intelligence Research, 2026.
- Up to 11%: the range solitairex.io's homepage gives for its players in its "expert-level" group, which includes four-suit Spider. It does not describe its method.
- 0.2%: our simple, no-undo program.
Perfect knowledge, real people and a basic program land at three very different points. The gap from 0.2% to 98.5% is the room that skill, patience and Undo can fill.
Four-suit rules on spidersolitaire.us
- Cards: two full decks, all four suits, 104 cards.
- Layout: ten columns, the first four with six cards and the rest with five. Top card face up.
- Stock: 50 cards, dealt ten at a time, one on each column. Five deals and no reuse.
- Building: any card may go on a card one rank higher, of any suit.
- Moving runs: only a run of one suit moves as a group.
- Empty columns: accept any card or movable run.
- Completing runs: a full King-to-Ace run of one suit is removed automatically. Eight runs wins.
If you try to deal while a column is empty, the game warns you.
How people climb this rung
Accept that it is a long game
Our player's rare wins took about 209 moves each, but with only 4 wins that average is shaky. What it does show is that four-suit games run much longer than one-suit wins (117.5 moves). Settle in.
Empty columns are worth almost anything
Our player cleared a column before its first deal in only 6 of 2,000 games and won 2 of those. That is far too few to measure precisely, but with one and two suits an early empty column went with a much higher win rate (51 of 52 and 15 of 21 games). With four suits it is also the rarest.
Build in suit, and when you cannot, build high
Every off-suit build creates a run you cannot move. If you must make one, do it with high cards that will not block the low cards you need soon.
Every face-down card is a question mark
A face-down card is an unknown. The more of them you turn, the better your plans get. Our program is rewarded for uncovering cards; so should you be.
Treat Undo as part of the game
The 98.5% figure comes from a solver that sees everything. You cannot, but unlimited Undo lets you find out what lies beneath a pile and change your mind. Our program never used it, which is likely part of why it won so rarely.
Features on spidersolitaire.us
Four-suit Spider is free to play in any modern browser on computer, tablet or phone. No download, no account, no sign-up, no email. You get large clear cards with a zoom control, Hint, unlimited Undo, Restart, Auto Play to finish a won game, and drag-and-drop or tap-to-move. Wins, time and moves are tracked for the current game; sound can be turned off; fullscreen and several table colors are available.
Behind the numbers: how we ran the study
In September 2026 the spidersolitaire.us team measured how much harder each extra suit makes Spider on this site. Four suits was the last step of that ladder.
- Started from the code that runs our game. 104 cards in four suits, ten columns, a 50-card stock dealt in five rounds, any card on one rank higher, only single-suit runs move, anything into an empty column, finished runs removed automatically. We assumed, without checking in a browser, that you cannot deal while a column is empty.
- Rebuilt Spider in code. One program covers one, two and four suits.
- Dealt numbered hands. Deals 100,001 to 102,000, each shuffled from its seed so anyone can repeat the test. The player's scoring was set earlier on deals 1 to 30, which are not part of these results.
- Chose the player. A program that looks a short way ahead using only the visible cards and never undoes. It is a very weak floor at four suits. We did not build a solver that sees every card, so our study gives no ceiling for this game; the published 98.5% fills that gap.
- Played 2,000 deals to the finish, half the number we used for one and two suits, because the result was already clear.
- Compared with published research. Blake and Gent's 98.5% is a different measure (every card known), so it cannot confirm or contradict our 0.2%. It tells you what is possible; ours tells you what a simple program manages.
- Calculated the margin of error. Four wins in 2,000 gives a 95% range of 0.08% to 0.5%.
- Stated the limits. With so few wins, the move count for won games and the early-empty-column figure are rough. The 0.2% says nothing about how often skilled players win.
The code behind the ladder
We wrote the simulation in Python 3.11 using only the standard library; Spider is the spider.py module. On each turn the program runs a best-first search through up to 250 positions reachable with the face-up cards, treating newly uncovered cards as unknown, and scores each position: plenty for same-suit links, a little for any link, minus points for every hidden card, plus points for empty columns and newly revealed cards. It deals from the stock only when nothing improves the score. A shared runner spread the 2,000 games over 19 worker processes on a 20-core machine, and the run finished in about three and a half minutes.
The ranges are Wilson score 95% intervals. Claude, Anthropic's AI assistant, helped us write and review the simulation code and draft this page; it did not produce any of the results. Every figure comes from running the code or from the cited sources, and the team checked them before publishing.
Ladder questions, answered with data
Can four-suit deals actually be beaten?
Almost always, in principle: 98.5% of deals with every card known, according to the published study. In practice it takes a lot of planning.
Why did your program do so badly?
It looks only a short distance ahead, cannot see face-down cards and never undoes. Four suits punish all three.
Should I start here?
Probably not. Start with one suit, then two suits. Each step teaches something the next one needs.
How is this different from FreeCell?
In FreeCell every card is face up from the start and our solver won over 99% of deals. Spider hides most of the cards and keeps dealing more.
Is there more detail on the four-suit study?
We write up the study in more detail on the blog, and every game is on the games page.
By Admin, spidersolitaire.us. Research run September 2026.