How much of Stockfish code is still from Tord Romstad?

Discussion of chess software programming and technical issues.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
Jouni
Posts: 2010
Joined: Wed Mar 08, 2006 7:15 pm

How much of Stockfish code is still from Tord Romstad?

Post by Jouni » Mon Sep 16, 2013 8:57 pm

Just curious. Still most of code? Only skeleton? Something between these two?
Jouni

Modern Times
Posts: 2410
Joined: Thu Jun 07, 2012 9:02 pm

Re: How much of Stockfish code is still from Tord Romstad?

Post by Modern Times » Tue Sep 17, 2013 5:19 am

Very good question, I was wondering this myself, and whether our description of Stockfish needed updating:

http://computerchess.org.uk/ccrl/4040/c ... 4-bit_4CPU

"Stockfish is based on Glaurung by Tord Romstad, with additional programming by Marco Costalba and Joona Kiiski. "

Seems to me that is out of date now.

User avatar
Don
Posts: 5106
Joined: Tue Apr 29, 2008 2:27 pm

Re: How much of Stockfish code is still from Tord Romstad?

Post by Don » Tue Sep 17, 2013 1:37 pm

Jouni wrote:Just curious. Still most of code? Only skeleton? Something between these two?
If it's anything like Komodo, most of the code has been touched so in some sense probably very little of the code is Tord's original code. You could say that about Komodo too comparing it to Doch or even not too recent versions of Komodo.

However, that is not particularly relevant (although cloners think it's the only thing that matters.) Stockfish is still very much based on the original Glaurung - and has its DNA all over the place.

I have discovered that you cannot easily rewrite a program in place - you would have to scrap it completely and restart from scratch to get something original and even that would be most difficult as you would be tempted to cut and paste large sections or clone the algorithms and heuristic themselves.

Nevertheless, Stockfish has undergone an amazing make-over or transformation over the past few years due the remarkable talent of Marco and Joona.
Capital punishment would be more effective as a preventive measure if it were administered prior to the crime.

User avatar
velmarin
Posts: 1600
Joined: Mon Feb 21, 2011 8:48 am

Re: How much of Stockfish code is still from Tord Romstad?

Post by velmarin » Tue Sep 17, 2013 1:51 pm

It is very difficult to get out than usual.
Write something new without thinking about what's in your head.
Write something new, once you are in the learning. And if you work out, to resign.

But Stockfish and Glaurung, still seem very familiar.

Surely a Glaurung facelift would do some miracle.
:wink:

Tord Romstad
Posts: 1808
Joined: Wed Mar 08, 2006 8:19 pm
Location: Oslo, Norway

Re: How much of Stockfish code is still from Tord Romstad?

Post by Tord Romstad » Tue Sep 17, 2013 8:52 pm

It is still similar enough that I can orientate myself in the code very easily, much more than I could in a foreign program. The least changed part of the code is the evaluation function, which is still very recognisably Glaurung. There are some new terms, some code changes, and lots of tweaked parameters, but conceptually it is extremely similar. The endgame knowledge also hasn't changed much, apart from stylistic changes. At the other end of the spectrum, the parallel search code (the non-parallel parts of the code haven't changed quite as much) is radically altered.

Taken as a whole, the changes aren't big enough that the program could pass off as an original engine in an official computer chess tournament (in the hypothetical scenario that Glaurung was still actively developed and entered the same tournament), but they are certainly big enough that Joona and Marco deserve a vast amount of credit for the incredible work they have done, and have more than earned their status of co-authors of the program.

Modern Times
Posts: 2410
Joined: Thu Jun 07, 2012 9:02 pm

Re: How much of Stockfish code is still from Tord Romstad?

Post by Modern Times » Tue Sep 17, 2013 8:57 pm

Tord Romstad wrote:........Joona and Marco deserve a vast amount of credit for the incredible work they have done, and have more than earned their status of co-authors of the program.
Yes, agreed.

Daniel Shawul
Posts: 3758
Joined: Tue Mar 14, 2006 10:34 am
Location: Ethiopia
Contact:

Re: How much of Stockfish code is still from Tord Romstad?

Post by Daniel Shawul » Tue Sep 17, 2013 9:58 pm

Why are you so focused on stirring trouble by raising ownership issues here and there, which no one asked you for anyway? Maybe you should run it thorugh your similarity test to be objective, he ?? I am sure stockfish authors are smart enough not to engage in such threads, but your efforts are irritating to say the least...

It is my opinion that it is extremely offensive to ask if Stockfish authors have earned authorship, even if you state it with so many adjectives which mean nothing. The real question should be stated the other way round. Have the original authors earned the improvements, which in stockfish case would be code evolution and elo as well? For me it is a no. Only in chess arena would you have to worship bare bone chess engine authors not to get persecuted by so many ready to do that.

Post Reply