Brainfish, a new concept of a chess engine

[harri200k]

Mio4 ctg Download Link please.

[velmarin]

Mio4 ctg Download Link please.

It is impossible to upload this. At least for me, sorry
It is easier to build one's own computer.
For example Ed material, this included

http://rebel13.nl/download/books.html

[Ozymandias]

Code: Select all

it is quite achievable to compute and store over thousands of billions chess positions with ArShah.

I find such statements to be suspicious. While I agree that it is achievable, it is also very time consuming, and the data would be unmanageable.

Besides that, there's the reliability of an engine score, take this position, from this season's TCEC:

5k2/2R5/5K2/1p6/p5b1/P7/8/8 b - - 26 75

At depth 61, Komodo evaluates this as +1.44. With access to 6 men syzygy, I would usually think that such an eval, is on the road to winning, but we're looking at a draw. If you minimax that score, you would need to navigate all the way down to move 75, before realising this isn't what you what. Not very helpful.

[Canoike]

Hello,

Version of today July 31. There is a problem with TZbook.*, Linux makes the difference between lower and upper case.
I have tried changing the upper/lower case, the compilation error comes later...

Regards.

[fern]

No 32 bits?

[clumma]

So what does talkchess think of this work, currently #1 on Pohl's rating lists?

* It's been done before, in a 3-page paper with no code from 1999
* It's just an opening book
* Maybe it violates the GPL?
* What, no Ubuntu version??
* What, no 32-bit version??

As a casual observer, I must say these responses are disappointing. And pretty funny, actually.

I'll even chip in: your website (zipproth.com) is slow as molasses and I haven't been able to get the Demo tab to load properly in either Chrome or Firefox. Also, an English version of the Cerebellum tab would be appreciated.

-Carl

[Dann Corbit]

So what does talkchess think of this work, currently #1 on Pohl's rating lists?

* It's been done before, in a 3-page paper with no code from 1999

Not sure if this exactly has been done before.

* It's just an opening book

It's a special opening book. It is different in several aspects, the most important of which is that you do not have to be an openings expert to generate one.

* Maybe it violates the GPL?

I doubt it

* What, no Ubuntu version??

If you have Ubuntu, you should know how to compile

* What, no 32-bit version??

There is source code, so someone can build one.

As a casual observer, I must say these responses are disappointing. And pretty funny, actually.

I'll even chip in: your website (zipproth.com) is slow as molasses and I haven't been able to get the Demo tab to load properly in either Chrome or Firefox. Also, an English version of the Cerebellum tab would be appreciated.

-Carl

[clumma]

Dann,

My bullet points were a critical summary of discussion that already took place in this thread.

-Carl

[Jesse Gersenson]

In order to rank moves correctly it would be useful to know more about each data point (move) in the 'book'. Knowing the nps, number of threads, depth and engine responsible for a data point would be very useful if one needed to filtering data. Also maybe a timestamp.

[shrapnel]

Don't listen to them, Thomas, they are all envious of your excellent work.
When do you plan to go commercial ?
Waiting eagerly...