Michael Paycer — SQL Server DBA and lifelong computer-chess tinkerer
Computer Chess · The Clone Wars

Code Provenance in Computer Chess

The most dramatic story in engine history isn't a game — it's a question of whose code is whose. Fruit into Rybka, Rybka into Strelka and the Ippolit family, Ippolit into Houdini, and finally Stockfish's own code turning up inside a €100 product. It's a lineage of borrowed brilliance, a lifetime ban, and a GPL lawsuit. And it happens to be exactly the kind of story a DBA who's inherited someone else's codebase feels in their bones.

Computer Chess Cluster
This is the Clone Wars page — part of Computer Chess, from a DBA's Chair. Companion reading: History · How Engines Were Built · Poke the Bear (where the unresolved parts of this story get argued out) · Glossary.
The King, Briefly

How far ahead Rybka was

From roughly 2005 to 2010, Rybka — written by Vasik Rajlich — didn't just lead the rating lists, it terrified them. It topped every list, won World Computer Chess Championships, and was so far ahead that it reshaped how human grandmasters prepared openings. For a while, "what does Rybka think?" was the last word in any analysis. That dominance is the setup for everything that follows, because a lead that large invites the question of how.

How to read a legacy engine's rating (the honest way)

You'll see Rybka quoted at a single Elo — "Rybka was 3186" — and you should distrust it. The honest treatment of any legacy engine is three parts: (1) what it did in its own era (Rybka: list-topping, terrifying, reshaped human prep) → (2) its legacy and controversy (the ICGA plagiarism ruling, titles stripped, the lifetime ban) → (3) why its modern number is soft (Rybka 4's "current" CCRL rating ranges from about 3057 to 3186 depending purely on which build, index, and snapshot you cite). Never lead with the number. The Disputes page has the receipts on that 130-point swing.

The Clean One

Fruit — the open engine that taught a generation

Two years before Rybka's reign, in 2004, Fabien Letouzey released Fruit: open source, strong, and — the crucial part — clean and readable. Its code became the template a whole generation of programmers studied. Fruit 2.2 took runner-up at the 2005 World Computer Chess Championship, behind the independent engine Zappa. Then, after the 2005 season, Letouzey closed Fruit and took it commercial. The dramatic irony writes itself: he closed his engine specifically to stop people cloning it in tournaments — right before the most famous cloning case in chess history broke.

The Reckoning · 2011

The ICGA investigation and ruling

In 2011 the International Computer Games Association convened a panel to investigate accusations that Rybka had derived from Fruit. Their finding cited substantial structural similarity to Fruit 2.1 (analyses circulated around the case put the overlap near 74%), plus evaluation and search similarities to Fruit and to Crafty. The ruling was severe: Rajlich was banned for life from ICGA events, and four World Computer Chess Championship titles (2007–2010) were vacated. Rajlich largely declined to mount a formal defense.

The counter-argument has never fully died, and it belongs on the record: critics of the verdict argue the "originality" standard was applied too strictly, that many of the cited ideas had already diffused across the whole field, and that the distinction between inspiration and line-by-line copying was never cleanly drawn. Even people who accept the ruling often think the ICGA process was messy. That unsettled residue is why this page links to Poke the Bear rather than pretending the matter is closed.

The Spine

The full provenance thread

The reason this is a page and not a paragraph is that the borrowing didn't stop with Rybka. Strength kept leaking from closed code into clones and open reimplementations, in a chain that runs all the way to a 2021 lawsuit:

Fruit2004 · open, clean RybkaICGA-banned 2011 Strelkaalleged RE clone Ippolit /RobboLito Houdiniacknowledged influence Stockfishopen, GPLv3 Fat Fritz 2 (2021)SF + swapped net GPL lawsuitsued 2021 · settled

Strength leaking from closed code into clones and reimplementations — until the borrowing hit an engine (Stockfish) whose licence had teeth.

Reading the chain in words: FruitRybka (ICGA-banned for deriving from it) → Strelka (2007, alleged reverse-engineered Rybka clone — Rajlich himself called the derivation "obvious," which is its own irony) → Ippolit / RobboLito / Igorrit (an anonymous ~2009 open-source family, widely believed Rybka-derived, legally murky) → Houdini (a strength leader of the early 2010s that openly acknowledged drawing on Ippolit/RobboLito ideas) → and finally Fat Fritz 2 (2021), which ChessBase sold for around €100 as a breakthrough. The Stockfish and Leela teams showed it was essentially Stockfish with a swapped neural network — a violation of the GPLv3, which lets you sell derived work but requires you to ship the licence and the corresponding source. Stockfish sued ChessBase in July 2021, and the case settled: ChessBase may keep distributing Stockfish-based products only with proper GPL compliance and disclosure.

The Centerpiece

Why this is a DBA's kind of story

The DBA bridge · inherited-code governance

This whole saga is about code provenance, derivative works, licensing, and attribution — the exact governance a DBA applies to an inherited schema or a third-party component. What is actually "your own IP" when you've built on someone else's codebase? What must you disclose? The GPL answered it bluntly for Fat Fritz 2: you can sell derived work, but you must disclose and share the source. That's the same discipline as tracking which stored procedures came from a vendor, which objects carry a licence obligation, and what you're allowed to ship. And the punchline is the one that matters most: open, auditable, licence-clean code (Stockfish) outlasted every brilliant closed engine. The black boxes were dazzling and are mostly gone; the auditable system is still on top. Any DBA who has chosen a boring, inspectable component over a magical opaque one knows exactly why.

Sources & Further Reading
  • ICGA investigation and ruling (2011) — the panel report; Chessprogramming Wiki, "Rybka Controversy."
  • Fruit timeline and Letouzey — fruitchess.com history; Chessprogramming Wiki (Fruit); Wikipedia (Fruit). Fruit 2.2 was runner-up at WCCC 2005 behind Zappa.
  • Rybka chronology and Strelka — Chessprogramming Wiki (Rybka, Strelka); Wikipedia (Rybka).
  • Fat Fritz 2 / GPL — Chess.com coverage; Stockfish's lawsuit post; the settlement.
  • Komodo/Kaufman cross-links (Kaufman also worked on Rybka 3) — Wikipedia (Komodo).
Keep Exploring the Cluster

The parts that were never settled

The ICGA process, Rybka's "true" originality, and the clone-of-a-clone recursion are still argued about. That's a page of its own.

Poke the Bear →  ·  The full History →  ·  Back to the Hub →