Alaska Solitaire: reading a deal where you build both ways
Alaska takes Yukon's deal — seven columns, no stock, 31 of 52 cards face up — and changes how cards join together. Instead of alternating colours, you build in the same suit, one rank up or down. A 7♥ can go on the 8♥ or the 6♥. That two-way building changes what a "good" opening looks like.
We dealt 10,000 Alaska games under this site's rules. On average 1.6 Aces started face down. A solver that could see every card proved at least 75.6% of 500 deals winnable and only 7.2% impossible. So most Alaska deals can be won — if you read them well.
Our simple look-ahead player, which sees only visible cards and never undoes, won only 1.5%. That's a very low floor from a weak player, and it tells you more about how much planning Alaska demands than about the deals themselves.
An Alaska deal, by the numbers
| What we looked at | Result | Deals | How we got it |
|---|---|---|---|
| Cards face up after the deal | 31 of 52 | every deal | From the rules |
| Aces face down, on average | 1.6 of 4 | 10,000 | Inspecting the deal |
| Deals proven winnable with every card visible | 75.6% (95% range 71.6–79.2%) | 500 | Solver, 50,000-position limit |
| Deals proven impossible | 7.2% | 500 | Same solver |
| Deals the solver couldn't settle | 17.2% | 500 | Same solver |
| Look-ahead player's wins | 1.5% (1.3–1.8%) | 10,000 | Visible cards only, no Undo — a floor |
| Cards the player got home, on average | about 5 of 52 | 10,000 | Same player |
Put together: between 75.6% and 92.8% of Alaska deals can be won with perfect knowledge (75.6% proven, plus up to 17.2% undecided).
The rules of Alaska on klondike.nz
Dealing
Seven columns of 1, 6, 7, 8, 9, 10 and 11 cards. The first column is one face-up card; the others have their last five cards face up. No stock and no waste.
The aim
Get all four suits onto the foundations from Ace to King.
Moves
- Any face-up card can move, taking every card on top of it along.
- It must land on a card of the same suit, one rank higher or lower.
- Only a King (with its stack) can go into an empty column.
- Face-down cards turn over when uncovered.
The alternating-colour version of this deal is Yukon.
Reading an Alaska deal
Look for suit neighbours
Scan the face-up cards for pairs in the same suit one rank apart — 9♣ and 10♣, or 4♦ and 3♦. Each pair is a ready-made landing spot, and because building goes both ways, every face-up card has up to two possible partners.
Think in both directions
Building down and up in suit means a column can grow 8, 7, 8, 9 if that's what frees a card. Don't limit yourself to descending runs; use whichever direction digs deeper.
Don't over-read the Aces
Our player actually did slightly better on deals with two or more Aces face down (1.9%) than with one or none (1.1%). That's a surprising result from a weak player on small win counts, so we wouldn't build a strategy on it — but it's a reminder that Ace position is not the key to an Alaska deal.
Keep an escape route for Kings
As in Yukon, only Kings fill empty columns. A King at the top of a tall column blocks everything beneath it until a column is free.
Expect a long game when it works
When our player did win, it took about 111 moves. Alaska rewards a plan you stick to.
A first-minute checklist for Alaska
An Alaska deal shows 31 of 52 cards, so you can plan a good part of the game before you start. Here is a routine built around the two-way, same-suit building rule.
- List every same-suit neighbour pair. For each face-up card, check for the card one rank above and one below in its suit. Those are your possible landing spots.
- Find the columns with the most hidden cards. They're on the right, with up to six face-down cards each.
- Look for a group move that uncovers a card. Because any face-up card can move with its stack, a single move can clear a lot of cards off a hidden one.
- Use both directions. If building down leads nowhere, check whether building up frees the same card.
- Earmark a King for the first empty column. As in Yukon, nothing else can go there.
- Don't judge the deal by its Aces. Our results give no reason to think buried Aces make a deal worse.
With 75.6% to 92.8% of deals winnable in principle, Alaska is a game you can usually beat with patience — and Undo.
Three ways players misread an Alaska deal
- "It's just Yukon." The deal is the same, but same-suit building changes which cards can meet. A card that would be stuck in Yukon may have two homes in Alaska, and the other way round.
- "Only build downwards." Many players forget that Alaska lets you build up as well. Checking both directions often reveals a move that seemed not to exist.
- "A low win rate means the game is unfair." Our weak player's 1.5% says little about the deals. The solver proved at least 75.6% winnable.
An Alaska deal rewards patient reading more than almost any other game on this site.
Play Alaska on klondike.nz
Free in any modern browser, on computer, tablet or phone. No download, no account, no email. Large cards with zoom; Hint, unlimited Undo, Restart and Auto Play; drag-and-drop or tap to move; wins, time and moves for the current game; mute, fullscreen and table colours. Try it on the Interislander crossing, if the swell allows.
Where these numbers come from
The klondike.nz team ran this study in September 2026.
- Rules from our code. Same deal as Yukon, same-suit building up or down, group moves, Kings only in empty columns.
- A program version. The Yukon engine with Alaska's building rule.
- Numbered deals. Seeds 1 to 10,000 for the player and 1 to 500 for the solver.
- Looking at each deal. We counted face-down Aces.
- Two players. A look-ahead player on visible cards with no Undo, and a solver that sees every card with a limit of 50,000 positions per deal.
- Comparison. The published study we check against doesn't cover Alaska, so there's nothing to compare with.
- Error bars. 71.6–79.2% is the 95% range around 75.6%.
- Limits. The player is weak, and 17.2% of deals were left undecided by the solver.
Under the bonnet
Python 3.11 with the standard library only; the game logic is in yukon.py with the Alaska rule switched on. A shared runner uses multiprocessing to spread deals over 19 worker processes. The 10,000-game player run took about a minute, the 500-deal solver run about three and a half. The player searches 300 positions ahead over visible cards; the solver is a depth-first search with a 50,000-position budget and no automatic moves. We report Wilson score 95% intervals. The code and this page were written with help from Claude, an AI assistant by Anthropic. The AI didn't produce any figures: every number is the output of the code.
Questions players ask
What makes Alaska different from Yukon?
You build in the same suit, up or down, instead of down in alternating colours.
How many Alaska deals can be won?
At least 75.6% with every card visible, and up to 92.8% if all undecided deals turn out winnable.
Why did your computer player win so rarely?
It sees only visible cards, has no Undo and a short look-ahead. It's a floor, not a forecast.
Is Alaska easier than Yukon?
Our solver proved more Alaska deals winnable (75.6% against 67.0%), but it also proved more impossible (7.2% against 3.4%), so it's closer than it looks.
Where can I read about the team?
See our about page, or browse all games on the games page.
By Admin, klondike.nz. Study date: 18 September 2026.