Soberango 0.00.0 realized.

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

User avatar
Luis Babboni
Posts: 464
Joined: Sat Feb 28, 2015 4:37 pm
Location: Argentina

Re: My test (Soberango 0.02.0)

Post by Luis Babboni »

Thanks HG!

Actually Soberango search method keep a ply counter that tracks how far is from the root :oops:

At each step improving my engine I try to do, I think is time to start again a cleaner code taking in account since the begining all the things I was learned in the road to reach where I am, but at the same time I think that as far as I could go without start a new code, more things I could add to this new code without need to start a 3rd code! :roll:
This seems to be one of those things that if I started a new code some days ago, today I think I need to start a 3rd code! :roll:

Hard but nice job ahead seems to wait for me in this adventure!!
User avatar
Luis Babboni
Posts: 464
Joined: Sat Feb 28, 2015 4:37 pm
Location: Argentina

Re: My test (Soberango 0.02.0)

Post by Luis Babboni »

Sven Schüle wrote: ... You can assign a draw score (e.g. 0) to all KNNK positions, and that will be fine in more than 99.99% of all cases. Or you can try to be even more accurate by taking care of the "mate-in-1" special case, with rules like those I mentioned above.
Please let me kown again if I understood right:

In the position you posted some posts ago:
[d] 6k1/8/6K1/5NN1/8/8/8/8 w - - 0 1
whites needs black colaborates to be checkmated in 2.

The suggestion is then to make a rule that see if there is checkmate in 1, not in 2 with colaboration.
For example after kxh8 in this position (if there is no bishop in the corner, we are in a KNNK case with need of colaboration):
[d] 6kB/4N3/6K1/6N1/8/8/8/8 b - - 0 1

In that case my "little tablebase" for KNNK seems to need just 238 positions without use simetries.
User avatar
Luis Babboni
Posts: 464
Joined: Sat Feb 28, 2015 4:37 pm
Location: Argentina

Re: My test (Soberango 0.02.0)

Post by Luis Babboni »

Soberango 0.04.2 realized.
Added two little "tablebases" to indentify checkmate in 1 and checkmate situations in KNNK endgames and not take those positions as tie cause insufficient material.
http://www.luisbtenis.com.ar/Soberango/Soberango.html
User avatar
Luis Babboni
Posts: 464
Joined: Sat Feb 28, 2015 4:37 pm
Location: Argentina

Re: My test (Soberango 0.02.0)

Post by Luis Babboni »

Fixed a bug with 50 moves rule.
Soberango 0.04.3 uploaded:
http://www.luisbtenis.com.ar/Soberango/Soberango.html

Soon it could be able to have some management of time.
User avatar
Luis Babboni
Posts: 464
Joined: Sat Feb 28, 2015 4:37 pm
Location: Argentina

Re: My test (Soberango 0.02.0)

Post by Luis Babboni »

After an entire month fighting with crashes, I could made stable version and uploaded Soberango 0.05.0 that adds a primitive management of time to reach differents search depth instead of fixed it in 5.

With this, Soberango reaches for first time the level of NSVChes v0.14* in 40/4 games (46-32-22 for Soberango in 100 games match) and be far over previous version (35-13-2 for 0.05.0 vs 0.04.3).
*: nº 494 in the 510 engines list of CCRL for 40/4 games:
http://www.computerchess.org.uk/ccrl/404/

Soberango´s web:
http://www.luisbtenis.com.ar/Soberango/Soberango.html
User avatar
Luis Babboni
Posts: 464
Joined: Sat Feb 28, 2015 4:37 pm
Location: Argentina

Re: My test (Soberango 0.02.0)

Post by Luis Babboni »

Version 0.06.0 realized.
Now Soberango valuates better faster checkmates when is Soberango who will mate and valuate worst faster checkmates if is the opponent wich mates Soberango.
This way there is no change in performance but its make less bored endgames that finish in mate.
http://www.luisbtenis.com.ar/Soberango/Soberango.html
Daniel Anulliero
Posts: 759
Joined: Fri Jan 04, 2013 4:55 pm
Location: Nice

Re: Soberango 0.00.0 realized.

Post by Daniel Anulliero »

Luis Babboni wrote:I took a list of weakest engines from CCRL and, started from low to high with engines found, Soberango have this results in blitz 5/0 games without books:

Image
*: the ELO is estimated by CCRL playing 40 moves in 40 minutes games so the results are just anecdotal:
With the elos you give , you have a perf around 700 elo :)
Congratulation for your release
I'll plan to release my engine too ... soon :wink:
User avatar
Luis Babboni
Posts: 464
Joined: Sat Feb 28, 2015 4:37 pm
Location: Argentina

Re: Soberango 0.00.0 realized.

Post by Luis Babboni »

Hola Daniel!
Eso era el 0.00.0, de a poco le fui agregando cosas y fue mejorando, voy por la versión 0.06.0.
Siempre en el modo 40/4, creo que debe andar por los 1000 de ELO ahora:

Image
Last edited by Luis Babboni on Mon Jun 13, 2016 9:42 pm, edited 1 time in total.
Daniel Anulliero
Posts: 759
Joined: Fri Jan 04, 2013 4:55 pm
Location: Nice

Re: Soberango 0.00.0 realized.

Post by Daniel Anulliero »

Luis Babboni wrote:Hola Daniel!
Eso era el 0.00.0, de a poco le fui agregando cosas y fue mejorando, voy por la versión 0.06.0.
Siempre en el modo 40/4, creo que debe andar por los 1000 de ELO ahora:

Image
Hello Luis

Please , right in english , I'm french and I dont understand the spanish :wink:
But I think you mean your last version is over 700 , 1000 yes ?
Good luck to you !
User avatar
Luis Babboni
Posts: 464
Joined: Sat Feb 28, 2015 4:37 pm
Location: Argentina

Re: Soberango 0.00.0 realized.

Post by Luis Babboni »

Sorry Daniel!!! Cause your name I thought you are spaniard!!!!
I did not know Daniel is in french the same as in spanish.
Yes, Soberango 0.05.0 and soberango 0.06.0 have around ELO 1000 in 40/4 I think.
I´m fighting with my new Dropbox account to use it to upload images to forums but still I cant do it. My old Dropbox account is full and the new acounts seems has not the option to public images, you need now to have the non free account.