New Houdini

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

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
User avatar
Laskos
Posts: 9960
Joined: Wed Jul 26, 2006 8:21 pm
Full name: Kai Laskos

Re: re: win adjudication/Re: New Houdini

Post by Laskos » Sun Oct 09, 2016 2:24 pm

Lyudmil Tsvetkov wrote:
Guenther wrote:
Lyudmil Tsvetkov wrote:
carldaman wrote:
Laskos wrote:
Number of wins which were adjudicated before reaching more than 1 pawn material advantage (mostly positional advantage):

Houdini: 13/34 (38%)
Stockfish: 7/30 (23%)
Komodo: 12/30 (40%)
A disappointing adjudication method is in use by TCEC, where the spectators don't get to see many of the wins played out properly - especially for a tournament that prides itself as being on display primarily "for entertainment" purposes.

CL
Why it says then: 650cps from both engines for 4 consecutive moves?
It means 1 pawn material advantage on the board not in the shown engine score.
How do you measure 1 point material advantage?
I might have 5 points material disadvantage and 11.50 points positional advantage, and still win.
This is not how it usually happens at the adjudication point. Material is 13359 here, by Scid.

User avatar
Laskos
Posts: 9960
Joined: Wed Jul 26, 2006 8:21 pm
Full name: Kai Laskos

Re: re: win adjudication/Re: New Houdini

Post by Laskos » Mon Oct 10, 2016 3:06 pm

Laskos wrote:
Lyudmil Tsvetkov wrote:
Guenther wrote:
Lyudmil Tsvetkov wrote:
carldaman wrote:
Laskos wrote:
Number of wins which were adjudicated before reaching more than 1 pawn material advantage (mostly positional advantage):

Houdini: 13/34 (38%)
Stockfish: 7/30 (23%)
Komodo: 12/30 (40%)
A disappointing adjudication method is in use by TCEC, where the spectators don't get to see many of the wins played out properly - especially for a tournament that prides itself as being on display primarily "for entertainment" purposes.

CL
Why it says then: 650cps from both engines for 4 consecutive moves?
It means 1 pawn material advantage on the board not in the shown engine score.
How do you measure 1 point material advantage?
I might have 5 points material disadvantage and 11.50 points positional advantage, and still win.
This is not how it usually happens at the adjudication point. Material is 13359 here, by Scid.
Another interesting thing - material in 13359 count at the adjudication of won games. Houdini has the highest, higher even than Stockfish. Seems punchy indeed.

Houdini: 30.0
Stockfish: 27.3
Komodo: 24.2

Dan Cooper
Posts: 184
Joined: Sun Nov 01, 2015 2:15 am

Re: New Houdini

Post by Dan Cooper » Mon Oct 10, 2016 3:19 pm

stavros wrote:i think the point is why houdart doesnt says the h5-h4 difference as he did it in the past with houdini versions before releases :) whats wrong? looks weird isn't it?
chess cold war espionage:P
This is Ordo run across the entire history of TCEC.

Code: Select all

   # PLAYER                  :  RATING  ERROR  POINTS  PLAYED   (%)  CFS(%)
   3 Stockfish 160716        :  3371.8   77.8    39.0      56    70      63
   7 Houdini 200716          :  3326.8   77.0    35.5      56    63      53
   9 Komodo 10.1             :  3314.4   77.7    34.5      56    62      57
  35 Houdini 4               :  3195.6   43.3   176.5     299    59      70

User avatar
Laskos
Posts: 9960
Joined: Wed Jul 26, 2006 8:21 pm
Full name: Kai Laskos

Re: re: win adjudication/Re: New Houdini

Post by Laskos » Tue Oct 11, 2016 3:43 pm

Laskos wrote:Another interesting thing - material in 13359 count at the adjudication of won games. Houdini has the highest, higher even than Stockfish. Seems punchy indeed.

Houdini: 30.0
Stockfish: 27.3
Komodo: 24.2
After about 3/4 of the games in TCEC Rapid, the following stats to have an impression on the new Houdini.

Houdini is the fastest winner (median length of the won game):

Houdini: 49 moves
Stockfish: 50 moves
Komodo: 51 moves

Houdini likes to keep material on the board for a win (average material on the board at win adjudication in 1,3,3,5,9):

Houdini: 29.1
Stockfish: 26.5
Komodo: 24.0

Houdini keeps the Queen longer to win (average number of Queens on the board at the win adjudication, significant result):

Houdini: 0.95
Stockfish: 0.75
Komodo: 0.67

User avatar
Houdini
Posts: 1471
Joined: Mon Mar 15, 2010 11:00 pm
Contact:

Re: re: win adjudication/Re: New Houdini

Post by Houdini » Tue Oct 11, 2016 6:30 pm

Laskos wrote:Houdini keeps the Queen longer to win (average number of Queens on the board at the win adjudication, significant result):

Houdini: 0.95
Stockfish: 0.75
Komodo: 0.67
Interesting because Houdini is playing without contempt in the TCEC tournament.
I'm not sure whether the numbers above are just a statistical fluke, or whether the (rather big) changes to the H5 evaluation function have produced this side-effect.

User avatar
Laskos
Posts: 9960
Joined: Wed Jul 26, 2006 8:21 pm
Full name: Kai Laskos

Re: re: win adjudication/Re: New Houdini

Post by Laskos » Tue Oct 11, 2016 8:08 pm

Houdini wrote:
Laskos wrote:Houdini keeps the Queen longer to win (average number of Queens on the board at the win adjudication, significant result):

Houdini: 0.95
Stockfish: 0.75
Komodo: 0.67
Interesting because Houdini is playing without contempt in the TCEC tournament.
I'm not sure whether the numbers above are just a statistical fluke, or whether the (rather big) changes to the H5 evaluation function have produced this side-effect.
This was the most significant result I got, but it still can be a fluke (10% or so fluke). Some more games (in fact Wins) are necessary, but I doubt the Superfinal will be very generous with decided games, so we must wait for the release of Houdini. My results seem to confirm the more artistic and often more penetrating impression some people get watching the games.

shrapnel
Posts: 1291
Joined: Fri Nov 02, 2012 8:43 am
Location: New Delhi, India

Re: re: win adjudication/Re: New Houdini

Post by shrapnel » Wed Oct 12, 2016 3:55 am

Houdini wrote: Interesting because Houdini is playing without contempt in the TCEC tournament.
H4 at Contempt 0 was an awesome Engine in terms of Defence.
Can't wait for the new one !
i7 5960X @ 4.1 Ghz, 64 GB G.Skill RipJaws RAM, Twin Asus ROG Strix OC 11 GB Geforce 2080 Tis

Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 10:41 am

Re: re: win adjudication/Re: New Houdini

Post by Lyudmil Tsvetkov » Wed Oct 12, 2016 6:32 am

Houdini wrote:
Laskos wrote:Houdini keeps the Queen longer to win (average number of Queens on the board at the win adjudication, significant result):

Houdini: 0.95
Stockfish: 0.75
Komodo: 0.67
Interesting because Houdini is playing without contempt in the TCEC tournament.
I'm not sure whether the numbers above are just a statistical fluke, or whether the (rather big) changes to the H5 evaluation function have produced this side-effect.
and what do the -5cps(stage 3) and -7cps(rapid) scores at drawing nodes/3-fold repetitions mean then?

User avatar
Houdini
Posts: 1471
Joined: Mon Mar 15, 2010 11:00 pm
Contact:

Re: re: win adjudication/Re: New Houdini

Post by Houdini » Wed Oct 12, 2016 10:57 am

Lyudmil Tsvetkov wrote:and what do the -5cps(stage 3) and -7cps(rapid) scores at drawing nodes/3-fold repetitions mean then?
That's a simple 3-fold draw score to avoid silly move repetitions. The score depends on the material that is still on the board. The 3-fold repetition score only operates in a narrow band around 0.00 and doesn't change whether queens or other pieces are kept on the board.

The real contempt used by Houdini and Komodo has the effect of modifying the piece values ("my pieces are worth slightly more than your pieces") so that exchanges are avoided unless they are really favorable. This contempt works all the time, not just in a very narrow band around 0.00.

Lyudmil Tsvetkov
Posts: 6052
Joined: Tue Jun 12, 2012 10:41 am

Re: re: win adjudication/Re: New Houdini

Post by Lyudmil Tsvetkov » Wed Oct 12, 2016 12:33 pm

Houdini wrote:
Lyudmil Tsvetkov wrote:and what do the -5cps(stage 3) and -7cps(rapid) scores at drawing nodes/3-fold repetitions mean then?
That's a simple 3-fold draw score to avoid silly move repetitions. The score depends on the material that is still on the board. The 3-fold repetition score only operates in a narrow band around 0.00 and doesn't change whether queens or other pieces are kept on the board.

The real contempt used by Houdini and Komodo has the effect of modifying the piece values ("my pieces are worth slightly more than your pieces") so that exchanges are avoided unless they are really favorable. This contempt works all the time, not just in a very narrow band around 0.00.
Is not this 'simple 3-fold draw score' a kind of contempt?

What people would really like to know is if this SDS (simple draw score) in any way affects H5 performance.

Would H5 with this SDS score higher than H5 (SDS=0) versus the standard pool of 20+ opponents used in most rating lists?
And, of course, weaker against the top 2?

Post Reply