The object of this solitaire game is to build a complete face-card
gallery of Jacks, Queens, and Kings.
You build sequences of cards on the
Foundation rows by suit (building rules: top row 2-5-8-Jack;
middle 3-6-9-Queen; bottom 4-7-10-King) using the cards of the
Tableau and the still incorrect cards of the Foundation. Deal cards from the Stock to the
Tableau as needed. At the end of the game, let the computer play out the same starting situation
a few times.

24 piles: 3 rows with 8 columns.
Empty spots can be filled with the base cards "2", "3", and "4", as above. Building rule: Add cards by suit, in increasing rank, by a difference of 3.
Cards at their correct position have an altered appearance (no number, plus a narrow frame).
8 piles are fed by the cards played from the Stock. Only completely visible cards are available for play.
The Stock initially consists of two decks (104 cards). As the game begins, the first 24 cards are played to the Foundation. Each time you click Stock, 8 cards move into play, one to each of the Tableau piles.
Ace Pile: Aces are automatically removed to the Aces pile and are not a part of play.
The game begins by dealing 24 cards, face up, to the Foundation and 8 cards to the Tableau. Some of the cards on the Foundation may now already be at their final position ("2", "3", "4"), and aces go directly to the ace pile. Just tap on a card to play it (it finds its correct position automatically).
You build a complete gallery of all face cards, removing all cards from the Tableau (score: 0)
The appearing photos are all from Blenio
Valley, Ticino, Switzerland
Or, you make a better score than the computer (see the "Statistics" tab). For your personal evaluation use whatever measure you like (better than the minimum or better than the median or the mean).
After the evaluation you start a New game or you Redo the same starting game.

"Jammed situations" are marked by a
blue bar.
Definitely jammed cards

During the evaluation process (simulation of 1'000 games) each individual result of the random process is given a small square. It's color depends on the result: blue: your result is better - white: drawn - red: computer is better - dark red: solved game (result 0); if your result was also 0: light gray.
Next to the random-tapping numbers, a second box labeled AI shows the score of a computer player using a Perfect Information Monte Carlo (PIMC) search: for every possible move, it briefly "peeks" at many random ways the hidden cards could be arranged, plays each one out to the end, and picks the move with the best average outcome. Like you, the AI gets exactly one attempt per deal - its score is never averaged over multiple tries, so it's a fair, single-shot comparison to your own game. Tap the AI box after finishing a game to step through its moves one by one and see exactly how it played your deal.

(in the lower right corner of the statistics graph). The small graph shows the same thing (at least the colors) as the summary table in the Statistics tab: the comparison indicators of the last games.
Export your game results in CSV format for further analysis.
Restore your game results from an older version exported with the button above.
Save your current settings for later re-import into a new version.
Load preferences saved from an earlier version.
Special thanks to Karen Morris for editing the texts, and to Marc Brühlmann for several of the photographs.
Vectorized Playing Cards 1.3, Copyright 2011, 2021 - Chris Aguilar - conjurenation at gmail dot com Licensed under: LGPL 3.0 - https://www.gnu.org/licenses/lgpl-3.0.html
Aerial photograph from 1923: ETH-Bibliothek Zürich, Bildarchiv/Stiftung Luftbild Schweiz / Photographer: Mittelholzer, Walter / LBS_MH01-003541
6.11.2026, 3.2.4: Solvability check with a known stock runs up to 45 s; tap orange to search twice as long. More unsolvable deals are detected. Cards that can never be played again are now marked. Dark mode fixes
24.10.2026, 3.2.3: Statistics show whether the deal is solvable with a known stock (green: a win was found, orange: no win found within 15 s, red: unsolvable), and the AI's degrees of freedom in bits
20.10.2026, 3.2.2: Stronger AI evaluation: within each phase between two stock deals, the AI now compares all reachable positions and picks the best one to deal from (often leaving well-parked cards on the tableau); once the stock is empty, it plays the rest perfectly
17.10.2026, 3.2.1: Stronger AI evaluation (PIMC): all candidate moves are compared on the same simulated stock orders, equivalent moves are evaluated only once
06.09.2026, 3.2: First version with a AI evaluation (PIMC) in the background
04.08.2026, 3.1.13: Fever curve now smooths the 'better or equal' trend in logit space, so gains and drops react more symmetrically near 0%/100%. Buttons
04.08.2026, 3.1.12: Fever curve now smooths the 'better or equal' trend in logit space, so gains and drops react more symmetrically near 0%/100%.
13.07.2026, 3.1.12: Fever curve now smooths the 'better or equal' trend in logit space, so gains and drops react more symmetrically near 0%/100%.
13.07.2026, 3.1.11: Removed the ELO indicator, focus on 'better or equal'. Added a color-coded 'better or equal' readout to the result panel.
17.02.2026, 3.1.8: Option for 4-color suits.
17.11.2025, 3.1: New indicator ("ELO"), new build and deployment procedures.
24.08.2025, 3.0.4: New evaluation graph. Import button for previously exported statistics
24.08.2025, 3.0.3: Removed a controversial indicator. Cleaned up the result panel
17.10.2024, 3.0.2: Fixed an error in the auto-move statistics introduced in 3.0.1
23.09.2024, 3.0.1: Added an indicator showing the risk of a stock-jam situation on the tableau after the next deal (small numbers display the probability of a new — or even final — stock jam)