Spider Solitaire, 4 suits: when the stock is the least of your problems
Most of solitaire-turn-3.com is about one question: how much does the way you draw cards change your odds? In Klondike, on our home page, it changes them a lot. A simple bot that turns one card at a time won 42.3% of 10,000 deals (95% range 41.3%–43.3%); turning three at a time, it won 10.6% (10.0%–11.2%).
Four-suit Spider is a different story. Its 50-card stock never becomes a waste pile you pick through. It is dealt five times, one card onto each of the ten columns, and never comes back. The draw is simple; what is hard is the table. Only same-suit runs move, and with all four suits in play, clean runs are scarce. Our look-ahead bot won 0.2% of 2,000 deals (0.08%–0.5%). That says far more about the bot than about the game: a solver that sees every card, reported by Blake and Gent (2026), wins about 98.5% (±1.5%).
At a glance: Spider with four suits
| Measure | Result | 95% range | Deals |
|---|---|---|---|
| Our bot (no peeking, no undo): games won | 0.2% | 0.08–0.5% | 2,000 |
| Suits completed per game (of 8) | 0.04 | 0.03–0.06 | 2,000 |
| Games with at least 4 suits completed | 0.2% | 0.08–0.5% | 2,000 |
| Moves in a won game (deals included) | 209.3 | 193.9–224.6 | 4 wins |
| Bot emptied a column before the first deal | 0.3% | 0.14–0.65% | 2,000 |
| Published: winnable with every card visible | 98.5% | ±1.5% | Blake & Gent 2026 |
The bot's figures are a very weak floor; its four wins make the moves figure little more than an anecdote. The published figure is a ceiling that no player who cannot see face-down cards can reach. Your real chances lie somewhere in between, and we have no measurement that narrows the gap.
Where the draw rule fits in
In Solitaire Draw 3 the pain is access: two of every three stock cards are stuck under a third, and our bot needed 3.81 passes through the stock per win (3.73–3.90) against 2.34 at Turn 1 (2.32–2.36). In four-suit Spider every dealt card is immediately usable, so access is not the problem. Coverage is. Each deal lays ten random cards on your columns, and with four suits most of them will not match the run underneath.
We did not simulate four-suit Spider with a different stock rule, so we cannot say how much the one-way, five-deal stock costs in winning chances. The honest reading of our data is only this: the bottleneck we could measure is the suits, not the stock.
Spider rules on this site
- Cards: 104, two of each of the four suits.
- Table: ten columns, four of 6 cards and six of 5, top card up.
- Stock: 50 cards in five deals of ten, no redeal.
- Empty columns: any card may fill one; no dealing while one is empty.
- Building: any card on one rank higher; same-suit runs move together.
- Win: complete all eight King-to-Ace suits.
Surviving four suits
Build in suit whenever you can
An off-suit build is a temporary parking spot, not progress. Undo it later if a same-suit home appears.
An empty column is worth more than a deal
With one free column you can take apart and re-sort mixed runs. Hold off dealing while you are close to clearing one.
Turn over face-down cards early
Forty-four cards start hidden. Each one revealed before a deal is a card the deal cannot bury further.
Use Undo generously
With a 98.5% published ceiling, most deals are winnable in principle. Rewinding and trying another line is the closest you can get to the solver's view.
Reading a floor and a ceiling together
This page shows the widest gap on the site between what a program without foresight achieves and what is possible. At the bottom, 0.2% of 2,000 deals for our bot. At the top, 98.5% (±1.5%) for a published solver that sees every card. Klondike Turn 3 has a gap too, but a much narrower one: our simple bot won 10.6% of 10,000 deals, while our own solver that sees every card solved 66.2% of 1,000 deals (63.2%–69.1%), proved 9.5% impossible (7.8%–11.5%) and left 24.3% undecided (21.7%–27.1%).
Why the difference? In Klondike, most of what a program cannot see is the order of the stock and the 21 face-down cards, and the three-card draw adds to the fog. In four-suit Spider, the program can see plenty; what it lacks is the patience to sort four suits through only ten columns without undoing. A human with Undo can explore lines the bot abandons, which is why we expect real players to land well above our floor. We have no human data for this game, though, so we cannot say where.
Playing on solitaire-turn-3.com
Four-suit Spider is free to play in any current browser on computer, tablet or phone; there is no account and nothing to download. Hint, unlimited Undo, Restart, Auto Play, zoom and fullscreen are all included, along with drag or tap moves, a win, time and move counter, and settings for sound and table color.
How the figures were made
The solitaire-turn-3.com team ran these tests in September 2026 on the Spider rules taken from this site's game code.
- Rebuilt in Python 3.11, standard library only.
- 2,000 numbered deals from seed 100001, apart from the seeds used to tune the bot.
- Bot: a 250-position look-ahead over visible cards, no knowledge of hidden cards or the stock.
- Published check: Blake and Gent (2026), 98.5% with a solver that sees every card. Not comparable with our bot.
- 95% ranges: Wilson score for percentages, normal intervals for averages.
An AI assistant, Claude by Anthropic, helped with the code and the draft of this page. It supplied no figures; every number is from the code or the linked paper, and the team checked it.
Four-suit questions
Is four-suit Spider nearly unwinnable?
No. The published solver figure is 98.5%. Our 0.2% only shows how badly a simple bot copes.
Is it harder than Klondike Turn 3?
For our bots, yes (0.2% vs 10.6%), but the bots are different programs, so treat that as a rough hint only.
Easier Spider?
Try two suits or one suit. All games are on the games page, and more reading is on the blog.
By Admin, solitaire-turn-3.com (Forty Media). Research run September 2026.