Scorpion Solitaire: the game where even perfect planning often runs out

Most games on this site tell the same story: think further ahead and you win far more often. Scorpion is the exception that proves how far that rule stretches. It runs on our Spider engine with one deck, builds only in suit, fills gaps only with Kings, and lets you move any face-up card with everything piled on top of it.

We gave Scorpion the strongest planner we have: a solver that knows every face-down card and the three stock cards before it moves. Across 3,000 deals on the rules used on spidersolitaireonline.us, it won 12.9% (95% range 11.7–14.1%) and proved 67.5% (65.8–69.2%) impossible. The other 19.6% (18.2–21.1%) stayed undecided. So even flawless play with full information can win at most about one deal in three.

We did not build a simple, non-planning Scorpion player, so we cannot show a greedy-versus-planning gap. What we can show is where planning hits a wall, and which parts of the deal matter less than you might think.

At a glance: the limits of strategy in Scorpion

The last two rows are a useful surprise. Hidden Kings look like a disaster, since only a King can fill a gap. Yet the solver did no worse with two or more of them face down (14.0%) than with none (12.1%); the ranges overlap. The layout of the Kings is not what decides Scorpion. What decides it is whether some card is locked where no sequence of moves can free it, and in two deals out of three that lock is already in place when the cards are dealt.

For contrast, Simple Simon, on the same engine, was solved in 86.5% of deals. There, planning wins most games. Here, it mostly tells you sooner that a game is lost.

The rules as played on spidersolitaireonline.us

• Deck: one 52-card deck.
• Tableau: seven columns of seven cards. In each of the first four columns the three back cards are face down, twelve hidden cards in all. Everything else is face up.
• Stock: the last three cards. One click deals one card onto each of the first three columns, whenever you choose. It happens once.
• Building: down by one rank in the same suit.
• Moving cards: pick up any face-up card and everything on top of it comes along, in order or not. The card you pick up must land on the next higher card of its own suit.
• Empty columns: Kings only, with whatever sits on them.
• Completing a suit: a King-to-Ace run in one suit is removed automatically. Remove all four to win.

Where planning still helps

Reveal the twelve hidden cards early

The solver's advantage is that it knows them. You can shrink that advantage by turning them over. Moves that reveal a card come first.

Look for locks before you commit

If a card sits above a lower card of its own suit in the same column, say the 5 of clubs above the 4 of clubs, the only way to split them is to move the upper card elsewhere. Spot these at the start and plan around them; they are often why a deal is lost.

Use the unordered move as your escape route

Moving a whole unsorted stack is Scorpion's most powerful move. Choose the stack whose move exposes the card you most need.

Time the stock deal

The three stock cards can unblock the first three columns, once. Deal them when those columns need help, not out of habit.

Do not judge a deal by its Kings

Hidden Kings barely moved the solver's win rate. Look for locked cards instead.

Know when to walk away

With about two deals in three proven impossible, a fresh deal is often the smartest move. Use Undo to test a promising line before you give up.

Playing here

Scorpion on spidersolitaireonline.us is free in any modern browser on computer, tablet or phone, with no account or download. Undo and Restart let you test lines and try again; Hint and Auto Play are there too. Zoom for larger cards, move by drag or tap, and check wins, time and moves in the counter. Sound, fullscreen and table color are in the settings.

How we ran the study

The spidersolitaireonline.us team ran this test in September 2026, on this site's own rules.

1. Read the rules from our Spider engine's Scorpion variant: seven columns of seven, three hidden cards at the back of each of the first four columns, a three-card stock dealt to the first three columns once at any time, same-suit building, any face-up card moving with its pile, Kings only in gaps, finished suits removed automatically.
2. Wrote the game in Python 3.11, standard library only, in spiderlike.py, shared with Simple Simon.
3. Dealt 3,000 numbered hands, seeds 1 to 3,000.
4. Ran a best-first solver that knows the hidden cards and the stock, with up to 200,000 positions per deal. No person has that knowledge, so its wins are an upper limit for real play.
5. Sorted each deal into solved, proven impossible or undecided, and recorded how many Kings started face down.
6. Double-checked "impossible" verdicts with a separate brute-force program written independently of the solver; it agreed on the deals we tested. Scorpion is not covered by Blake and Gent (2026), so there is no published figure to compare.
7. Added 95% ranges (Wilson score intervals).
8. Named the limits. One deal in five stayed undecided, so the true winnable share lies between 12.9% and 32.5%, and even that is only reachable with knowledge of hidden cards.

An AI assistant, Claude by Anthropic, helped write and review the solver and draft this page. Every figure comes from running the code, checked by the team before publishing.

Questions

Why do I lose Scorpion so often?

Because most deals cannot be won. Even seeing every card, our solver proved 67.5% impossible.

Does a moved pile have to be in order?

No. Only the card you pick up must fit; the cards on it can be in any order.

Is Scorpion related to Spider?

Closely: same engine, completed suits removed automatically. The unordered pile move comes from Yukon.

Where does planning pay off more?

Try Simple Simon or one-suit Spider. All games are on the games page, with more on the blog.

By Admin, spidersolitaireonline.us (Solitairen OOD). Research run September 2026.