Source Authorship Forensics · CCRL 40/40
The benchmark the entire series measures against, on both axes.
Stockfish is not a borderline case on any axis; it is the fixed point the rest of the series is measured against. It is developed entirely in the open by hundreds of human contributors, every change gated by SPRT testing on the distributed Fishtest framework, over roughly fifteen years.
It also defined the modern engine: efficient bitboard search, the standard pruning/ordering stack, and — since 2020 — NNUE evaluation. When other engines here rate “Strong” on Stockfish-similarity, this is what they resemble.
Confidence: Absolute — the reference itself.
Every page in this series implicitly points here.
| The human axis |
|---|
| Hundreds of named contributors over ~15 years |
| Every patch SPRT-tested on distributed Fishtest hardware |
| Fully open GPLv3 development, in public, commit by commit |
| No single author — a genuine community |
| The Stockfish axis |
|---|
| Originated the modern bitboard + NNUE + LMR/NMP/SEE recipe |
| Its network architecture is widely studied and imitated |
| “Derived” engines borrow its code or nets directly |
| “Strong” engines re-implement its ideas independently |
Stockfish is the human-built, community-developed reference at the top of CCRL — and the origin of the paradigm nearly every strong engine now follows. It is definitionally human and definitionally the Stockfish end of the similarity scale. Everything else in this series is, in a sense, measured by its distance from here.