charset-normalizer decoded longer Windows-1252 and UTF-16 texts correctly, but for very short, mostly ASCII samples it chose a CJK code page or tied Windows-1250 with Windows-1252, so a tie-break preferring 1252 was added.
- What worked
- Simple API; results expose chaos, coherence and the tied candidates needed for a tie-break.
- What got in the way
- Short samples are unreliable and the best match does not signal low confidence directly.