Similarity test for Capivara > version 0.0.8 available?

[Guenther]

Does anyone ever has made or posted similarity test results for
Capivara versions 0.0.8 and 0.0.9?
I confess I have some doubts about it since the huge jump beginning
after version 0.0.7. It seems +600-700 and then 200 again.
(so +900 after version 0.0.7 until 0.0.9)

What was the time difference between last version 7 and 8?
I also see now in Leos past lists release of sources were stopped
after version 7. Also it appears dozens are perhaps hundreds(?)
betas are available with version 8 at once?
IIRC it started as a (legal) TSCP derivate, correct?

I wonder why I didn't find similar questions so far in the CCC search.
The only thing I found now is that in the beginning of versions 8 it displayed nothing, just the move it made.

You can compare the versions strength-wise e.g. in Vincents list:

http://home.scarlet.be/vincentlejeune/r ... 160414.txt


Guenther

[Sven]

The CPW article about Capivara refers to at least two forum posts where one seems to be written around the 0.07 time and the other around the time of 0.08 development. Within the latter the author also explained why he did not publish the source code for recent versions. He also made clear that his program was based on TSCP but with lots of changes.

There are one and a half year between both threads which could be a sufficient time frame for substantial improvements, especially if major bugs of early versions were fixed. Nevertheless a similarity test is always interesting. Indeed the major jump in playing strength between 0.07 and 0.08 is remarkable.

[Guenther]

The CPW article about Capivara refers to at least two forum posts where one seems to be written around the 0.07 time and the other around the time of 0.08 development. Within the latter the author also explained why he did not publish the source code for recent versions. He also made clear that his program was based on TSCP but with lots of changes.

There are one and a half year between both threads which could be a sufficient time frame for substantial improvements, especially if major bugs of early versions were fixed. Nevertheless a similarity test is always interesting. Indeed the major jump in playing strength between 0.07 and 0.08 is remarkable.

According to Leos old news pages, which I have found all now (wayback and more) and integrated for historical reasons and for verification into
my new RWBC chronology, there were 13 months between both.
IMO 0.0.8 had nothing to do at all with TSCP and surely not with fixed bugs in 0.0.7. (0.0.7 was not overwhelmingly better than TSCP it seems.)
It is also interesting that there is an incredible swing (300! diff) in ratings for several 0.0.9 versions independent from what version was older/newer.
(This is normally also an indication for something else happening)

[Guenther]

BTW what was the conclusion for the thread below, if any?

'Fritz and Naum shown as Rybka/Strelka clones'

http://talkchess.com/forum/viewtopic.ph ... 20&t=40795

[Adam Hair]

BTW what was the conclusion for the thread below, if any?

'Fritz and Naum shown as Rybka/Strelka clones'

http://talkchess.com/forum/viewtopic.ph ... 20&t=40795

An inspection of Naum 4.2 code by way of disassembly confirmed the similarity results with Strelka 2.1. I do not know of any independent confirmation (code similarities) for Fritz.