From 667267d9bba60036db965eb60ebbbc295ede8a5c Mon Sep 17 00:00:00 2001
From: Dorian Burton <120594826+dburton95@users.noreply.github.com>
Date: Wed, 19 Aug 2026 14:56:08 -0400
Subject: [PATCH] Fixes the greyscale cutoff for boulder.png

    A fixed cutoff (e.g. "<=200 is ink") only makes sense for sprites with a light background to split against; a mostly-opaque 16x16 icon has almost no pixel above that cutoff, so every such icon collapsed onto the same "all ink" hash and was flagged as a near-duplicate of anything else that also collapsed -- which was most of them, boulder.png included. Thresholding against the image's own mean keeps the split meaningful (and roughly balanced) no matter how light or dark the source is.
---
 tools/modkit.py | 19 +++++++++++++++----
 1 file changed, 15 insertions(+), 4 deletions(-)

diff --git a/tools/modkit.py b/tools/modkit.py
index 64b212d5..99d98089 100644
--- a/tools/modkit.py
+++ b/tools/modkit.py
@@ -988,14 +988,25 @@ def cmd_add_release_workflow(args, repo):
 # ---------------------------------------------------------------- lint
 
 def ahash(image):
-    """Ink-mask hash over the 8x8 downscale: background (the lightest GB
-    shade) vs ink.  Swapping the three ink shades -- the classic recolor --
-    leaves the mask intact, which is exactly what MK302 wants to catch."""
+    """Ink-mask hash over the 8x8 downscale: background vs ink, split at
+    THIS image's own average brightness rather than a fixed shade.  Swapping
+    the three ink shades -- the classic recolor -- leaves the mask intact,
+    which is exactly what MK302 wants to catch.
+
+    A fixed cutoff (e.g. "<=200 is ink") only makes sense for sprites with a
+    light background to split against; a mostly-opaque 16x16 icon has almost
+    no pixel above that cutoff, so every such icon collapsed onto the same
+    "all ink" hash and was flagged as a near-duplicate of anything else that
+    also collapsed -- which was most of them, boulder.png included.
+    Thresholding against the image's own mean keeps the split meaningful
+    (and roughly balanced) no matter how light or dark the source is."""
     from PIL import Image
     small = image.convert("L").resize((8, 8), Image.LANCZOS)
     raw = (small.get_flattened_data() if hasattr(small, "get_flattened_data")
            else small.getdata())
-    return sum((1 << i) for i, p in enumerate(raw) if p <= 200)
+    raw = list(raw)
+    average = sum(raw) / len(raw)
+    return sum((1 << i) for i, p in enumerate(raw) if p <= average)
 
 
 def hamming(a, b):