Scorpion Solitaire: at least two deals in three cannot be won
Scorpion looks like a gentle game. All but nine cards are face up from the start, you can move any group of face-up cards, and there are only three cards in the stock. It is not gentle. Its ceiling is among the lowest of any game we have tested.
We gave a solver every advantage: it could see the twelve face-down cards and knew exactly which three cards the stock would deal. On 3,000 deals it solved only 12.9%. It proved 67.5% impossible, and left 19.6% undecided. So even with perfect information, at most 32.5% of Scorpion deals on this site can be won, and at least two in three are lost from the deal.
That is worth knowing before you start. Losing at Scorpion is normal. Winning one is a real achievement. Play it here for free, with nothing to install.
The ceiling at a glance
| Result | Value | Deals | How |
|---|---|---|---|
| Solved with every card visible | 12.9% (95% range 11.7–14.1%) | 3,000 | Solver that sees face-down cards and the stock (a ceiling) |
| Proven impossible | 67.5% (95% range 65.8–69.2%) | 3,000 | Same solver |
| Undecided at the search limit | 19.6% | 3,000 | Up to 200,000 positions per deal |
| Most deals that could be winnable | 32.5% | 3,000 | Solved plus undecided |
| Deals with at least one King face down | 65.2% | 3,000 | Deal inspection |
| Solved when 2+ Kings start face down | 14.0% | 645 | Same solver |
| Solved when no King starts face down | 12.1% | 1,045 | Same solver |
We found no published winnability figure for Scorpion to compare with.
Why so many deals are dead on arrival
Two rules make Scorpion so unforgiving on this site.
- Only Kings go into empty columns. You cannot use a space to park an awkward card, only to start a new King-led column.
- Cards must be built in suit. A 6 of spades will only take a 5 of spades. There is exactly one card in the deck that can go on any given card.
Put those together and a card can easily end up trapped below its only possible partner in the same column, with no way to separate them. The solver proves that happens in most deals.
You might expect buried Kings to be the main problem, since 65.2% of deals start with at least one King face down. Our results say it matters little: deals with two or more hidden Kings were solved 14.0% of the time, against 12.1% for deals with none.
The rules as this site plays them
Forty-nine cards in seven columns
- Seven columns of seven cards, which uses 49 cards.
- In the first four columns, the bottom three cards are face down. Everything else is face up.
- The last three cards form the stock.
Suit building and group moves
- Build down in the same suit.
- You can move any face-up card together with everything on top of it, in order or not.
- An empty column accepts only a King, carrying any cards above it.
The three reserve cards
At any time, click the stock to deal its three cards, one each onto the first three columns. There is only one stock deal.
Clearing the table
Build four complete suits, King down to Ace, in the columns. Each finished suit is removed automatically. Remove all four to win.
Playing a Scorpion deal as well as it allows
Free the twelve face-down cards
The twelve face-down cards in the first four columns hold most of the surprises. Free them early so you know what you are working with.
Think twice before you deal the stock
The three stock cards land on the first three columns and can bury exactly what you need. Deal when it helps, not by habit.
Create an empty column for a King
An empty column is only useful with a King to fill it. Line up a King-led group before you clear a column.
Watch for suit traps
If a card sits below its only possible partner in the same column, look for a way to split them early, before more cards pile on.
Two in three deals are already lost
Even the solver could win at most about one deal in three. If a game locks up, start another.
Spotting a dead deal early
Since at least two Scorpion deals in three cannot be won even with perfect information, it helps to recognize a hopeless one quickly instead of playing it for twenty minutes. These are the patterns to look for.
- A card above its own partner. If the 4 of diamonds sits somewhere above the 5 of diamonds in the same column, the 4 can only get onto the 5 after the 5 has been moved out from underneath. With Kings-only spaces, that is often impossible.
- Two suits locked together. If the 9 of spades is above the 10 of hearts in one column while the 9 of hearts is above the 10 of spades in another, each card needs the other column freed first. Scorpion has many knots like this.
- No King available for a space. An empty column does nothing until a King can fill it. If all the free Kings are already at the base of their columns, clearing a column may not help.
None of these guarantees a loss by itself, and the face-down cards can still surprise you. But if you see several at once, the deal is very likely one of the many that the solver proved lost. Starting a fresh deal is a reasonable choice, not a failure.
Inside the test, one step at a time
The solitaire-online.us team tested Scorpion in September 2026.
- Reading the rules from our game code. Seven columns of seven with three face-down cards in the first four, a three-card stock dealt one card to each of the first three columns at any time, in-suit building, any face-up group moving together, and Kings only in empty columns.
- Rebuilding the game. We wrote Scorpion as a Python program with those rules.
- Dealing the cards. 3,000 deals, each from a numbered seed.
- Choosing the player. A "thoughtful" solver that knows the face-down cards and the stock. Its result is a ceiling: the best any player could possibly do.
- Running it. Up to 200,000 positions per deal: solved, proven impossible or undecided.
- A second opinion. We re-checked some of the solver's "impossible" verdicts with a completely separate brute-force program. On the deals we tested, it agreed.
- Working out the margin of error. A 95% range means the true value would fall inside it in 19 of 20 repeats. Proven impossible: 65.8–69.2%.
- What the numbers cannot tell you. How often a person wins. Without seeing the hidden cards, real players will win less often than the solver.
The code behind the ceiling
The study is written in Python 3.11 using only the standard library. Scorpion shares the module spiderlike.py with Simple Simon, and a shared runner spreads the deals across processor cores with Python's multiprocessing. The separate cross-check is verify_scorpion_bruteforce.py.
The solver is a best-first search: it expands the most promising position first and keeps a record of positions already seen. All ranges on this page are 95% Wilson score intervals.
We had Claude, Anthropic's AI assistant, help write the solver, review it and draft this page. It did not produce any figures: every number comes from running the programs.
By Admin, solitaire-online.us (Solitairen OOD). Study date: September 19, 2026.
Ceiling questions from players
Is Scorpion the hardest game on the site?
It is one of the two hardest by ceiling. Only Baker's Solitaire, with this site's Kings-only rule, came close: 68.2% of its deals are proven impossible, against at least 67.5% for Scorpion.
Is it my fault when I lose?
Usually not. At least 67.5% of deals cannot be won at all.
Do the cards I move have to be in sequence?
Yes. Pick up any face-up card and the cards above it come along, sorted or not.
What is the best moment for the three-card deal?
Whenever it helps. You can deal it at any time, but only once.
What is a friendlier game?
Spider Solitaire also removes finished suits, and Yukon Solitaire also lets you move any face-up group.