Strong Malt 1.0

[Rebel]

Strong Malt is our second guest engine. Single Malt is a cooperate HCE chess engine written by Bart Weststrate and Hans Secelle rated about 2300 elo. I was asked by Jeroen Noomen if I could introduce NNUE programming to them.

....

https://rebel7775.wixsite.com/rebel/guest-engines

[Bart Weststrate]

Many thanks Ed. Quite a surprising result. Hans and myself are verry happy with the brand new Strong Malt.

[Gabor Szots]

Strong Malt is our second guest engine. Single Malt is a cooperate HCE chess engine written by Bart Weststrate and Hans Secelle rated about 2300 elo. I was asked by Jeroen Noomen if I could introduce NNUE programming to them.

....

https://rebel7775.wixsite.com/rebel/guest-engines

Hi Ed,

Reading the guest engine page I was surprised to learn that Rebel codebase is in fact that of CSTal. Will you please clarify from which Rebel version that is so? Just to make sure that CCRL family grouping is correct.

[Rebel]

Strong Malt is our second guest engine. Single Malt is a cooperate HCE chess engine written by Bart Weststrate and Hans Secelle rated about 2300 elo. I was asked by Jeroen Noomen if I could introduce NNUE programming to them.

....

https://rebel7775.wixsite.com/rebel/guest-engines

Hi Ed,

Reading the guest engine page I was surprised to learn that Rebel codebase is in fact that of CSTal. Will you please clarify from which Rebel version that is so? Just to make sure that CCRL family grouping is correct.

Search is CSTAL, evaluation and polyglot book is Rebel.

[Peter Berger]

Strong Malt is our second guest engine. Single Malt is a cooperate HCE chess engine written by Bart Weststrate and Hans Secelle rated about 2300 elo. I was asked by Jeroen Noomen if I could introduce NNUE programming to them.

....

https://rebel7775.wixsite.com/rebel/guest-engines

Hi Ed,

Reading the guest engine page I was surprised to learn that Rebel codebase is in fact that of CSTal. Will you please clarify from which Rebel version that is so? Just to make sure that CCRL family grouping is correct.

What is difficult about this?

Rebel 13 was the last completely independent version.

Rebel 14 was the maybe a bit controversial open source version partly based on Fruit evaluation. For later versions Ed has already given an explanation.

This just with a most basic Google search, so please blame Alphabet for potential errors.What is this family idea btw? This one Google can't answer.

[Peter Berger]

Out of sheer curiosity I pasted this into ChatGPT - the result looks pretty nice to me:

What you’ve dug up is basically touching a long-running—and slightly messy—lineage issue in computer chess. The confusion isn’t surprising.

1) What’s actually going on with Rebel vs CSTal vs Fruit
Rebel (up to version 13)
This is the original, fully independent codebase developed by Ed Schröder over decades.
→ Your statement is correct: Rebel 13 is the last “pure” Rebel.
Rebel 14+ (2022 onward)
This is where things change fundamentally:
It is explicitly based on Fruit 2.1 code (open source).
It replaces the evaluation with NNUE and adds modern components.
It is not a continuation of the old Rebel codebase in a strict genealogical sense.

So if someone says “Rebel codebase is actually CSTal,” they are either:

mixing up evaluation style / authorship influences, or
misunderstanding the post-14 reboot, which is Fruit-based, not CSTal-based.
2) Where CSTal likely comes in

CSTal (Chess System Tal) is tied to:

Chris Whittington
NNUE work and evaluation networks used in modern engines

Rebel 14+ mentions:

NNUE nets “in the famous Chess System Tal style”
→ That’s probably the connection people are seeing.

But that’s not the same as saying the engine “is CSTal”. It’s more like:

“Uses NNUE ideas / nets associated with CSTal-style work.”

3) Why this becomes “difficult”

Because you’re mixing three different notions of “family”:

Codebase lineage (strict)
Rebel ≤13 → independent
Rebel ≥14 → Fruit-derived
Evaluation lineage (NNUE / HCE ideas)
CSTal influences via NNUE nets
Authorship / collaboration lineage
Shared contributors (Noomen, Whittington, etc.)

People often blur these together, but in engine classification (like CCRL), they matter a lot.

4) About “family grouping” (your question)

This is not a formal concept Google can answer cleanly—it’s more of a community convention used in rating lists like CCRL:

Engines are grouped into “families” to avoid inflated diversity
(e.g., counting 20 Stockfish derivatives as independent engines)

Typical criteria:

Shared codebase (strongest criterion)
Direct derivation (forks)
Significant copied components (evaluation/search)

So the real question is:

Should Rebel 14+ be grouped with Rebel ≤13?

Strict answer:

No (code lineage) → Rebel 14 belongs to the Fruit family
Yes (branding/author) → still “Rebel” in a historical sense

That’s exactly why people argue about it.

5) Bottom line
Your summary is basically right:
Rebel 13 = last original Rebel
Rebel 14+ = Fruit-based reboot with NNUE
The “CSTal codebase” claim is incorrect or overstated
The real difficulty is defining what “family” even means:
code vs ideas vs authorship

My personal complaint here would be, that this AI thingy should really learn to be a bit more succinct

Peter

[Peter Berger]

and i should add, that ChatGPT competely misses the next major change, but it sounds nice, doesn*t it?

[smatovic]

Out of sheer curiosity I pasted this into ChatGPT - the result looks pretty nice to me:
[...]

CPW disagrees with Chatbot:

https://www.chessprogramming.org/Rebel#Rebel_with_NNUE

--
Srdja

[Peter Berger]

Out of sheer curiosity I pasted this into ChatGPT - the result looks pretty nice to me:
[...]

CPW disagrees with Chatbot:

https://www.chessprogramming.org/Rebel#Rebel_with_NNUE

--
Srdja

CPW really makes no sense at all. It is easy to understand the confusion of ChatGPT, but how come CPW is confused?

[smatovic]

Out of sheer curiosity I pasted this into ChatGPT - the result looks pretty nice to me:
[...]

CPW disagrees with Chatbot:

https://www.chessprogramming.org/Rebel#Rebel_with_NNUE

--
Srdja

CPW really makes no sense at all. It is easy to understand the confusion of ChatGPT, but how come CPW is confused?

I added the 15 and 16 parts, all covered by references IIRC, if I missed something, please correct me.

--
Srdja