TrySolitaire Card Decks — Noir & Four Color
===========================================

WHAT THIS COVERS

    assets/cards/Noir/*              (52 SVG sources + WebP renders)
    assets/cards/FourColor/*         (52 SVG sources + WebP renders)
    assets/cards/NoirMobile/*        (52 SVG sources + WebP renders)
    assets/cards/FourColorMobile/*   (52 SVG sources + WebP renders)
    assets/cardBacks/Obsidian.*
    assets/cardBacks/Emerald.*
    assets/cardBacks/IvoryDeco.*

These seven designs are original work by trysolitaire.com. Suit shapes,
court emblems, pip layouts, card backs and palettes were drawn from
scratch — they are not traced from, derived from, or recoloured from any
existing deck.

The two Mobile decks are the same two palettes in a second layout built
for small screens: one corner index instead of two, a rank about twice
the height (3.9x the ink area), and a single large centre glyph in place
of the pip grid.

    Licensed under Creative Commons Attribution 4.0 International
    (CC BY 4.0)

    https://creativecommons.org/licenses/by/4.0/
    https://creativecommons.org/licenses/by/4.0/legalcode

You may use, modify, remix and redistribute these decks, including in
commercial products, provided you give appropriate credit. Something
like this is enough:

    Card artwork by trysolitaire.com (CC BY 4.0)
    https://trysolitaire.com

The generator that draws them, scripts/generate-card-art.js, is code
rather than artwork and is covered by the repository's MIT license.
Running it reproduces every file above byte-for-byte.


WHAT THIS DOES NOT COVER

Every other deck in this directory came from an outside source under its
own terms, and nothing here changes or relicenses them:

    Classic/           Vectorized Playing Cards 3.2 — LGPL 3.0
    FrenchClassic/     SVG-cards by David Bellot — LGPL 2.1
    Plain/, Single/    Vectorized Playing Cards derivatives — LGPL 3.0
    ornamental/        OpenClipart — CC0
    Simple/, LargeSimple/, MobileSimple/,
    NotoSerif/, SerifFaces/, TimesNewRoman/

    Card backs other than the three listed above — OpenClipart, CC0

See src/pages/credits.html for the full attribution chain. If you want a
freely-licensed deck without copyleft obligations, Noir and Four Color
are the two in this repository that qualify.


A NOTE ON TYPE

The corner indices (A, 2-10, J, Q, K) are outlines traced from DejaVu
Sans Bold and baked into the SVG sources as paths. Rendering type to
outlines inside artwork is use of a font, not redistribution of Font
Software, so no font-license obligation travels with these decks — you
do not need DejaVu to use, modify or redistribute them, and the SVGs
render identically on a machine that has never seen it.

DejaVu Sans is itself under the Bitstream Vera license, noted here for
provenance rather than obligation:

    https://dejavu-fonts.github.io/License.html


ACCESSIBILITY NOTE ON FOUR COLOR

Four Color exists so suit identity does not depend on telling red from
black. Spades are slate, hearts red, diamonds blue, clubs green — the
conventional four-colour scheme, but with the specific values solved for
rather than picked by eye. Green is not the problem; the wrong green is.

The four inks are optimised against CIEDE2000 under two dichromat models
(Vienot 1999 and Machado 2009), across normal vision, protanopia and
deuteranopia, subject to every ink clearing 4.5:1 on the card body.
Worst pair over those five conditions is dE00 21.9, under deuteranopia.
The floor is the number to quote, not the pair that produces it: two
independent implementations of this measurement agree on the floor to
within 0.3% (21.96 and 21.91) and disagree on which pair binds it
(diamonds against clubs under Vienot, hearts against clubs under
Machado), because the published LMS matrices for these models differ
between sources.

Contrast on the body spans a 2.39x band, from 4.50:1 (clubs) to 10.77:1
(spades). That spread is deliberate and cannot be flattened: dichromats
keep lightness, so the separation the hue loses under simulation comes
back as luminance. Equalising the four inks by weight would undo what
the deck is for.

The pipeline matters as much as the number, because reviewers using a
different one have arrived at three different floors for this same
palette. It is: sRGB -> linear light -> the model's matrix -> back to
sRGB -> CIELAB -> CIEDE2000. Applying the matrix and then reading its
linear-light output as though it were sRGB skips an encode step and
produces figures that are not comparable to these.

Tritanopia is not part of the objective, and it is the clearest case of
why the pipeline has to be stated: measurements of this palette under
tritanopia range from dE00 4.8 to 14.8 depending on whose matrices you
use — a factor of three, against 0.3% agreement for the two dichromacies
that actually drive the design. It affects on the order of 1 in 10,000
people. The solve is constrained not to make it worse than the palette
it replaced (6.1 by the same implementation used throughout this file,
diamonds against clubs), and suit shape carries identity in that case.

For scale: a plausible-looking hand-picked version of the same scheme
measures 9.2, which means hearts and clubs are effectively the same
colour to a protanope — the exact confusion this deck exists to avoid.

If you redistribute a modified palette, re-check it against a
colour-vision simulator, and use CIEDE2000 rather than CIE76. CIE76 is
badly over-permissive in saturated reds and yellows, which is where
these inks live: it scores a palette 26 that CIEDE2000 scores 11. The
four are solved as a set, so moving one hue by eye reliably breaks
another pair.