dragon 4.2

Discussion about development of draughts in the time of computer and Internet.
Post Reply
MichelG
Posts: 244
Joined: Sun Dec 28, 2003 20:24
Contact:

dragon 4.2

Post by MichelG » Tue Jul 16, 2013 20:00

I am happy to announce dragon version 4.2

This release improves playing strength, improves the interface and addresses several bugs.

program improvements:
* new board graphics
* dragon now automatically saves and reloads the game when you restart
* new way to edit its options
* open game now has a preview option
* button to toggle between table and text view of the game
* some bugfixes for saving and opening pdn files

engine improvements:
* increased performance: dragon is now (finally) fully bitboard based. This gives about 18% more evaluations per second. Also uses popcount instruction when available (+5% speed)
* dragon has a new opening book of about 150.000 positions (up from 26.000). The book has been automatically generated.
* more effective transposition tables


http://mdgsoft.home.xs4all.nl/draughts/ ... v4.2.0.msi

Michel

Rein Halbersma
Posts: 1722
Joined: Wed Apr 14, 2004 16:04
Contact:

Re: dragon 4.2

Post by Rein Halbersma » Tue Jul 16, 2013 22:50

MichelG wrote:I am happy to announce dragon version 4.2

This release improves playing strength, improves the interface and addresses several bugs.

program improvements:
* new board graphics
* dragon now automatically saves and reloads the game when you restart
* new way to edit its options
* open game now has a preview option
* button to toggle between table and text view of the game
* some bugfixes for saving and opening pdn files

engine improvements:
* increased performance: dragon is now (finally) fully bitboard based. This gives about 18% more evaluations per second. Also uses popcount instruction when available (+5% speed)
* dragon has a new opening book of about 150.000 positions (up from 26.000). The book has been automatically generated.
* more effective transposition tables


http://mdgsoft.home.xs4all.nl/draughts/ ... v4.2.0.msi

Michel
Looks very nice, Michel! Any plans to include the Killer engine again? I really like that feature!

Catherine
Posts: 129
Joined: Tue Aug 14, 2012 22:24
Real name: Catherine Bourneuf

Re: dragon 4.2

Post by Catherine » Thu Jul 18, 2013 15:17

Thank very much Michel for this new version of your wonderful Dragon. Thank again, I will run a dxp match against Truus to test the new strenght. God bless you.

Ed Gilbert
Posts: 859
Joined: Sat Apr 28, 2007 14:53
Real name: Ed Gilbert
Location: Morristown, NJ USA
Contact:

Re: dragon 4.2

Post by Ed Gilbert » Fri Jul 19, 2013 19:29

As I was testing an updated version of kingsrow for release, I ran a dxp match of 3-move openings against the new dragon. Dragon has been getting stronger with the last few releases, and though it didn't come out ahead in this match, it was a lot closer than the matches from a year ago.

Kingsrow 1.54 vs Dragon (64 bit, standaard): 130 wins, 40 losses, 804 draws, 14 unknowns

Common settings for both engines:

time control: 75 moves in 2 minutes
hashtables: 256 mb
search threads: 4
opening books: on
pondering: off
endgame db: 6 pieces

As I recall dragon was ahead after about 60 games, which again shows that very many games are needed to average out the randomness when the engines are not that far apart in strength.

-- Ed
Attachments
kr vs dragon 7-19-2013.zip
Zip file with games
(1.12 MiB) Downloaded 221 times

MichelG
Posts: 244
Joined: Sun Dec 28, 2003 20:24
Contact:

Re: dragon 4.2

Post by MichelG » Mon Jul 22, 2013 12:37

Ed Gilbert wrote:As I was testing an updated version of kingsrow for release, I ran a dxp match of 3-move openings against the new dragon. Dragon has been getting stronger with the last few releases, and though it didn't come out ahead in this match, it was a lot closer than the matches from a year ago.

Kingsrow 1.54 vs Dragon (64 bit, standaard): 130 wins, 40 losses, 804 draws, 14 unknowns
As I recall dragon was ahead after about 60 games, which again shows that very many games are needed to average out the randomness when the engines are not that far apart in strength.

-- Ed
Thanks for publishing this result. The fairly large number of losses for dragon suprised me, and it appears there is a bug in the handling of dragon's new opening book. When it is playing white, dragon loses many games straight out of the book. I suspect it is selecting the worst instead of the best opening moves in some cases...

With dragon playing black, the score seems more balanced: 44 wins, 24 losses, 418 draws and 7 unkowns, from kingrows perspective.

I'll try to fix this as soon as possible.

Michel

Catherine
Posts: 129
Joined: Tue Aug 14, 2012 22:24
Real name: Catherine Bourneuf

Re: dragon 4.2

Post by Catherine » Tue Jul 23, 2013 23:05

Hi Michel, just a question about the opening book of Dragon 4.2.
What is the difference beetwen the old and the new Dragon opening book?
on what critera the new opening book, automatically generated, is based on? Thank
Catherine.

MichelG
Posts: 244
Joined: Sun Dec 28, 2003 20:24
Contact:

Re: dragon 4.2

Post by MichelG » Wed Jul 24, 2013 09:22

Catherine wrote:Hi Michel, just a question about the opening book of Dragon 4.2.
What is the difference beetwen the old and the new Dragon opening book?
on what critera the new opening book, automatically generated, is based on? Thank
Catherine.
The old book was based on human games. The new book is generated by automatically investigating the best moves (it's called the DOE algorithm)

There was a bug in the v4.2.0 opening book, causing dragon to lose from the book.

At new version is now available from:
http://mdgsoft.home.xs4all.nl/draughts/install.html

Michel

Catherine
Posts: 129
Joined: Tue Aug 14, 2012 22:24
Real name: Catherine Bourneuf

Re: dragon 4.2

Post by Catherine » Wed Jul 24, 2013 10:15

Thank Michel for fixing the bug in The opening book of the new version 4.2.1. I runed a short games and see that the opening is better. I was the hope to see like you promised to solve the commodity of auto move 35 - 30. thank.

Post Reply