Synapse Chess

[mjlef]

>Attempts to contact the Synapse guys have bounced back

don't understand why you are so critical about this program;
had a look at it, and it appears to be a neural network (as the name also already suggests), a bit like AlphaGo, and then probably only using Stockfish and Komodo for some tuning or treesearch; in fact having analyzed Ed S's huge book now with Synapse, i found to get overwhelming advantage for 1.c4! (1.d4 doesn't work anymore because of some improve Gruenfeld variations for Black); so although i used to think that chess is a draw with best play, it now appears that White can win, although solving the game with the updated English variations will still take me some months. And this all thanks to Synapse !
jef

Jef,

Finally some comedy on this site! Although I find stealing programs no so funny.

Just to be clear the Synapse output exactly matches Gull, Stockfish or Komodo depending on the version. No neural network. Just copyright infringement. If they have any decency, they will return any money they took.

Mark

[bob]

>Attempts to contact the Synapse guys have bounced back

don't understand why you are so critical about this program;
had a look at it, and it appears to be a neural network (as the name also already suggests), a bit like AlphaGo, and then probably only using Stockfish and Komodo for some tuning or treesearch; in fact having analyzed Ed S's huge book now with Synapse, i found to get overwhelming advantage for 1.c4! (1.d4 doesn't work anymore because of some improve Gruenfeld variations for Black); so although i used to think that chess is a draw with best play, it now appears that White can win, although solving the game with the updated English variations will still take me some months. And this all thanks to Synapse !
jef

Are you serious?

[AndrewM]

Lol...

Jeff, as you might be aware that Synapse team had sent me a copy of their program as noted on their website.

I did extensive analysis with Synapse and came to the following conclusion:

1. It is no neural network nothing.
2. It just has a strong, original program below a wrapper layer.
3. The wrapper layer adds absolutely no value to the original engine except for polluting the output stream with random numbers under the IDDFS tag.
4. Effectively they only have created a wrapper to hide the origins which are either stockfish, gull or Komodo depending on the Synapse version you chose.

This can very easily be verified with various fixed depth searches. Key attributes to look at are node count and the principal variation.

It's amusing where people come up with or believe the neural network none sense with Synapse.

I agree with Mark that the money they have collected should have rightfully gone to respective programmers, for Komodo its copyright infringement, and in case of Gull and Stockfish, it's GPL violation.

It is not funny when it comes to real money some people have lost to these guys.

Regards,Andrew

[jefk]

[quote="Steve Maughan"]Jef,
(or is it an early April Fools joke )- Steve[/quote]

hmm it's still 1/4 but anyway, so my reply might
not be taken seriously, but, believe it or not,
here in NL we always are about six hours earlier
(that's because we are so smart you see, some
scientist suggest it also has something to do
with movement of the sun, or supreme horror,
the rotation of the earth); now i think of it, it was
rather later here, 1.30 Am or so past midnight, now i
think of it, could it already have been April ?
don't know really (still working on my English (1.c4)



[quote="jefk"][...]although solving the game with the updated English variations will still take me some months[...][/quote]
ROTFL!

[Amateur]

I tested this engine for several months and I can say that it is anything but a komodo clone. I regret that by a few weeks the authors do not send me any versions to testing.
I hope that Alejandro Torres is still working on the engine Synapse and soon to contact his friends and supporters, with a new version to testing.