Baker's Solitaire: a low ceiling you can know for certain

Baker's Solitaire looks like FreeCell's twin: the same eight columns, the same four free cells, every card face up. Two rule changes make it a very different game. You build by suit rather than by alternating color, and, on this site, only a King may fill an empty column.

Those two rules put the ceiling far lower than most players expect. We ran a solver on 10,000 deals, and for once it reached a verdict on every single one. 31.8% can be won. 68.2% are provably impossible, no matter how well you play. There is no undecided middle this time; those are exact answers for those 10,000 deals.

So if Baker's feels harder than FreeCell, you are not imagining it. About two deals in three cannot be won at all. You can play it free here in any browser, with nothing to download.

Baker's Solitaire, measured

The published figure is from Blake & Gent's 2026 paper in the Journal of Artificial Intelligence Research. Our classic-rule result of 76.0% agrees with it, which tells us the solver is sound. The lower 31.8% is specific to the Kings-only rule used here.

One rule that more than halves the ceiling

In the classic version of Baker's Game, any card can go into an empty column, and three deals in four can be won. On this site, only a King may go there. That single change drops the winnable share from 76.0% to 31.8%.

Why so much? In FreeCell-style games, an empty column is your most flexible tool. You can park any awkward card there, or use it to shuffle a long run from one place to another. When only Kings may enter, an empty column is useless until a King is free to move, and in Baker's that often means the King is already stuck on top of cards you need. On this site an empty column also does not count toward moving several cards at once, so it cannot help you shift runs either.

The effect: many deals that would be comfortable under the classic rule become tightly locked. Knowing that helps. On most deals your job is simply to find out whether this is one of the winnable third.

The rules behind the low ceiling

• All 52 cards are dealt face up in eight columns: four of seven, four of six.
• Four free cells hold one card each. Four foundations build up by suit, Ace to King.
• Columns build down in the same suit: the 7 of hearts goes only on the 8 of hearts.
• Only a King may be placed in an empty column.
• Runs can move together, but only using empty free cells; empty columns do not add to how many cards move at once.

Getting the most out of a hard game

Locate the Kings before anything else

Since only Kings fill empty columns, their positions decide the game. A King lying at the bottom of a column is harmless. A King buried in the middle of a pile, on top of low cards, is a problem you need a plan for.

Count what's above each Ace

About half of our deals, 49.7%, had an Ace playable straight away. In the rest, count the cards sitting on each Ace and pick the cheapest one to free.

Protect your free cells even more than in FreeCell

With empty columns so limited, the free cells are your main working space. Fill one only when the move clearly advances you.

Don't empty a column without a King ready

An empty column that no King can reach is wasted. Clear it only when a King is free to drop in, ideally one that is blocking something important.

Accept the verdict and deal again

Two deals in three cannot be won under this site's rules. If you have tried a few lines with Undo and nothing opens, it is very possible there was nothing to find.

Baker's on the ceiling ladder

Baker's Solitaire has the lowest ceiling of any game we could measure precisely:

• FreeCell: 99.68% solved, 0 proven impossible.
• Golf: 92.7% winnable, exact.
• Baker's Solitaire: 31.8% winnable, exact.

It shares FreeCell's layout, so the gap between those two is almost entirely down to rules: same-suit building and the Kings-only empty column.

Inside the test, one step at a time

The solitaire-online.us team ran this study in September 2026 to find out exactly how many Baker's deals on this site can be won.

1. Reading the rules from our game code. The same-suit building, the Kings-only rule for empty columns and the fact that empty columns do not help multi-card moves all come from this site's game files. The classic literature version lets any card fill an empty column; ours does not.
2. Rebuilding the game. We added Baker's rules to the same program used for FreeCell, so both games share one tested engine.
3. Dealing the cards. Seeds 1 to 10,000 with this site's rules, and seeds 1 to 2,000 with the classic rule for comparison.
4. Choosing the player. A solver that searches for a full solution. Nothing is hidden in Baker's, so it sees exactly what you see.
5. Running it. Each deal had up to 200,000 positions. With these rules the search finished on every deal, so each one is either solved or proven impossible.
6. Checking against published research and a second program. With the classic rule we got 76.0%, against the published 75.053%. We also re-checked deals 1 to 5 with a separate brute-force program; the two agreed on every one.
7. Working out the margin of error. Even an exact answer for 10,000 deals is a sample of all possible shuffles. The 95% range, 30.9–32.8%, is where the true share for all deals would fall in 19 repeats out of 20.
8. What the numbers cannot tell you. A solvable deal still has to be found by a person, and some solutions are long. The site also lets a card come back off a foundation, which our solver did not use; that could only push the true share slightly higher.

The code behind the ceiling

The code is Python 3.11 with only the standard library. Baker's Solitaire runs on the FreeCell module, freecell.py, with a switch for the Baker's rules, and runner.py spreads the work across 19 processes on a 20-core machine. The independent check lives in verify_freecell_bruteforce.py.

The solver is a best-first search that always expands the most promising position, keeps a table of positions already seen, and only makes foundation moves that are provably safe. Because the search ran to completion on every deal, "impossible" here means the program tried every reachable position and none led to a win. The 10,000-deal run took about 16 minutes and the 2,000-deal classic-rule run about 3 minutes. For the percentage ranges we used the Wilson score interval.

Claude, an AI assistant made by Anthropic, helped us write and review the code and draft this page. It produced none of the figures on it: every result comes from running the code, and we checked the classic-rule result against the published study before trusting the rest.

Because each shuffle comes from a numbered seed, anyone with the same code gets exactly the same deals and the same verdicts.

By Admin, solitaire-online.us. Study date: September 18, 2026.

Ceiling questions from players

How many Baker's Solitaire games can be won?

With this site's rules, 31.8% of our 10,000 deals. The other 68.2% are provably impossible.

Why is it so much lower than the numbers I've seen elsewhere?

Most published figures, about 75%, use the classic rule that lets any card fill an empty column. This site only allows Kings there.

Is Baker's harder than FreeCell?

Much. FreeCell had no provably impossible deal in our test; Baker's had 6,816 of 10,000.

How many moves does a solved deal need?

The first solution our solver found averaged about 39 moves. That is not the shortest possible, and people usually take longer.

Which game has the same layout but more winnable deals?

Try FreeCell for the same layout with a far higher ceiling, or Golf Solitaire for a quick, winnable game. Everything else is on the games page.