ENGINE SPEED AND STRONG

[Joost Buijs]

Hi Joost.

I have a question for you. In the folder with the engine is a file called argusv106 bin, I understand that this is a book debut.

Hi Krzysztof,

No, these are the weights for the pattern evaluation, I don't have a book yet. Besides gaining some time a book for draughts does not seem very important.

Yesterday I started looking at the engine again because I wanted to improve the time-control and add a console mode to enable you to play games with it, while looking at the code I found another bug that was already there for weeks, after a fail-high at the root I increase beta in several steps and do a research while (score >= beta), after this I rechecked the condition (score >= beta) while it has to be (score > alpha), because of this the engine sometimes missed a better move. After fixing this the engine became stronger, I'm very happy that I found this, usually I overlook these things a hundred times.

The rest of this week I don't have much time for it but I will try to upload the latest version this weekend.

Joost

[Krzysztof Grzelak]

Thank you very much Joost.

[Catherine]

Hi Joost.

I have a question for you. In the folder with the engine is a file called argusv106 bin, I understand that this is a book debut.

Hi Krzysztof,

No, these are the weights for the pattern evaluation, I don't have a book yet. Besides gaining some time a book for draughts does not seem very important.

Yesterday I started looking at the engine again because I wanted to improve the time-control and add a console mode to enable you to play games with it, while looking at the code I found another bug that was already there for weeks, after a fail-high at the root I increase beta in several steps and do a research while (score >= beta), after this I rechecked the condition (score >= beta) while it has to be (score > alpha), because of this the engine sometimes missed a better move. After fixing this the engine became stronger, I'm very happy that I found this, usually I overlook these things a hundred times.

The rest of this week I don't have much time for it but I will try to upload the latest version this weekend.

Joost

Hi Joost,

Thank again to working on the engine.
Like i said an engine is like a paint, it never finish but it's also good like you are doing to permit to the others persons to see the progress, thank again. We are waiting for the latest version.

This question is for you and all the others programmers:
What are the elements that can do the "perfect engine" this wich can for example will never lose.
It's the hardware on what it's launched, search evaluation, speed of search?

Thank to all

Catherine.

[Joost Buijs]

Hi Joost.

I have a question for you. In the folder with the engine is a file called argusv106 bin, I understand that this is a book debut.

Hi Krzysztof,

No, these are the weights for the pattern evaluation, I don't have a book yet. Besides gaining some time a book for draughts does not seem very important.

Yesterday I started looking at the engine again because I wanted to improve the time-control and add a console mode to enable you to play games with it, while looking at the code I found another bug that was already there for weeks, after a fail-high at the root I increase beta in several steps and do a research while (score >= beta), after this I rechecked the condition (score >= beta) while it has to be (score > alpha), because of this the engine sometimes missed a better move. After fixing this the engine became stronger, I'm very happy that I found this, usually I overlook these things a hundred times.

The rest of this week I don't have much time for it but I will try to upload the latest version this weekend.

Joost

Hi Joost,

Thank again to working on the engine.
Like i said an engine is like a paint, it never finish but it's also good like you are doing to permit to the others persons to see the progress, thank again. We are waiting for the latest version.

This question is for you and all the others programmers:
What are the elements that can do the "perfect engine" this wich can for example will never lose.
It's the hardware on what it's launched, search evaluation, speed of search?

Thank to all

Catherine.

Hi Catherine,

A perfect engine is not possible for 10x10 Draughts because the game is too complex, maybe in the future when they build quantum computers it might be possible (I doubt they will ever exist in a usable form).

Like you said, engine strength is determined by the quality of the evaluation-function and the search-depth. The search-depth depends upon 3 things, the speed of the hardware, the efficiency of coding, and the efficiency of the algorithm itself.

Lately I get more and more the feeling that the current top-engines are already so strong that on fast hardware and with not a too short time per move all the games will be drawn, I guess it is possible to make a stronger engine but the only effect will be that it needs less time per move to reach the same strength. I have no idea how much room (Elo wise) there is still left, but time will tell.

I just ran a match of 100 games between the last version of my engine and Kingsrow with 1 minute and 90 moves for the whole game, the outcome was rather equal, 1 win, 1 loss, 96 draws and 2 unknowns (which also seem to be draws). I never tried a match with longer time but I expect that the draw rate will go up.

The engine I uploaded two days ago containes a few bugs and is about 35 Elo weaker, combine that with the slower version that you have to use on your hardware I expect it to be 60 Elo below Kingsrow, it can be that both engines behave differently on your hardware. I still have a core i3 over here with 2 cores at 3400 MHz., this weekend I will try what it does, I have to check the egdb code for compatibility anyway.

Joost

[Krzysztof Grzelak]

Joost and you can write on what hardware you play matches with another program and how you set up your program.

[Joost Buijs]

Joost and you can write on what hardware you play matches with another program and how you set up your program.

Krzysztof,

I use a core i7-6950x 10 threads at 4GHz., I don't use the egdb and the hash-size is usually set at 1024 MB. Because most other programs are very cumbersome to use I currently test only against Kingsrow. Kingsrow runs on my core i7-5960x 8 threads at 4GHz., out of the box it doesn't support more than 8 threads otherwise I would have put it on the other machine as well.

Joost

[Krzysztof Grzelak]

Thank you for your answer Joost. Try your settings on my laptop. I will not use the base endings and the hash-size is give 1024 MB.

[Joost Buijs]

This is an interesting story, I don't know what to think of it.

http://talkchess.com/forum/viewtopic.ph ... w=&start=0

[Krzysztof Grzelak]

This is an interesting story, I don't know what to think of it.

http://talkchess.com/forum/viewtopic.ph ... w=&start=0

I have used this program during the tournament. Please take a look here http://www.usacheckers.com/forum/viewto ... f=2&t=3738 Very good program.

[Joost Buijs]

Thank you for your answer Joost. Try your settings on my laptop. I will not use the base endings and the hash-size is give 1024 MB.

On a slow machine with less cores 256 MB. will probably be enough, it also depends upon the time per game you use to play the match, for slow games you may set it higher.

[Krzysztof Grzelak]

On a slow machine with less cores 256 MB. will probably be enough, it also depends upon the time per game you use to play the match, for slow games you may set it higher.

Excuse me for asking a match you're playing on the same computer or on two computers using dxp.

[Joost Buijs]

On a slow machine with less cores 256 MB. will probably be enough, it also depends upon the time per game you use to play the match, for slow games you may set it higher.

Excuse me for asking a match you're playing on the same computer or on two computers using dxp.

Since I have two fast computers available I play with DXP on 2 different computers, with 2 computers you can also use pondering and 8p egdb, with one computer pondering is not possible and egdb performance goes down because both programs try to read from the same disk at the same time, of course you can use 2 different disks for the egdb.

[Rein Halbersma]

This is an interesting story, I don't know what to think of it.

http://talkchess.com/forum/viewtopic.ph ... w=&start=0

What do you mean, Joost? This is a straightforward database computation, 8x8 checkers (not 10x10 draughts!), 8 pieces, could have been done 20 years ago.

[Krzysztof Grzelak]

Since I have two fast computers available I play with DXP on 2 different computers, with 2 computers you can also use pondering and 8p egdb, with one computer pondering is not possible and egdb performance goes down because both programs try to read from the same disk at the same time, of course you can use 2 different disks for the egdb.

Please play the match using dxp but on one computer, not on the two computers.Then you will see the true power of your engine. Dxp done so in order to play on one computer as in chess.

[Catherine]

Hi Krzysztof,

No, these are the weights for the pattern evaluation, I don't have a book yet. Besides gaining some time a book for draughts does not seem very important.

Yesterday I started looking at the engine again because I wanted to improve the time-control and add a console mode to enable you to play games with it, while looking at the code I found another bug that was already there for weeks, after a fail-high at the root I increase beta in several steps and do a research while (score >= beta), after this I rechecked the condition (score >= beta) while it has to be (score > alpha), because of this the engine sometimes missed a better move. After fixing this the engine became stronger, I'm very happy that I found this, usually I overlook these things a hundred times.

The rest of this week I don't have much time for it but I will try to upload the latest version this weekend.

Joost

Hi Joost,

Thank again to working on the engine.
Like i said an engine is like a paint, it never finish but it's also good like you are doing to permit to the others persons to see the progress, thank again. We are waiting for the latest version.

This question is for you and all the others programmers:
What are the elements that can do the "perfect engine" this wich can for example will never lose.
It's the hardware on what it's launched, search evaluation, speed of search?

Thank to all

Catherine.

Hi Catherine,

A perfect engine is not possible for 10x10 Draughts because the game is too complex, maybe in the future when they build quantum computers it might be possible (I doubt they will ever exist in a usable form).

Like you said, engine strength is determined by the quality of the evaluation-function and the search-depth. The search-depth depends upon 3 things, the speed of the hardware, the efficiency of coding, and the efficiency of the algorithm itself.

Lately I get more and more the feeling that the current top-engines are already so strong that on fast hardware and with not a too short time per move all the games will be drawn, I guess it is possible to make a stronger engine but the only effect will be that it needs less time per move to reach the same strength. I have no idea how much room (Elo wise) there is still left, but time will tell.

I just ran a match of 100 games between the last version of my engine and Kingsrow with 1 minute and 90 moves for the whole game, the outcome was rather equal, 1 win, 1 loss, 96 draws and 2 unknowns (which also seem to be draws). I never tried a match with longer time but I expect that the draw rate will go up.

The engine I uploaded two days ago containes a few bugs and is about 35 Elo weaker, combine that with the slower version that you have to use on your hardware I expect it to be 60 Elo below Kingsrow, it can be that both engines behave differently on your hardware. I still have a core i3 over here with 2 cores at 3400 MHz., this weekend I will try what it does, I have to check the egdb code for compatibility anyway.

Joost

Hi Joost,

I think that the engine crashed because of the cfg setting with the egdb path wich no matched. Now that i saw how it is, the engine run perfectly.
It seem that the new engine is more stronger, wich all the fings you done and fixed.
Thank again.

Catherine