New uci engine: Rofchade

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

Moderators: Harvey Williamson, bob, hgm

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Dokterchen
Posts: 115
Joined: Wed Aug 15, 2007 10:18 am
Location: Munich

Re: New uci engine: Rofchade

Post by Dokterchen » Wed Aug 29, 2018 11:18 am

Ronald wrote:
Tue Aug 28, 2018 11:30 am
One of the reasons why asymmetrical tables seem to work best (for now?) probably is more short castling, but I think chess is not symmetric at all, due to the fact that we have 1 queen and king, and we have alternating colors for the squares (so bishop 1 can not reach square C1 for instance and bishop 2 van not reach C8).
WRONG! :) 35 years ago, when I was around 18 years old, I played a blitz tournament and I desperately needed a white colored bishop to attack the opponent's king, but I have had just the black one. Very easy to solve. I moved the Bc1 to h5! :D
Torsten

F. Bluemers
Posts: 860
Joined: Thu Mar 09, 2006 10:21 pm
Location: Nederland
Contact:

Re: New uci engine: Rofchade

Post by F. Bluemers » Wed Aug 29, 2018 4:49 pm

Congrats with your strong engine.
I have been playing with a psq only eval,texel tuned also.It was about 150 - 170 elo weaker than the regular evaluation.
So you can look forward for some more nice elo
Best Fonzy

PK
Posts: 863
Joined: Mon Jan 15, 2007 10:23 am
Location: Warsza
Contact:

Re: New uci engine: Rofchade

Post by PK » Thu Aug 30, 2018 5:43 am

These tables actually might turn quite strong. so far I did only a quick experiment, replacing rows 2 to 5 in Rodent's midgame piece/square table (Your values in rows 6-7 are clearly compensating for the absence of passed pawn evaluation). Surprisingly, this passed the test, very narrowly, but passed. And we are talking about just one table, used partially, and not synchronized with the rest of Rodent evaluation. Usually things like that just don't happen.

Which brings another question. Good tables imply good set of test positions for tuning (and/or some twist to the tuning code). Are You using some known position set or have You created something special?

User avatar
Ronald
Posts: 112
Joined: Tue Jan 23, 2018 9:18 am
Location: Rotterdam
Full name: Ronald Friederich
Contact:

Re: New uci engine: Rofchade

Post by Ronald » Thu Aug 30, 2018 8:27 am

F. Bluemers wrote:
Wed Aug 29, 2018 4:49 pm
Congrats with your strong engine.
I have been playing with a psq only eval,texel tuned also.It was about 150 - 170 elo weaker than the regular evaluation.
So you can look forward for some more nice elo
Best Fonzy
Thanks!

I would expect that you could gain more elo (double like 300?) with a good evaluation function, but maybe tuned PS tables can "simulate" more evaluation elements then expected.

User avatar
Ronald
Posts: 112
Joined: Tue Jan 23, 2018 9:18 am
Location: Rotterdam
Full name: Ronald Friederich
Contact:

Re: New uci engine: Rofchade

Post by Ronald » Thu Aug 30, 2018 8:57 am

PK wrote:
Thu Aug 30, 2018 5:43 am
These tables actually might turn quite strong. so far I did only a quick experiment, replacing rows 2 to 5 in Rodent's midgame piece/square table (Your values in rows 6-7 are clearly compensating for the absence of passed pawn evaluation). Surprisingly, this passed the test, very narrowly, but passed. And we are talking about just one table, used partially, and not synchronized with the rest of Rodent evaluation. Usually things like that just don't happen.

Which brings another question. Good tables imply good set of test positions for tuning (and/or some twist to the tuning code). Are You using some known position set or have You created something special?
:D It was your idea to organise a challenge for PST! With that topic you showed that PST can have a serious influence on strength (stand alone) and it resulted in an interesting discussion also. I'm really curious how much a tuned PST can simulate evaluation parts of a regular evaluation function. I hope to find out the coming months for myself...

Concerning the testset, I used the "quiet-labeled".epd set created by Alexandru Mosoi (Zurichess).

User avatar
Scally
Posts: 135
Joined: Thu Sep 28, 2017 7:34 pm
Location: Bermondsey, London
Full name: Alan Cooper
Contact:

Re: New uci engine: Rofchade

Post by Scally » Tue Sep 11, 2018 11:54 am

Hi Ronald,

I’ve tried leaving comments on your Rofchade website but they get deleted.

When do you expect the source code to be available, are we talking weeks or Months?


Thanks,

Al.
Alan Cooper
My Chess Computers

User avatar
mclane
Posts: 18131
Joined: Thu Mar 09, 2006 5:40 pm
Location: US of Europe, germany
Full name: Thorsten Czub
Contact:

Re: New uci engine: Rofchade

Post by mclane » Tue Sep 11, 2018 12:14 pm

There was an engine called breakthrough by Werner Koch.

IMO it computed very complex amount of PST etc. for all the positions in the tree and this way the program
Was often capable to make the opponent nearly not move at all.
Kind of stalemate. Where any more creates a loss,


Of course it was an experiment and breakthrough was never that strong but the games were very
Funny and reminded on a draughts game,

https://www.game-ai-forum.org/icga-tour ... php?id=196
What seems like a fairy tale today may be reality tomorrow.
Here we have a fairy tale of the day after tomorrow....

User avatar
Ronald
Posts: 112
Joined: Tue Jan 23, 2018 9:18 am
Location: Rotterdam
Full name: Ronald Friederich
Contact:

Re: New uci engine: Rofchade

Post by Ronald » Tue Sep 11, 2018 5:36 pm

Scally wrote:
Tue Sep 11, 2018 11:54 am
Hi Ronald,

I’ve tried leaving comments on your Rofchade website but they get deleted.

When do you expect the source code to be available, are we talking weeks or Months?


Thanks,

Al.
Sorry, I that replies on the website are not posted directly, I first have to approve them before they are show on the site, I didn't know that..., so your reply was received, but not approved.

I'm reorganizing the source code at the moment. What should not have done is at the same time making some changes to the code, it all works now but it's better to not do 2 things at once.... I still want to do some cleanup/commenting on the source code before I make it available, it will probably be next week.

From your reply I understand that you would like to have/make a raspberry pi version for playing with picochess. I own a Digital PI of DGT myself and a few months ago it was a special moment to play on a DGT board against my own engine! I lost of course but it was great fun.
I use Visual studio for the windows builds and clang for MacOS, and the raspberry pi build from a few months ago was done with GCC, so I don't know if the current source will compile in GCC (usually not right away...). I will try to build a raspberry pi version tonight or tomorrow, as soon as I have one I will send you a message/email.

User avatar
Scally
Posts: 135
Joined: Thu Sep 28, 2017 7:34 pm
Location: Bermondsey, London
Full name: Alan Cooper
Contact:

Re: New uci engine: Rofchade

Post by Scally » Tue Sep 11, 2018 6:03 pm

Hi Ronald,

Thanks for your reply, I look forward to your Engine.

I have Updated the base Raspberry Pi compilers of gcc, gdc, g++ & c++ to version 8.1.0; clang to version 6.0.0 plus added Go, ldc2, java and pascal compilers. I’ve used all to compile a bunch of engines running on Raspbian Stretch on my DGT Pi.

Recently I ported these over to the ASUS Tinker Board. This runs on TinkerOS Debian Stretch, slightly different but the majority of the engines ran immediately. The remaining 4 compiled on the Tinker Board and connect to my DGT Board fine.


Thanks,

Al.
Alan Cooper
My Chess Computers

User avatar
Ronald
Posts: 112
Joined: Tue Jan 23, 2018 9:18 am
Location: Rotterdam
Full name: Ronald Friederich
Contact:

Re: New uci engine: Rofchade

Post by Ronald » Wed Sep 12, 2018 7:44 pm

I managed to create a version for the Raspberry Pi on my DGT Pi, with the help of my son. I did a short test and everything seems to work, even multi threading. The version can be downloaded from the website below. Let me know if there are problems with the version. :D

Post Reply