The Five Solitaires: What Actually Differs

I wrote the engines for all five of these games, and the thing that surprised me wasn't that they play differently. It's how measurably differently they play, from the very first card.

One Word, Five Different Jobs

"Solitaire" is one word covering five games that have almost nothing in common except a deck. I got here by accident. Spider Solitaire bored me years ago, the moves went automatic, and that boredom is why half of Cliko's card catalogue exists. When I sat down to write the engines for all five, I expected variations on a theme.

I found five sports.

What follows is the comparison I wish somebody had handed me. Every number in it comes from the code running on this site, not from a rules page, and where the code and the reputation disagree I've said so.

How Much the Deal Hides

The most useful way to sort these five is by how much of the deck you see on move one. It sounds like trivia. It decides everything.

FreeCell shows you all of it. Fifty-two cards across eight columns, four columns of seven and four of six, every single card face up before you touch anything. Hidden information: zero.

TriPeaks comes next. This build lays out thirty peak cards rather than the twenty-eight you'll read about elsewhere, and every one of them is face up from the deal. Twenty-two cards go to the stock, one flips immediately to the waste, and twenty-one stay hidden.

Pyramid puts twenty-eight cards on the felt, all visible, with twenty-four face down in the stock. Blocked cards are dimmed rather than turned over, so you can read the whole structure even when you can't touch most of it.

Then the gap opens. Klondike deals twenty-eight cards to seven columns and turns exactly seven of them face up, leaving twenty-one buried plus a twenty-four card stock. Forty-five of fifty-two cards are hidden when you start. Spider is worse: two decks, 104 cards, fifty-four dealt across ten columns with only the ten column tops showing, and a fifty-card stock. Ninety-four of 104 cards hidden.

Zero, 21, 24, 45, 94. That spread is the whole argument, and everything people say about which solitaire is "luckier" reduces to that one list.

Winnable Is Not the Same as Filtered

Here's a thing I should be honest about, because it's the claim I see repeated most often and it's half wrong. None of these five games checks whether the deal it just handed you can be won. There is no solver in the deal loop. Each game runs a plain Fisher-Yates shuffle over a seeded generator and puts the cards down exactly where they land.

That matters most for FreeCell, which has a reputation for being solvable by design. It isn't. Nearly every FreeCell deal is winnable, but that's a property of the game rather than a filter sitting in front of it: four free cells, eight columns and complete information leave so much slack that random deals almost always have a line through them. Of the 32,000 numbered deals from the old Microsoft version, exactly one has been proven unwinnable. One.

Klondike is the other pole. Nobody has computed the exact fraction of Klondike deals that a perfect player could win. It's a genuinely open problem, and the estimates that exist put draw-one somewhere in the low eighties as a percentage. So roughly one Klondike deal in five may be dead before you touch it, and the game will happily let you spend twenty minutes discovering that.

What Each Game Does When You Get Stuck

All five carry a rescue, which is a decision I still argue with myself about.

Klondike's is the most aggressive. When no move can gain ground, it sweeps every card still in the tableau, stock and waste into a pool, reshuffles, and re-deals a fresh layout while your foundations stay put. It's free when you're dead and costs 25 points when you call it voluntarily. FreeCell has the same shape. Spider's stays greyed out until nothing on the board gains ground, Pyramid's unlocks only when no two exposed cards sum to thirteen, and TriPeaks instead recycles drawn cards back into the deck against a per-difficulty budget: three at Beginner, none at Cyborg.

The purist's objection is fair. A deal that was unwinnable from the first card now ends in a re-deal rather than a wall, so the game quietly stops being the game it advertises. I kept it anyway. Losing to a shuffle you never had a chance against teaches you nothing.

What Each One Actually Trains

FreeCell trains lookahead, full stop. The number that governs the whole game is how many cards you can move as a unit, and it's exactly (free cells + 1) doubled once for every empty column. Park three cards carelessly and your maximum move collapses to two. The real difficulty dial isn't a difficulty menu, it's the mode: Classic gives you four cells, Challenge gives two, Expert gives one, and Expert is a different game.

Spider trains patience with space. An empty column is the only place a broken run can be taken apart and rebuilt in suit, and the suit count (one, two or four) changes the game more than any other setting on this site. Klondike trains move selection under uncertainty: which face-down card do you flip, given that a flip revealing a card with nowhere to go is worse than no flip at all. Pyramid trains something rarer, counting unlocks rather than counting sums. Any child can find two cards making thirteen. Knowing which thirteen frees the most structure is the game.

TriPeaks trains one skill only: scanning the whole board instead of the part you were just looking at. The rank chain wraps, so an ace reaches both a king and a two, and the card that saves your run is almost always sitting on the peak you stopped checking.

One honest note while I'm being precise about difficulty. Pyramid's seven levels move hints and undo counts and nothing else. The stock recycles without limit at every level, Cyborg included. The rules card inside the game still describes a per-difficulty pass limit, which is a leftover from an earlier build and a fix I owe.

Where You Should Start

If you want a game where losing is your fault, start with FreeCell. Nothing is hidden, nothing is random after the deal, and the distance between five cards of lookahead and twenty-five is learnable. It's the purest game on this site and I'd defend that against anyone.

If you want a game to think inside for twenty minutes, take Spider at two suits. One suit is a warm-up and four is an argument with probability, but two is where it gets honest. If you want speed, TriPeaks, which is built so that nothing at all sits between your finger and the next card. If you want the game your parents played, Klondike is right there being the friendly liar it has always been.

And if none of the five grabs you, follow the instinct that sent me looking past Spider. Mahjong Solitaire scratches the same clear-the-board itch on completely different geometry, and the rest of the card catalogue came out of that search. I also wrote about why most decks have 52 cards, and everything above is downstream of that one number.