Robert 2 Questions

[kgburcham]

I read where you said Vas originally was asking a lot of questions.

Question #1. How many years between time Vas was asking all the questions and the release of the first version of Rybka?

If answer to question #1 is 2,3,4,6 years whatever,
Question #2. If the answer is ? years, is this enough time to write from scratch a first release of a new program that is capable of beating all other commercial programs? Also is that enough time to clear up 98% of bugs? I remember I used to get so mad about the cheap chessmaster program because it was always released with so many bugs. Rybka was very clean with first release and has been with release 2 and 3.

I find the above suspect.

I think based on your answer I will conclude we either have a Michael Jordan/Tiger Woods type Programmer or someone made some quick easy money. I still have not ruled out the first possibility.

kburcham

[bnemias]

Also is that enough time to clear up 98% of bugs?

It is certainly possible to write code in a way that keeps bugs from lasting long, chess engines notwithstanding. (Yes, I know I'm not the Robert you were addressing. heh)

[bob]

I read where you said Vas originally was asking a lot of questions.

Question #1. How many years between time Vas was asking all the questions and the release of the first version of Rybka?

If answer to question #1 is 2,3,4,6 years whatever,
Question #2. If the answer is ? years, is this enough time to write from scratch a first release of a new program that is capable of beating all other commercial programs? Also is that enough time to clear up 98% of bugs? I remember I used to get so mad about the cheap chessmaster program because it was always released with so many bugs. Rybka was very clean with first release and has been with release 2 and 3.

I find the above suspect.

I think based on your answer I will conclude we either have a Michael Jordan/Tiger Woods type Programmer or someone made some quick easy money. I still have not ruled out the first possibility.

kburcham

You are asking something I can't recall with any accuracy, because there was no reason to commit the first date that he posted here or asked questions to memory. But if you want a guess, I would say somewhere between 12 and 24 months maybe...

As far as his programming skills, I don't see how they can be criticized. Even if he copied fruit line for line, he made enough remarkable changes to it to produce a program that is stronger than all others. I assume nobody doubts that part of what has happened. The only question is about the "early days", not the final product, which is outstanding.

My only interest in staying in the discussion is to keep a lot of urban myth out of the discussion, such as how likely is it that two different programmers will produce many identical sections of code? It is _extremely_ unlikely, not a daily occurrence, as some would have you believe.

As far as the "bug" question, I don't think that is too short a time. Crafty played its first game on ICC in December of 2004. By the Summer of 2005 it was basically "complete" (null-move and everything) and playing thousands of games a week without crashing. It took me about 6 months from the time I put the bare-bones program together (something that would play a complete game without crashing) until the final version which included most everything I had done in Cray Blitz (although obviously rewritten since I changed from fortran/assembly/vector-processing to pure C using bitboards.) So the time frame is, to me, reasonable.

[Uri Blass]

I read where you said Vas originally was asking a lot of questions.

Question #1. How many years between time Vas was asking all the questions and the release of the first version of Rybka?

If answer to question #1 is 2,3,4,6 years whatever,
Question #2. If the answer is ? years, is this enough time to write from scratch a first release of a new program that is capable of beating all other commercial programs? Also is that enough time to clear up 98% of bugs? I remember I used to get so mad about the cheap chessmaster program because it was always released with so many bugs. Rybka was very clean with first release and has been with release 2 and 3.

I find the above suspect.

I think based on your answer I will conclude we either have a Michael Jordan/Tiger Woods type Programmer or someone made some quick easy money. I still have not ruled out the first possibility.

kburcham

What are the bugs that you talk about that chessmaster has and rybka does not have?
I think that Rybka is strong inspite of some bugs that it has.

People already posted examples when rybka mp blundered and traded to KP vs K drawn endgame or KBP vs K endgame inspite of using the nalimov tablebases and it is clear that rybka can simply ignore the nalimov tablebases because of the bug(I am not talking about wrong evaluation of simple drawn tablebase positions and it can be considered as design decision assuming Vas had not the time to implement the knowledge and prefered to improve rybka by other ways).

Uri

[frankp]

Crafty played its first game on ICC in December of 2004. .

2004 ?

(Wrong decade ?)

[kgburcham]

I do not have interest in discussing chessmaster program.
Like I said, I used to get mad at John because I thought CM was released too soon before bugs were worked out---he said the CM team was small. CM team used to release the program with bugs and then as the feedback came in, patches were available. I know this is standard process but I thought CM had too many bugs.

My point in making the above reference to buggy CM is that I thought Rybka had very few bugs for being a new program written from scratch.
I had one of the early downloads for the first Rybka release---I cant recall one bug.
I did not have any trouble with Rybka 2.3.2a mp 64---I cant recall one bug.
I find this remarkable.

Uri I have read about the few bugs that are in Rybka. I am aware of what has been posted about Rybka bugs. I still have the opinion that Rybka bugs are very few.

kburcham

[bob]

Crafty played its first game on ICC in December of 2004. .

2004 ?

(Wrong decade ?)

ah crap. 1994. Right after the last ACM tournament was held.

You are right. One decade off...

[Mike S.]

I did not have any trouble with Rybka 2.3.2a mp 64---I cant recall one bug.

Unfortunatley, in fact the Rybkas do have several bugs, despite being so strong overall. Many different examples have been shown on the message boards, and I found most of those I tried myself to be reproducable. Often, they are related to late endgame positions, without but also with tablebases.

From my viewpoint, the King engine was the much more "clean" one in terms of comparable bugs, at least the versions I have from 2.x to 3.23.

Maybe it's also because Rybka is more scrutinized than other engines, or simply used more often so more bugs are found, just because of that. - But to be honest, I have the impression that Rybka 2.3.2a really has more bugs than it is typical for an engine of that category. I don't have version 3 yet but as far as I read, it's not bugfree either and people are hoping for a patch.

Rybka 2, and Rybka 3 especially, is so strong that losing one or two (half) points every 100 games due to such bugs, doesn't affect the Elo much, relative to the total gap to the competitors. But does that compensate for missing confidence in analysis results? I hope the various bugfixes required, have top priority now over attempts to make Rybka even stronger - which is almost unimaginable anyway.