NNUE

Discussion about development of draughts in the time of computer and Internet.
Post Reply
BertTuyt
Posts: 1511
Joined: Wed Sep 01, 2004 19:42

Re: NNUE

Post by BertTuyt » Wed Feb 10, 2021 10:38

Sidiki,

I think the situation with International 10x10 Draughts is different, compared with Chess.
First of all with Draughts we might be already quite close to the Draw Black-Hole, with excellent programs as Scan, and Kingsrow.
So with NNUE we will most likely not surpass the current pattern based programs (at least that is my expectation), it would be already a huge achievement if we reach on-par level.

But we might definitely grow into that situation, although there is much work to do, both in HW as in SW.
Current standing is that I was able to reduce the gap with Scan (Scan nnue implementation) towards around 10 ELO (as posted and shared in this forum), so step by step......

Bert

Krzysztof Grzelak
Posts: 1028
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Re: NNUE

Post by Krzysztof Grzelak » Wed Feb 10, 2021 13:50

Looking at all this, you have to consider which method is better. Of course, NNUE is a constant game of changing parameters in the program.

BertTuyt
Posts: 1511
Joined: Wed Sep 01, 2004 19:42

Re: NNUE

Post by BertTuyt » Wed Feb 10, 2021 16:46

My 5 cents...
If you want to write the best program , then stick to the patterns based solution as implemented in Scan and Kingsrow.
If one want to embark on the new developments in neural networks (both SW as HW), and use draughts as a test environment, then try nnue.

Bert

Krzysztof Grzelak
Posts: 1028
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Re: NNUE

Post by Krzysztof Grzelak » Wed Feb 10, 2021 17:06

BertTuyt wrote:
Wed Feb 10, 2021 16:46
My 5 cents...
If you want to write the best program , then stick to the patterns based solution as implemented in Scan and Kingsrow.
If one want to embark on the new developments in neural networks (both SW as HW), and use draughts as a test environment, then try nnue.

Bert
I am only giving some thought to one thing - and how somebody will invent something better than Ed and Fabian. Programmers should go their own way, and not be based on other programs. In the end they have a head and a mind.

Sidiki
Posts: 203
Joined: Thu Jan 15, 2015 16:28
Real name: Coulibaly Sidiki

Re: NNUE

Post by Sidiki » Fri Feb 12, 2021 11:10

BertTuyt wrote:
Wed Feb 10, 2021 10:38
Sidiki,

I think the situation with International 10x10 Draughts is different, compared with Chess.
First of all with Draughts we might be already quite close to the Draw Black-Hole, with excellent programs as Scan, and Kingsrow.
So with NNUE we will most likely not surpass the current pattern based programs (at least that is my expectation), it would be already a huge achievement if we reach on-par level.

But we might definitely grow into that situation, although there is much work to do, both in HW as in SW.
Current standing is that I was able to reduce the gap with Scan (Scan nnue implementation) towards around 10 ELO (as posted and shared in this forum), so step by step......

Bert
Thanks Bert,

As you said we already reached a high level with patterns, and it's true.

We hope that, steps by steps, as you said, we will find something greater.

Thanks again for all you and the others done and doing for this community.

God bless you.

Sidiki.

Sidiki
Posts: 203
Joined: Thu Jan 15, 2015 16:28
Real name: Coulibaly Sidiki

Re: NNUE

Post by Sidiki » Fri Feb 12, 2021 11:18

Krzysztof Grzelak wrote:
Wed Feb 10, 2021 17:06
BertTuyt wrote:
Wed Feb 10, 2021 16:46
My 5 cents...
If you want to write the best program , then stick to the patterns based solution as implemented in Scan and Kingsrow.
If one want to embark on the new developments in neural networks (both SW as HW), and use draughts as a test environment, then try nnue.

Bert
I am only giving some thought to one thing - and how somebody will invent something better than Ed and Fabian. Programmers should go their own way, and not be based on other programs. In the end they have a head and a mind.
Hi Krzysztof,

In year 1990, we never thought that another program will be stronger than Truus, and then came Buggy, Damy, Kingsrow, Scan without forget Damage , Dragon, Maximus.

All these stronger programs are being innovated, creation not really exist, we perfom something that already exist to have something else much better.

Krzysztof Grzelak
Posts: 1028
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Re: NNUE

Post by Krzysztof Grzelak » Fri Feb 12, 2021 13:31

Sidiki wrote:
Fri Feb 12, 2021 11:18
Hi Krzysztof,

In year 1990, we never thought that another program will be stronger than Truus, and then came Buggy, Damy, Kingsrow, Scan without forget Damage , Dragon, Maximus.

All these stronger programs are being innovated, creation not really exist, we perfom something that already exist to have something else much better.
This is your wrong thinking. Each programmer should have their own code and not use any other code.

Sidiki
Posts: 203
Joined: Thu Jan 15, 2015 16:28
Real name: Coulibaly Sidiki

Re: NNUE

Post by Sidiki » Fri Feb 12, 2021 13:52

Krzysztof Grzelak wrote:
Fri Feb 12, 2021 13:31
Sidiki wrote:
Fri Feb 12, 2021 11:18
Hi Krzysztof,

In year 1990, we never thought that another program will be stronger than Truus, and then came Buggy, Damy, Kingsrow, Scan without forget Damage , Dragon, Maximus.

All these stronger programs are being innovated, creation not really exist, we perfom something that already exist to have something else much better.
This is your wrong thinking. Each programmer should have their own code and not use any other code.
Who spoken of copying idea or plagia code? each program has his own eval and others things that make his style of playing: Kingsrow, Scan, Damage, Maximus, Ares, Dragon, Truus, Flits and Mobydam haven't the same playing style. But all of them are based on, by caterogy, on a kind of algorithm, so CODE.

What you want to say, and i see that you don't love to wait, it's that :"itsn't possible to surpass the level of Kingsrow, Scan or Damage without reproduce exactly the source code of these programs.

This it's wrong, because, you don't know anything about programmation.

Krzysztof Grzelak
Posts: 1028
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Re: NNUE

Post by Krzysztof Grzelak » Fri Feb 12, 2021 14:10

Sidiki wrote:
Fri Feb 12, 2021 13:52
Who spoken of copying idea or plagia code? each program has his own eval and others things that make his style of playing: Kingsrow, Scan, Damage, Maximus, Ares, Dragon, Truus, Flits and Mobydam haven't the same playing style. But all of them are based on, by caterogy, on a kind of algorithm, so CODE.

What you want to say, and i see that you don't love to wait, it's that :"itsn't possible to surpass the level of Kingsrow, Scan or Damage without reproduce exactly the source code of these programs.

This it's wrong, because, you don't know anything about programmation.
We don't understand each other. I don't need to know programming, and neither do other people. I believe that a programmer should write a program without looking at other programs. Unfortunately, this is not the case these days.

Sidiki
Posts: 203
Joined: Thu Jan 15, 2015 16:28
Real name: Coulibaly Sidiki

Re: NNUE

Post by Sidiki » Sat Feb 13, 2021 09:23

Krzysztof Grzelak wrote:
Fri Feb 12, 2021 14:10
Sidiki wrote:
Fri Feb 12, 2021 13:52
Who spoken of copying idea or plagia code? each program has his own eval and others things that make his style of playing: Kingsrow, Scan, Damage, Maximus, Ares, Dragon, Truus, Flits and Mobydam haven't the same playing style. But all of them are based on, by caterogy, on a kind of algorithm, so CODE.

What you want to say, and i see that you don't love to wait, it's that :"itsn't possible to surpass the level of Kingsrow, Scan or Damage without reproduce exactly the source code of these programs.

This it's wrong, because, you don't know anything about programmation.
We don't understand each other. I don't need to know programming, and neither do other people. I believe that a programmer should write a program without looking at other programs. Unfortunately, this is not the case these days.
This mean that this programmer must come from another world,

Because in the life, not only in programmation, we create anything by looking for an already existing thing or concept. I'm sure that in a near future we will gain again in strenght with the program. As i said yesterday, who will believe that the great Truus will be crushed today as a baby by Scan or Kingsrow?

Sidki

Krzysztof Grzelak
Posts: 1028
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Re: NNUE

Post by Krzysztof Grzelak » Sat Feb 13, 2021 14:56

Sidiki wrote:
Sat Feb 13, 2021 09:23
This mean that this programmer must come from another world,
There is 2015 Fabien Letouzey is writing the wonderful program. The program is called Scan. Scan has own beautiful ideas and has an opened code. He is playing the program wonderfully and is getting the top one spots at the Olympics. Whether Fabien Letouzey comes from other world - doesn't come from the earth.

Sidiki
Posts: 203
Joined: Thu Jan 15, 2015 16:28
Real name: Coulibaly Sidiki

Re: NNUE

Post by Sidiki » Sun Feb 14, 2021 11:35

Krzysztof Grzelak wrote:
Sat Feb 13, 2021 14:56
Sidiki wrote:
Sat Feb 13, 2021 09:23
This mean that this programmer must come from another world,
There is 2015 Fabien Letouzey is writing the wonderful program. The program is called Scan. Scan has own beautiful ideas and has an opened code. He is playing the program wonderfully and is getting the top one spots at the Olympics. Whether Fabien Letouzey comes from other world - doesn't come from the earth.
Krzysztof,

You don't understood, or yourself don't know what about you are speaking.

You are telling to us that you are tired to see programs based on existing concept, so algorithm or code if you want.

I asked to Fabien about pattern, and he said and it's true, that this concept was taken from hotello, Michel of Dragon already used this concept.

What you need to know it's that in all thing in this world, anything you will "create" was inspired by you from an existing thing.

If we continue to explain it to you we risk to finish to remote to GOD, at the beginning of the world.

Sidiki.

Krzysztof Grzelak
Posts: 1028
Joined: Thu Jun 20, 2013 17:16
Real name: Krzysztof Grzelak

Re: NNUE

Post by Krzysztof Grzelak » Sun Feb 14, 2021 19:58

Sidiki wrote:
Sun Feb 14, 2021 11:35
Krzysztof,

You don't understood, or yourself don't know what about you are speaking.

You are telling to us that you are tired to see programs based on existing concept, so algorithm or code if you want.

I asked to Fabien about pattern, and he said and it's true, that this concept was taken from hotello, Michel of Dragon already used this concept.

What you need to know it's that in all thing in this world, anything you will "create" was inspired by you from an existing thing.

If we continue to explain it to you we risk to finish to remote to GOD, at the beginning of the world.

Sidiki.
Sorry, we don't understand each other.

BertTuyt
Posts: 1511
Joined: Wed Sep 01, 2004 19:42

Re: NNUE

Post by BertTuyt » Mon Mar 29, 2021 19:03

Most recent match with tmgr (the tool developed by Ed).
Conditions Scan_31nnue vs Scan_31, played on an Intel Core i7 8700K, 2-move start positions, TC 75 moves in 1 minute, books off, 6-piece dbs, 1 search thread.

Code: Select all

Match stats Scan 3.1 nnue vs. Scan 3.1

[ 1]: 0.491 score,   632 games,    2 wins,   13 losses,   617 draws,   0 unk
[ 2]: 0.491 score,   632 games,    3 wins,   14 losses,   612 draws,   3 unk
[ 3]: 0.496 score,   632 games,    2 wins,    7 losses,   621 draws,   2 unk
[ 4]: 0.495 score,   632 games,    4 wins,   10 losses,   616 draws,   2 unk
[ 5]: 0.490 score,   632 games,    2 wins,   14 losses,   615 draws,   1 unk
[ 6]: 0.498 score,   632 games,    5 wins,    8 losses,   618 draws,   1 unk
total 0.494 score,  3792 games,   18 wins,   66 losses,  3699 draws,   9 unk
elo diff -4.4
Bert

BertTuyt
Posts: 1511
Joined: Wed Sep 01, 2004 19:42

Re: NNUE

Post by BertTuyt » Sun Apr 04, 2021 10:09

Most recent match with tmgr (the tool developed by Ed).
Conditions Scan_31nnue vs Scan_31, played on an Intel Core i7 8700K, 2-move start positions, TC 75 moves in 2 minute, books off, 6-piece dbs, 1 search thread.

Code: Select all

[ 1]: 0.502 score,   632 games,    6 wins,    3 losses,   623 draws,   0 unk
[ 2]: 0.502 score,   632 games,    5 wins,    3 losses,   624 draws,   0 unk
[ 3]: 0.501 score,   632 games,    6 wins,    5 losses,   621 draws,   0 unk
[ 4]: 0.498 score,   632 games,    4 wins,    6 losses,   620 draws,   2 unk
[ 5]: 0.497 score,   632 games,    4 wins,    8 losses,   620 draws,   0 unk
[ 6]: 0.498 score,   632 games,    2 wins,    5 losses,   625 draws,   0 unk
total 0.500 score,  3792 games,   27 wins,   30 losses,  3733 draws,   2 unk
elo diff -0.3
Bert

Post Reply