Penalty score

The four rules

After data placement, the encoder applies each mask pattern in turn and scores the result:

Rule	Looks for	Penalty
N1	Runs of 5+ same-colour modules in a row or column	3 + (length − 5)
N2	Every 2 × 2 block of one colour	3 each
N3	The 1:1:3:1:1 dark/light ratio of a finder, with 4 light modules beside it	40 each
N4	Deviation of overall darkness from 50%	10 per 5% step

Lowest total wins. N3's weight of 40 exists because a fake finder pattern in the data region is the most expensive defect possible: a scanner could lock onto the false corner and fail to find the symbol at all. The chosen mask number is then recorded in the format information.

Why identical data can yield different codes

The rules sound exact but are implemented unevenly: several major encoders score N3 without checking the four light modules the standard requires next to the 1:1:3:1:1 run, and a one-point difference flips which mask wins. Both results decode identically, the score only tunes scannability. This is a principal reason two generators produce visually different codes from the same input, and why "my code looks different" is not a bug report. The full scoring walkthrough lives at mask selection penalty scores; the exact per-mask formulas are in the mask pattern reference.

FAQ

What is a good penalty score for a QR code?

There is no target number. The score only ranks the eight mask candidates for one symbol against each other; the encoder simply keeps the lowest and discards the value.

Does the penalty score affect whether my QR code scans?

Marginally. Masking prevents pathological patterns (big single-colour areas and false finders), but any of the eight masks produces a valid, decodable symbol. Contrast, size and quiet zone matter far more.