Opening analyses: SF / Komodo to 88% similar, not nice!!

[Lyudmil Tsvetkov]

Where is the similarity tester by Don Dailey?

certainly, a human or engine playing 2.d5 should be considered alienated, but I do not see where the similarity test fits in all this.

[leavenfish]

I have complete certainty the stonewall is completely drawn in a short while, only reasonable move for white seeking an advantage after 1...f5 is 2.d5.
.

Sometimes the less one says the better...you may have thrown your credibility out the window with this statement.

[Lyudmil Tsvetkov]

I have complete certainty the stonewall is completely drawn in a short while, only reasonable move for white seeking an advantage after 1...f5 is 2.d5.
.

Sometimes the less one says the better...you may have thrown your credibility out the window with this statement.

why?

2.c4 e6 3.d5 transposes:

[d]rnbqkbnr/pppp2pp/4p3/3P1p2/2P5/8/PP2PPPP/RNBQKBNR b KQkq - 0 3

while 2.c4 c6 already improves for black


so 2.d5 seems to be better than 2.c4

as said, the Dutch stonewall is fully and easily drawn: I have checked that thousand times.

of course, if white insists on winning, it might also easily get worse.

modern theory is mostly completely irrelevant, built upon strictly personal choices other people liked for some unspecified reason, as well as random events: a strong player playing that line, for example, even if it might objectively be weak.

sometimes, theory did not budge ahead for half-centuries.

for example, one pet theoretical line is Botvinnik's QGD treatment: 1.d4 d5 2.c4 e6 3.Nc3 Nf6(as said, f5 is better) 4.cd5? ed5 5.Bg5:

[d]rnbqkb1r/ppp2ppp/5n2/3p2B1/3P4/2N5/PP2PPPP/R2QKBNR b KQkq - 0 5

and now both Botvinnik and theory(that followed exclusively Botvinnik, of course) give 5...Be7 for black, with a long continuation leading to very promising white play.

of course, above, after the cd5 exchange mistake, black easily gets a small edge instead after 5...c6, followed by h6 and developing the bishop to d6.

most human players will still believe the line favours white, as this is repeated via all possible channels, books, chess lessons, databases, and most players would just take in the readily available information, while in actual fact the line includes a couple of very serious mistakes, and 5...c6 easily gives black small edge.

so much about modern theory.

[corres]

Similarity tester of Don Dailey is a tool to investigate similarity between chess engines. That was not a far time when "experts" based on the results of this tool stated the verdict about an engine to be a clone or not.

[Guenther]

Similarity tester of Don Dailey is a tool to investigate similarity between chess engines. That was not a far time when "experts" based on the results of this tool stated the verdict about an engine to be a clone or not.

Well, obviously you don't know how the simtest works and what positions and search times are used for it... this is a completely different similarity - no need to hijack that thread.

[Lyudmil Tsvetkov]

Similarity tester of Don Dailey is a tool to investigate similarity between chess engines. That was not a far time when "experts" based on the results of this tool stated the verdict about an engine to be a clone or not.

still don't make sense of the full connection.

[Frank Quisinsky]

Hi Guenther,

thats right.

In middlegame both programs have in my opinion an other style (FCP Rating List move statistics). Nice to have Stockfish and Komodo.

For openings the Excel workbook results should be clear. Means if we put a template over the analyses of both engines interesting things are to see. Without more words because I am sure that each one can understand the grafics I added before.

Best
Frank

[mjlef]

Similarity tester of Don Dailey is a tool to investigate similarity between chess engines. That was not a far time when "experts" based on the results of this tool stated the verdict about an engine to be a clone or not.

Don made it very clear that you cannot declare one program is the clone of another just by using the similarity tester. It is meant to be a tool to raise suspicions which would lead to investigating the programs involved.

[mjlef]

Frank,

The link does not seem to work.

I don;t understand what you mean by 88% similar. DO you mean evals, or move matching?

[Frank Quisinsky]

Hi Mark,

oh, sorry!
Link changed yesterday ...
http://www.amateurschach.de/download/bo ... 20done.zip

88% ...
22.032 opening end positions analyzed.
In 88% the same first move or better in 19.422 of 22.034 = 88,15%. That is a lot if I am looking in FEOBOS results from others I have.

Also the other stats ...
Same average of ply, same average of nps ... have a look in the grafics. Average of Eval after 22.032 is 0.09 higher ...

So I am wrote that for opening analyses differents between Stockfish and Komodo arn't easy to see if you put a template about the analyses of 22.032 postions we do with the Excel workbook for the still running FEOBOS book project.

Best
Frank