BM25 (Saturation and Length)

Advanced ranking algorithm accounting for term saturation (successive word repetitions have diminishing effect: like salt in soup, 2 pinches good, 20 pinches inedible) and document length penalties (longer text isn't automatically better; what matters is density of rare, specialized words). BM25 is the standard in lexical retrieval used by Google.

Also appears as a content extraction method in Crawl4AI — text blocks compared with H1 using BM25 method. BM25 explains why keyword stuffing doesn't work: after reaching saturation threshold, additional keyword repetitions don't help and may hurt.

In practice, instead of repeating the same phrase 20 times, use it 3-5 times and fill the rest with specialized terms having high IDF. This gives better signal than brute-force repetitions.