Internet engine matches

Discussion about development of draughts in the time of computer and Internet.
Post Reply
TAILLE
Posts: 968
Joined: Thu Apr 26, 2007 18:51
Location: FRANCE

Re: Internet engine matches

Post by TAILLE » Thu May 10, 2012 11:32

Hi Bert,
BertTuyt wrote:Gerard,

any news regarding (thematic) Damy matches against KingsRow.
I play nowadays mini -matches (first 30 ballots) and i aim to play ( at least) 30 draws.
Last match 1 lost game, but it gave a clue :) for improvement, and thats the good news !!

Bert
Oops are you really satisfied by a long serie of draws? Why not try and increase the number of wins instead of limiting the number of losses?
My appoach is the following:
1) each time I see a Damy win I analyse if Damy anticipate correctly this win and I try to improve this early advantage detection
2) each time I see a Kingsrow win I try to add the corresponding winning strategy to Damy's program. As a consequence Damy will not only be able to avoid similar loss but it will be able to apply a new winning strategy.
I do not consider a draw as a good news. Let's imagine a 30 games match between Damage and Kingsrow and taking the point of view of Damage wich result do you prefer:
1) 10 wins, 20 losses
2) 30 draws
As far as I am concerned I prefer the first scenario because in this scenario I know Damage would have the potential to win a lot of games. The second scenario seems very poor for potential improvment isn't it?
Gérard

BertTuyt
Posts: 1377
Joined: Wed Sep 01, 2004 19:42

Re: Internet engine matches

Post by BertTuyt » Thu May 10, 2012 17:52

Gerard, im not satisfied with a long series of draws. :wink:
But im also realistic.

Before you can win you should at least draw.
So thats what I do, improving the program, so that the latest short match was 29 Draw , 1 Lost game.
Of course the next step is finding ways to win.

Anyway that still leaves my question unanswered, how far you are with thematic matches, and/or real matches, is it already possible to provide a WDL (Win/Draw/Lose) answer?

Bert

TAILLE
Posts: 968
Joined: Thu Apr 26, 2007 18:51
Location: FRANCE

Re: Internet engine matches

Post by TAILLE » Thu May 10, 2012 19:07

Hi Bert,
BertTuyt wrote:Gerard, im not satisfied with a long series of draws. :wink:
But im also realistic.

Before you can win you should at least draw.
So thats what I do, improving the program, so that the latest short match was 29 Draw , 1 Lost game.
Of course the next step is finding ways to win.

Anyway that still leaves my question unanswered, how far you are with thematic matches, and/or real matches, is it already possible to provide a WDL (Win/Draw/Lose) answer?

Bert
Image

My last match were played the 7/4 with the following result
For Damy point of view : 13 wins, 17 losses, 116 draws (146 games)
I am not quite satisfied with only 13 wins because Damy missed some other good opportunities. Anyway the 17 Kingsrow wins showed me some other winning strategy I have to take into account. For that I have first to restructure a little my eval function of positions with one king and I have planned another match end of May.
I keep you posted
Gérard

BertTuyt
Posts: 1377
Joined: Wed Sep 01, 2004 19:42

Re: Internet engine matches

Post by BertTuyt » Thu May 10, 2012 19:42

Gerard for my understanding, there a 9 * 9 = 81 games possible in your example, * 2 for the color to move = 162.
So a full 2-move ballot match is 162
So i do not fully understand the 146?

Bert

TAILLE
Posts: 968
Joined: Thu Apr 26, 2007 18:51
Location: FRANCE

Re: Internet engine matches

Post by TAILLE » Thu May 10, 2012 22:52

Bert,
BertTuyt wrote:Gerard for my understanding, there a 9 * 9 = 81 games possible in your example, * 2 for the color to move = 162.
So a full 2-move ballot match is 162
So i do not fully understand the 146?

Bert
Yes Bert you are right. I just have to face a Damy crash during the 147th game. Now this bug I encountered is fixed.
Did you yourself run a similar match against Kingsrow, starting with the same position?
Gérard

BertTuyt
Posts: 1377
Joined: Wed Sep 01, 2004 19:42

Re: Internet engine matches

Post by BertTuyt » Thu May 10, 2012 23:06

Gerard, no i didn't start a match with the position you used.

So far the 2-ballot games are hard-coded in my program.
So i need to change something in my code to make this possible..
Anyway, it is interesting, maybe i might find some time during the weekend.
But definitely I will put it on my "things to do list"..

Bert

BertTuyt
Posts: 1377
Joined: Wed Sep 01, 2004 19:42

Re: Internet engine matches

Post by BertTuyt » Sat May 12, 2012 18:29

I was able to quickly modify code, so i could also start a match with a non-standard position.
The stats so far (from the perspective of Damage) and base position as used by Gerard.

Damage - KingsRow: 50 games, 2 Win, 3 Lost, 43 Draws, 2 Unknowns (so i need to have look at the unknowns to provide a better score).

Bert

BertTuyt
Posts: 1377
Joined: Wed Sep 01, 2004 19:42

Re: Internet engine matches

Post by BertTuyt » Sun May 13, 2012 15:19

Unfortunately i still had ( apparently) some implementation errors, as the match stopped after 158 games (which is the number for the match with the initial position).
Anyway here the 158 games stats from the perspective of Damage.

Damage - KingsRow 12 Win, 10 Lost, 136 Draw.

Bert

TAILLE
Posts: 968
Joined: Thu Apr 26, 2007 18:51
Location: FRANCE

Re: Internet engine matches

Post by TAILLE » Sun May 13, 2012 15:42

Hi Bert,
BertTuyt wrote:Unfortunately i still had ( apparently) some implementation errors, as the match stopped after 158 games (which is the number for the match with the initial position).
Anyway here the 158 games stats from the perspective of Damage.

Damage - KingsRow 12 Win, 10 Lost, 136 Draw.

Bert
For my previous match Damy-Kingsrow (13 wins, 17 losses, 116 draws) 5 wins were made with white and 25 wins were made with black.
Could you give us your results from white/black point of view?
Gérard

BertTuyt
Posts: 1377
Joined: Wed Sep 01, 2004 19:42

Re: Internet engine matches

Post by BertTuyt » Sun May 13, 2012 17:02

Gerard in the match Damage played Kingsrow, there were 10 + 12 non-draw games.
8 White Wins and 14 Black Wins.

Bert

jj
Posts: 155
Joined: Sun Sep 13, 2009 23:33
Real name: Jan-Jaap van Horssen
Location: Zeist, Netherlands

Re: Internet engine matches

Post by jj » Sun May 13, 2012 17:03

TAILLE wrote:Image
Hi all,

Interesting experiment! My results, from Maximus' point of view:

1. Maximus - Kingsrow(6p): win 19, loss 6, draw 137 (7 white wins, 18 black wins)
2. Maximus - Kingsrow(9p): win 10, loss 4, draw 148 (4 white wins, 10 black wins)

Both matches played with 3 cores each and pondering on my 6-core (hyperthreading switched off), tempo 5min/75moves. In match 2, Kingsrow was given an egdb cache of 16000 mb.

Gérard, did you already play 158-game ballot matches from the initial position? And what can you tell us about the time-to-depth speedup of the search on your 12-core? Do you use 12 threads or can you use 24?

Jan-Jaap
www.maximusdraughts.org

Krzychumag
Posts: 145
Joined: Tue Sep 01, 2009 17:31
Real name: Krzysztof Grzelak

Re: Internet engine matches

Post by Krzychumag » Sun May 13, 2012 17:48

BertTuyt wrote:Unfortunately i still had ( apparently) some implementation errors, as the match stopped after 158 games (which is the number for the match with the initial position).
Anyway here the 158 games stats from the perspective of Damage.

Damage - KingsRow 12 Win, 10 Lost, 136 Draw.

Bert
Bert then programme Damage lost game from KingsRow, or he won with programme KingsRow. Gerard if you can what equipment write on you are working above your program.

TAILLE
Posts: 968
Joined: Thu Apr 26, 2007 18:51
Location: FRANCE

Re: Internet engine matches

Post by TAILLE » Sun May 13, 2012 18:03

Hi Jan-Japp,
jj wrote:
TAILLE wrote:Image
Hi all,

Interesting experiment! My results, from Maximus' point of view:

1. Maximus - Kingsrow(6p): win 19, loss 6, draw 137 (7 white wins, 18 black wins)
2. Maximus - Kingsrow(9p): win 10, loss 4, draw 148 (4 white wins, 10 black wins)

Both matches played with 3 cores each and pondering on my 6-core (hyperthreading switched off), tempo 5min/75moves. In match 2, Kingsrow was given an egdb cache of 16000 mb.

Gérard, did you already play 158-game ballot matches from the initial position? And what can you tell us about the time-to-depth speedup of the search on your 12-core? Do you use 12 threads or can you use 24?

Jan-Jaap
BTW all matches I played against Kingsrow were against Kingsrow 1.52a 9p.

Your results against Kingsrow are a little surprising because you suffered 6 losses against Kingsrow(6p) but only 4 against Kingsrow(9p). It looks like Kingsrow uses its 9p egdb to avoid loss but with the drawback of decreasing the number of win. Strange isn't it? Maybe Ed. has an explanation for that.

I am not surprised to see Damy obtained more wins (but also more losses !) because I concentrated my program on winning strategies, avoiding "quite" draw strategies. I am quite sure to be able to increase greatly my score but I have to be patient.

No Jan-Jaap Damy is not yet able to play a match from the initial position due to still large holes in my new eval function.
Concerning the number of cores I use, the current Damy version is not able to use more than 8 cores. I have to work a little on my exclusion function to use 12 cores or more.
Gérard

BertTuyt
Posts: 1377
Joined: Wed Sep 01, 2004 19:42

Re: Internet engine matches

Post by BertTuyt » Sun May 13, 2012 19:33

Bert then programme Damage lost game from KingsRow, or he won with programme KingsRow. Gerard if you can what equipment write on you are working above your program.
Damage won the Match, although the margin was not significant.
Also i had switched off pondering (both programs), and also both programs only used 1 core.

Bert

BertTuyt
Posts: 1377
Joined: Wed Sep 01, 2004 19:42

Re: Internet engine matches

Post by BertTuyt » Sun May 13, 2012 19:43

Another interesting observation, but i dont have a clue how to explain it.
The 10 wins by Kingsrow was 2 times with white and 8 times with black.
The Damage wins was 12 times with black.

Bert

Post Reply