Time to rethink what Vasik Rajlich has done?

[Madeleine Birchfield]

Vasik Rajlich implemented a lot of his own code in his evaluation and used a unique search in Rybka, something that many other chess engine developers at the time of his banning in 2011, like Norman Schmidt and Robert Houdart, never did. Vasik Rajlich was one of the first engine authors who tuned their engine's evaluation function as well, at a time when most other chess engines refused to even consider automatic tuning methods. Time to rethink what Vasik Rajlich has done? From what I could tell, Vasik Rajlich's efforts are far more unique than Albert Silver's, whose only accomplishments of note is taking the Leela and Stockfish engines and calling it Fat Fritz 1 and Fat Fritz 2, and taking ideas from jjoshua about training Leela and Stockfish networks and calling it his own.

[kranium]

Vasik Rajlich implemented a lot of his own code in his evaluation and used a unique search in Rybka, something that many other chess engine developers at the time of his banning in 2011, like Norman Schmidt and Robert Houdart, never did. Vasik Rajlich was one of the first engine authors who tuned their engine's evaluation function as well, at a time when most other chess engines refused to even consider automatic tuning methods. Time to rethink what Vasik Rajlich has done? From what I could tell, Vasik Rajlich's efforts are far more unique than Albert Silver's, whose only accomplishments of note is taking the Leela and Stockfish engines and calling it Fat Fritz 1 and Fat Fritz 2, and taking ideas from jjoshua about training Leela and Stockfish networks and calling it his own.

Hey Madeleine
I believe Vas pioneered parallel testing at ultra-fast time controls, where you could play many games and get results with a very high LOS in a relatively short amount of time, a now old-fashioned technique that I still use today. He had great success with this approach, but I don't believe his eval tuning was automated in any way. It was still a matter of manually incrementing or decrementing the value of a variable, testing for 40000 13 sec games (1000 ms + 100ms for ex), measuring the result, and changing the var again until you have definitive (99% or greater LOS) elo gain.
Reading a text file of parameter on engine startup was the the easiest wat to do this, much faster than creating UCI options for the targeted variable.

Not sure why I'm included in this post...my work back then consisted of stabilizing and improving Ippolit...for ex: the creation of the 1st SMP versions (Iggorit), all of which was published opensource GPL by myself and Milos S.
This has all been very well documented.
http://users.telenet.be/chesslogik/robbolito.htm
https://www.chessprogramming.org/RobboLito

For a short time open-source GPL'd SMP Robbolito was the strongest engine that existed...about 50 Elo better than Rybka 3...until Houdini 1.0 (based on Robbolito, but a bit stronger) appeared.

I do believe Vas benefitted from open-source Fruit (and possibly Crafty as well), but I also think Vas deserves a lot of credit.
When presented with Strelka source code (RE'd Rybka). Fabien L. recognized things, and even stated "it's ok it's been re-written" (paraphrased)

In any event, none of this really seemed to matter and is all water under the bridge, as you know, both Rybka and Houdini continued commercially for many years with great success...just like Fat Fritz will I guess.

[Modern Times]

The most disgusting thing here was the way the world press handled this, involving Vas's wife as well. I don't know what role the ICGA played in that. Certainly they must have issued a press release, and then the press did their worst. I took a photocopy of the printed London Metro article at the time:


http://ccrl.chessdom.com/public/Metro_A ... 0Rybka.PNG

[Enzo]

Hear, hear.

Vas was treated outrageously.

Levy and other nonentities, consumed by envy, should be strung up. Simply no question about it.

[mar]

so what has he done? he sent some old crafty to compete in Olivier's chesswar (or openwar) disguised as Rybka 1.6.1 (or something) - this is lame and stupid and hard to understand. this has been proven beyond any doubt

as for "post-fruit" Rybkas - a wrong version was compared to Fruit (and even that comparison fell short, because Wegner et al were high and seeing "copied code" where there was none or imaginary at best)
so he allegedly "took too much" according to the panel, but too much is simply vague
of course, Wegner had a motivation of his own as he won WCCC with Cozzie's program later and retired as a "champ" when Vas was out of the way

if we'd look at todays top open source engines with 2011 optics, very few (if any) would pass the originality criteria back then.

[towforce]

Let me quickly play the role of "captain obvious" here:

* he entered a tournament in which the rules were that code used from other programs must be declared

* he didn't declare any used code

* a panel found that his program did contain such code

* if his code was free of other sources, he could have proven it. He chose not to.

I have looked at the evidence, and I personally think that it's good enough. If it wasn't, his supporters could have provided counter examples anyway.

His program was stronger than the others at the time, but when you enter a competition, you must agree to the rules, and abide by the referee's decision. Referees need respect and protection as well!

[flok]

I have looked at the evidence, and I personally think that it's good enough.

It's that disassembled code, right?

[Rebel]

Let me quickly play the role of "captain obvious" here:

* he entered a tournament in which the rules were that code used from other programs must be declared

* he didn't declare any used code

* a panel found that his program did contain such code

* if his code was free of other sources, he could have proven it. He chose not to.

I have looked at the evidence, and I personally think that it's good enough. If it wasn't, his supporters could have provided counter examples anyway.

His program was stronger than the others at the time, but when you enter a competition, you must agree to the rules, and abide by the referee's decision. Referees need respect and protection as well!

Look into the ICGA verdict, not a word about copied code.

[towforce]

Reading and understanding the documents takes time: I have done it, but I'm not going to do it again. There seem to be links to the documents here.

[chrisw]

so what has he done? he sent some old crafty to compete in Olivier's chesswar (or openwar) disguised as Rybka 1.6.1 (or something) - this is lame and stupid and hard to understand. this has been proven beyond any doubt

as for "post-fruit" Rybkas - a wrong version was compared to Fruit (and even that comparison fell short, because Wegner et al were high and seeing "copied code" where there was none or imaginary at best)
so he allegedly "took too much" according to the panel, but too much is simply vague
of course, Wegner had a motivation of his own as he won WCCC with Cozzie's program later and retired as a "champ" when Vas was out of the way

if we'd look at todays top open source engines with 2011 optics, very few (if any) would pass the originality criteria back then.

I knew Zach Werner actually quite well, given social media, and discussed with him a few times by email. At the time he was still a student, not written any chess program, and his main motivation, as far as I could tell, was a new found social acceptance into a world of “famous” people, increased much by working on the “evidence” document. He was just a student and he told me afterwards that the document was just a first shot on which he expected critique and discussion. It wasn’t meant to be any final paper (Zach’s words, to me, in email). In fact, he said, the reception was no critique at all, just an immediate “great, this is just what we need”. Again, in fact, any college supervisor actually reading it with some knowledge on the subject and the necessary objectivity would have put a big red line through it and either marked him “Fail” or “Start Again”. Zach later realised that the social group he so aspired to be accepted into had collective feet of clay and were mostly either idiots or really rather unpleasant, that he had been used, and he washed his hands of computer chess and all to do with it, leaving, never to return. I guess he got a useful lesson at a relatively early age but it was not very nice for him, he being a victim of it too.
The lesson? The entire thing against Vas was fired by social group dynamic, it was a large social group, the inner core had “titles”, people are attracted to the idea of being “in”, they believe “titles”, the social group welded itself together by creating an “outside” hate figure. The hate figure suited many of them for commercial/professional/status/personal reasons. The evidence wasn’t meant to be read, it was meant to be bulky. The evidence wasn’t read. It wasn’t critiqued. The was no red team, only a blue one, and they fixed it so there was no internal disagreement.
An interesting lesson in how things can go catastrophically wrong, a lesson in how everything a social group knows can be wrong, shame it had to destroy somebody’s life in the process.