CDrill 1800

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

Moderators: hgm, Rebel, chrisw

Frank Quisinsky
Posts: 6808
Joined: Wed Nov 18, 2009 7:16 pm
Location: Gutweiler, Germany
Full name: Frank Quisinsky

Re: CDrill 1800 for TMET ... vs. the Mephisto's?

Post by Frank Quisinsky »

Hi Brendan,

for FEOBOS is 8-10 move length really a problem!?

Short explanation:
Often ECO codes formed after 1 or 2 moves. If I create now an opening book undo 8-10 moves, to many of good lines for such openings are in the book and opening will be to often on the board. Much other ECO formed formed after 7 moves ... and not enough lines are in the book (other example).

So for FEOBOS:
3 moves after ECO codes formed. Often for ECO codes, formed after one move ... I am working all the time with 500 databases for opening book optimation.

That is indeed a problem because with book prioritys by GUIs lines with more moves will be play more often.

Example:
B01 is formed after 1. ... d5
B01 is normaly in GM theory from 500 ECO codes a TOP-50 code (around on place 30-40). If I added now 8-10 moves for B01 ... B01 is to often on the board with random book options by chess GUIs.

Most book creators forget it.

FEOBOS 20.1 is here very strong.
If lesser balanced positions (for each code 3 moves after ECO Code formed) ECO codes will be play rarely. And if you now compare with GM theory ... most popular lines will be comparable to FEOBOS.

To 100% we can't create the perfect book but in the near of 100% is possible.

:-)

Best
Frank

PS: With FEOBOS and 10 engines analysing each of the lines in book ... the topic is over for me. I have the "perfect" book for eng-eng testing ... created in team work by 10 strong engines with our Excel file. Have a look on FEOBOS Detail page ... all databases in *.pgn, *.epd the Excel ... 33.009 lines sorted with a ranking system in *.pgn file are online. Each of the 33.009 have an own ranking (important if you like to create a test-set).
User avatar
Rebel
Posts: 6991
Joined: Thu Aug 18, 2011 12:04 pm

Re: BTW: Short question to the black Mirage board!

Post by Rebel »

Frank Quisinsky wrote:Hi Ed,

BTW: A short question
Sorry to the others for my BTW question!

Mephisto Mirage board is working fine with MMI, Mirage and Rebell 5.0 modules (means not with the keyboard of modules, means with the sensors on the board).

Is this random or do you support the black Mirage board with an extra work?

Thats great, I like the black mirage board a lot.

I have two Rebell 5.0 modules.
One is tuned to 9.4Mhz. The tuned module is not working with sensors.

If you have time for an answere ... only for my personal interest on Mirage board.

...

:-)

Best
Frank
Hi Frank :wink:

Here some info about the board.

http://www.schaakcomputers.nl/hein_veld ... 09-1984%20[C-2676]%20Mephisto%20-%20Mirage-Brett.pdf

Be well.
BrendanJNorman
Posts: 2526
Joined: Mon Feb 08, 2016 12:43 am
Full name: Brendan J Norman

Re: CDrill 1800 for TMET ... vs. the Mephisto's?

Post by BrendanJNorman »

Frank Quisinsky wrote:Hi Brendan,

for FEOBOS is 8-10 move length really a problem!?

Short explanation:
Often ECO codes formed after 1 or 2 moves. If I create now an opening book undo 8-10 moves, to many of good lines for such openings are in the book and opening will be to often on the board. Much other ECO formed formed after 7 moves ... and not enough lines are in the book (other example).

So for FEOBOS:
3 moves after ECO codes formed. Often for ECO codes, formed after one move ... I am working all the time with 500 databases for opening book optimation.

That is indeed a problem because with book prioritys by GUIs lines with more moves will be play more often.

Example:
B01 is formed after 1. ... d5
B01 is normaly in GM theory from 500 ECO codes a TOP-50 code (around on place 30-40). If I added now 8-10 moves for B01 ... B01 is to often on the board with random book options by chess GUIs.

Most book creators forget it.

FEOBOS 20.1 is here very strong.
If lesser balanced positions (for each code 3 moves after ECO Code formed) ECO codes will be play rarely. And if you now compare with GM theory ... most popular lines will be comparable to FEOBOS.

To 100% we can't create the perfect book but in the near of 100% is possible.

:-)

Best
Frank

PS: With FEOBOS and 10 engines analysing each of the lines in book ... the topic is over for me. I have the "perfect" book for eng-eng testing ... created in team work by 10 strong engines with our Excel file. Have a look on FEOBOS Detail page ... all databases in *.pgn, *.epd the Excel ... 33.009 lines sorted with a ranking system in *.pgn file are online. Each of the 33.009 have an own ranking (important if you like to create a test-set).
My apologies Frank, I misunderstood (I think).

Yeah, 8-10 moves (or even up to 15 moves at 2100 level) is fine.

In fact, it'd be cool if the UCI options determined how deep the book went (if set at 2100 it uses the book to move 15, if 1800 it uses to move 10 etc).

Nothing worse than playing against a computer, thinking for ten minutes only to have the computer respond instantly from it's book on move 19 :lol:

This is was I was originally referring to (and fearing!) :wink:
Frank Quisinsky
Posts: 6808
Joined: Wed Nov 18, 2009 7:16 pm
Location: Gutweiler, Germany
Full name: Frank Quisinsky

Re: BTW: Short question to the black Mirage board!

Post by Frank Quisinsky »

Hi Ed,

so my observation is right.
I added the link on my site.

Thanks!

Best
Frank
User avatar
Rebel
Posts: 6991
Joined: Thu Aug 18, 2011 12:04 pm

Re: CDrill 1800 for TMET ... vs. the Mephisto's?

Post by Rebel »

BrendanJNorman wrote:
Rebel wrote:A while ago I released ProDeo 2.4 but received hardly any feedback if there is value in the chosen framework for improvements.

http://rebel13.nl/prodeo/prodeo24.html
Hi Ed,

Can this Elo adjustable version use personalities from ProDeo 2.6?

The ones created with your online tuner, I mean...

I wouldn't mind trying some of the personalities I've created at say 2200 Elo to see the resulting style. 8-)
You will need the 2.2 online editor.

http://rebel13.nl/pers.html
User avatar
PeterO
Posts: 215
Joined: Sun Jul 31, 2016 6:35 pm

Re: CDrill 1800 for TMET ... vs. the Mephisto's?

Post by PeterO »

Hi Ferdinand,

are there news about CDrill? Have you alteady realised ideas from this forum?

Peter
Ferdy
Posts: 4833
Joined: Sun Aug 10, 2008 3:15 pm
Location: Philippines

Re: CDrill 1800 for TMET ... vs. the Mephisto's?

Post by Ferdy »

PeterO wrote:Hi Ferdinand,

are there news about CDrill? Have you alteady realised ideas from this forum?

Peter
No major updates really except that I am trying to improve its time control handling for lichess bot.

I created CDrill lichess bot in the hope that I can gather games from humans. This is a good platform because the games are saved so I can review it later anytime.

Currenlty I allowed CDrill bot to use Chris Kofinas' setting.

Code: Select all

Material Weight 280 
Pawn Structure Weight 170 
Passed Pawn Weight 90 
Mobility Weight 200 
Threat Weight 300 
King Attack Weight 30 
King Shelter Weight 250 
Repeat Score -50 
GregNeto
Posts: 35
Joined: Thu Sep 08, 2016 8:07 am

Re: CDrill 1800 for TMET ... vs. the Mephisto's?

Post by GregNeto »

Would it be possible to make CDrill's strenght hardware-independent?
Ferdy
Posts: 4833
Joined: Sun Aug 10, 2008 3:15 pm
Location: Philippines

Re: CDrill 1800 for TMET ... vs. the Mephisto's?

Post by Ferdy »

GregNeto wrote:Would it be possible to make CDrill's strenght hardware-independent?
Yes it is possible by controlling its nodes or depths searched. At some point I will add such option.
GregNeto
Posts: 35
Joined: Thu Sep 08, 2016 8:07 am

Re: CDrill 1800 for TMET ... vs. the Mephisto's?

Post by GregNeto »

Thank you. A hardware-independent, stable, well calibrated and easy to change (playing-style wise) uci-engine in the clubplayer regions (1400 to 2200) would be great