Honey-XPro Release

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

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
MikeB
Posts: 3289
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Honey-XPro Release

Post by MikeB » Sun May 26, 2019 5:52 am

The maiden release of Honey-XPro, a UCI chess playing engine derived from McCain (MichaelB7), Stockfish, SugaR (Marco Zerbinati) and Glaurung 2.1

Honey-Xpro is named in honor of my paternal grandmother, Honey (née Sullivan) Byrne and she was, without a doubt, the sweetest person I have ever known. So I hope you find this chess engine as sweet as she was. Her grandfather, John J Sullivan , immigrated to the United States from Ireland around 1860 when he was about 20 years old, settling in as coal miner in the Scranton PA area. At the age of 40, he graduated from University of Maryland Medical School as an MD, returning to Olyphant PA to practice medicine for the next 47 years, until his death in 1927. HIs son and Honey's father , John J Sullivan, Jr., who also became a doctor and graduated from the University of Pennsylvania Medical School at the turn of the century.

Honey-XPro has a new score output option , where the user can elect to have the score outputed in a expected score percentage. When using in chess GUI such as xBoard, you must also check the xBoard option , otherwise the score percentages will be displayed as decimals. This was derived from reversed engineering Lc0 conversion function of win percentage to centipawns and then tailoring it specifically to Honey's centipawns evaluation. It is strictly a conversion, the engine itself is not using the score percentage. See uci.cpp for the formula - it could easily be transferable to any chess engine.

Details, source, binaries and books (scroll to the bottom for binaries and books)
https://github.com/MichaelB7/Stockfish/ ... s/tag/XPro

Below were games played today in the monthly HGM tourney using the scoring % output ( with 0% indicates losing without a doubt, 50.00 indicates a drawish position and 100% indicates certain win) ( scores like 1000.05 - is xboard output for mate-in-5). The most interesting game played was against LeelaRB, 5th game down. The McBrain name below is just my user name - it was Honey-XPro playing.



I hope you are able to enjoy using Honey-XPro as much as I enjoyed putting it together.

Modern Times
Posts: 2370
Joined: Thu Jun 07, 2012 9:02 pm

Re: Honey-XPro Release

Post by Modern Times » Sun May 26, 2019 6:37 am

MikeB wrote:
Sun May 26, 2019 5:52 am
The most interesting game played was against LeelaRB, 5th game down.
36. Bd3 seems to have been a blunder by Lc0. When you analyse with Komodo, prior to that move Komodo saw 0.00. After that move, -14

Probably with more time and a better GPU, Lc0 would have chosen differently.



Edit: Just analysed this with Lc0 T6.532. On the GTX 1050 it took 22 seconds for it to move away from Bd3 and favour a different move. So that explains it - in the match there wasn't enough time for it to make a better choice. On a fast RTX, that would probably have taken 2 seconds !!
.

Opinions expressed here are my own, and not necessarily those of the CCRL Group.

MikeB
Posts: 3289
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: Honey-XPro Release

Post by MikeB » Sun May 26, 2019 6:48 am

Modern Times wrote:
Sun May 26, 2019 6:37 am
MikeB wrote:
Sun May 26, 2019 5:52 am
The most interesting game played was against LeelaRB, 5th game down.
36. Bd3 seems to have been a blunder by Lc0. When you analyse with Komodo, prior to that move Komodo saw 0.00. After that move, -14

Probably with more time and a better GPU, Lc0 would have chosen differently.

It was a blunder alright, Honey went from a scoring percentage of 49.46% to 99.39% - a game that was nearly drawn at that point, became dead lost. Interestingly, after 36 Bd3, 39. ...Qxf5!! wins outright as well, but not quite as fast. It is mate-in-2 if one mistakenly plays 40. exf5 with a pretty Bh2 , any K move then mate with Ng3#.
Last edited by MikeB on Sun May 26, 2019 7:01 am, edited 1 time in total.

Modern Times
Posts: 2370
Joined: Thu Jun 07, 2012 9:02 pm

Re: Honey-XPro Release

Post by Modern Times » Sun May 26, 2019 6:58 am

MikeB wrote:
Sun May 26, 2019 6:48 am

It was a blunder alright, Honey went from a scoring percentage of 49.46% to 99.39% - a game that was nearly drawn at that point, became dead lost.
See the edit above I just analysed this with Lc0 T6.532. On the GTX 1050 it took 22 seconds for it to move away from Bd3 and favour a different move. So that explains it - in the match there wasn't enough time for it to make a better choice. On a fast RTX, that would probably have taken 2 seconds. So a combination of the budget GPU and the blitz time control. But it still made second place in the tournament.
.

Opinions expressed here are my own, and not necessarily those of the CCRL Group.

MikeB
Posts: 3289
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: Honey-XPro Release

Post by MikeB » Sun May 26, 2019 7:05 am

Modern Times wrote:
Sun May 26, 2019 6:58 am
MikeB wrote:
Sun May 26, 2019 6:48 am

It was a blunder alright, Honey went from a scoring percentage of 49.46% to 99.39% - a game that was nearly drawn at that point, became dead lost.
See the edit above I just analysed this with Lc0 T6.532. On the GTX 1050 it took 22 seconds for it to move away from Bd3 and favour a different move. So that explains it - in the match there wasn't enough time for it to make a better choice. On a fast RTX, that would probably have taken 2 seconds. So a combination of the budget GPU and the blitz time control. But it still made second place in the tournament.
Also after Bd3 , Qxf5 wins out right as well - see my edit above with cute mate-in-2 if one plays exf5.

User avatar
Ovyron
Posts: 2347
Joined: Tue Jul 03, 2007 2:30 am

Re: Honey-XPro Release

Post by Ovyron » Sun May 26, 2019 9:26 pm

Thanks Mike! I've been relying on McBrain/McCain in MultiPV for my correspondence chess analysis, and giving opponents hundreds of points stronger than me a run for their money on the ICCF, so great results, looks like Honey is the engine to use :mrgreen:

I have read that Honey is partly based on SuGaR, which I've been using under Chess Openings Wizard because it seems to detect automatically how many cores I have and uses them all by default, while Honey does like most other SF derivatives and sticks to 1cpu by default, I was wondering what's up with that (or if SuGaR uses 4 by default and they happen to match my number of cores.)
Make someone happy today.

MikeB
Posts: 3289
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: Honey-XPro Release

Post by MikeB » Sun May 26, 2019 11:17 pm

Ovyron wrote:
Sun May 26, 2019 9:26 pm
Thanks Mike! I've been relying on McBrain/McCain in MultiPV for my correspondence chess analysis, and giving opponents hundreds of points stronger than me a run for their money on the ICCF, so great results, looks like Honey is the engine to use :mrgreen:

I have read that Honey is partly based on SuGaR, which I've been using under Chess Openings Wizard because it seems to detect automatically how many cores I have and uses them all by default, while Honey does like most other SF derivatives and sticks to 1cpu by default, I was wondering what's up with that (or if SuGaR uses 4 by default and they happen to match my number of cores.)
Hi - good to hear about your successes - regarding cores , I didn’t bring code that over since most of my testing is with one core - but I believe most GUI’s let you make persistent the CPU setting for the number of cores or threads you want to use on a regular basis.

User avatar
Ovyron
Posts: 2347
Joined: Tue Jul 03, 2007 2:30 am

Re: Honey-XPro Release

Post by Ovyron » Sun May 26, 2019 11:30 pm

MikeB wrote:
Sun May 26, 2019 11:17 pm
I believe most GUI’s let you make persistent the CPU setting for the number of cores or threads you want to use on a regular basis.
Chess Openings Wizard doesn't, we're lucky it runs UCI engines at all! :mrgreen: But anyway, this is only about automatic analysis of leaf nodes in a Tree, SuGaR is sufficient for that (ideally, C.O.W. would use the engine at 1CPU, one engine instance per core, anyway, so this is clearly poor GUI design).

I wish the best with Honey :)
Make someone happy today.

WuShock
Posts: 166
Joined: Thu Jul 19, 2007 1:13 am

Re: Honey-XPro Release

Post by WuShock » Mon May 27, 2019 11:37 am

I have it running fine on F15 , but curious about Book #1 . There are boxes to enter pathway for Books #2 and #3 , but not for #1 ?

Thanks for all your efforts .....

Tom G

komodoslash001
Posts: 10
Joined: Fri Feb 01, 2019 10:56 am
Full name: Jeremy Tyrania

Re: Honey-XPro Release

Post by komodoslash001 » Mon May 27, 2019 2:11 pm

is it possible to add support for large pages? I tried to but it seems a bit different than adding it for stockfish i got a bunch of errors. Would be great! :) And cool engine, being able to use 3 books is awesome.

Post Reply