create Graphs from existing programs

[janflorijn]

I want to create graphs using the analyses of existing draught programs 10x10. In order te realise that, I need the value in pieces for each played move in a game.

F.e. Truus can create a .ana file, but this only gives values for the "bad" moves. Setting the value to 0 does not work. As far as I can see Flits does the same.

Does any one know a way/program to read the value for each played move of a game?

Kind regards,


Jan Flikweert

[BertTuyt]

Jan, i dont now a program which is able to extract this type information from a Draughts program.
To my knowledge the only program who output search information in a log file is Kingsrow.
As this is a text file, it should be not that difficult to find this information.

The straightforward way would be if programs would include search information as comment after each move in a .pdn file, but also here I'm not aware of anyone doing this...

Bert

[janflorijn]

Hi Bert,

Thanks for your reply. I will try to get a logfile from Kingsrow in order to import the nessecary data.

I have made a first idea:

http://www.mediafire.com/?3s07c4czpbm7j2y

But this needs to be worked out.
It will be fully freeware. Due to my limitations it is a lot of work for a small program. But I think very usefull.

Drag the mouse over the graph-line and you see all info about that move.

In this program i used a game imported from the analyse of the French program Dream. Nice to start, but I think not of the level of Truus/Flits/Kingsrow.

Kind regards,

Jan Flikweert

[janflorijn]

Hi all,

Is there any one who used Kingrow to analyse games and who want to sent me a log file from Kingsrow?

Kind regards,

Jan Flikweert

[MichelG]

Dragon draughts can show a graph of the score that it finds in analyses and gameplay. (menu-item view-->graph-->score)

It's also exported in the pdn file.

[janflorijn]

I will look at this. When for each analysed move in the game the score is stored in the pdn file I should be very happy and use it in my program.
I had notice from the graph in this program. I like this program because you can play games with it on your own level.

Jan FLorijn

[Rein Halbersma]

I will look at this. When for each analysed move in the game the score is stored in the pdn file I should be very happy and use it in my program.
I had notice from the graph in this program. I like this program because you can play games with it on your own level.

Jan FLorijn

A game like this

Code: Select all

[Event ""]
[Date "2012.12.17"]
[White ""]
[Black ""]
[Result "*"]
1. 32-28 18-23 2. 33-29 23x32 3. 37x28 12-18 4. 41-37 17-21 5. 37-32 19-23 6. 28x19 13x33 7. 39x28 *

will result in an analysis file like this

Code: Select all

[Event ""]
[Date "2012.12.17"]
[White ""]
[Black ""]
[Result "*"]
1. 32-28 {34-30(0/29)*  34-29(0/28)*  31-27(0/31)?  32-27(0/31)?  31-26(0/28)?  32-28(0/26)?  33-29(2/26)?  33-28(2/26)?  35-30(6/26)?}
   18-23 {17-21(0/25)*  17-22(0/28)*  20-25(0/25)?  18-23(0/25)?  19-23(0/24)?  16-21(-2/27)?  18-22(-2/27)?  19-24(-2/24)?  20-24(-2/24)?}
2. 33-29 {34-29(0/24)*  31-26(0/19)?  33-29(0/18)?  38-32(2/16)?  34-30(4/19)?  35-30(16/11)?}
   23x32 {23x32(0/17)*}
3. 37x28 {37x28(0/16)*  38x27(4/15)?}
   12-18 {17-22(0/17)*  20-25(0/15)*  19-24(0/21)?  20-24(0/16)?  16-21(0/14)?  17-21(0/7)?  19-23(-2/11)?  13-18(-4/10)?}
4. 41-37 {value=-206,  depth 19/24.0/41,  4.5 sec,  13848 kN/s,  pv 28-23 19x28 29-24 20x29 34x21 16x27 31x33 7-12 40-34 11-17 41-37}
   17-21 {value=0,  depth 17/21.7/40,  5.4 sec,  13743 kN/s,  pv 16-21 29-24 20x29 34x12 7x18 37-32 21-26 40-34 26x37 42x31 1-7}
5. 37-32 {value=0,  depth 17/21.7/40,  5.2 sec,  13335 kN/s,  pv 29-24 20x29 34x12 7x18 40-34 1-7 39-33 7-12 37-32 21-26 44-39}
   19-23 {value=2,  depth 17/21.9/42,  5.0 sec,  13615 kN/s,  pv 7-12 39-33 19-23 28x19 14x23 46-41 21-26 41-37 11-17 44-39 10-14}
6. 28x19 {value=-12,  depth 5/9.0/15,  0.0 sec,  1313 kN/s,  pv 28x19 13x33 39x28 7-12 44-39 20-24}
   13x33 {value=-6,  depth 5/7.9/17,  0.0 sec,  2637 kN/s,  pv 13x33 39x28 11-17 44-39 18-22 39-33}
7. 39x28 {value=-6,  depth 17/20.2/38,  3.8 sec,  13156 kN/s,  pv 39x28 7-12 44-39 21-26 39-33 26x37 42x31 11-17 50-44 20-24 31-27}
         {value=-6,  depth 17/19.9/39,  6.4 sec,  13453 kN/s,  pv 21-26 44-39 26x37 42x31 7-12 31-27 11-17 46-41 20-24 41-37 14-19}
*

The first 3 moves, Kingsrow is in its opening book, and you see the (score/depth) valuation. After Kingsrow drops out of its book, it starts searching on its own and you see the value, depth (nominal, average, maximum), time, nodes per second and the pv. The above output was generated with a search of 5 seconds per move, you will get more output if you use more time. There is no score threshold like in Flits/Truus. Instead, you get output after every move. You can use your favorite scripting language (Perl/Python) to extract only pv moves that are better than Delta compared to the game move.

Hope this helps.

[janflorijn]

Hi,

Thanks. Is this written to a pdn file, or is it part of the log file?

I will add it to the program.

For the most recent version download the file

setupdamgrafiek.exe

from

http://www.mediafire.com/?p1490pjn3v835

Kind regard,


Jan Flikweert

[Ed Gilbert]

Hi Jan,

I got your private message, and I see Rein has already answered some of your questions. Some more info:

- The analysis file is named analysis.txt and is found under the user's MyDocuments directory. In Windows7, it is \users\username\MyDocuments\Ed Gilbert\Kingsrow International\analysis.txt.

- From the example that Rein posted you can see that the analysis is given as comments in a PDN format. If the original PDN file that is given to the Analysis command has multiple games, there will be multiple games in the analysis.txt file.

- The search values use the convention that positive values are good for black, negative good for white. The value of a man is 100. A large positive or negative value of magnitude greater than 2000 indicates that Kingsrow sees a database win or database loss.

- In positions where the player has a forced move, kingsrow only searches to a very shallow depth of about 5 plies, so these search scores are not accurate and could be excluded from the graph.

-- Ed

[janflorijn]

Hi Ed,

Are you sure that thew value of a man is 100 and not 1000?

Kind regards,


Jan Flikweert

[janflorijn]

Hi Reindert,

Thanks for the example. It seems to be clear. One question:
1. 32-28 {34-30(0/29)* 34-29(0/28)* 31-27(0/31)? 32-27(0/31)? 31-26(0/28)? 32-28(0/26)? 33-29(2/26)? 33-28(2/26)? 35-30(6/26)?}

In the first move the score for 32-28 would be zero?

I think I use for the first three moves these values. My first lesson statistics the teacher told us: Statistics is a not accurate science.

For the most actual trie out with import single game pdn from Kingsrow,Dragon Draught and Dream download:

setupdamgrafiek.exe

from:

http://www.mediafire.com/?p1490pjn3v835


Kind regards,

Jan Flikweert

[Ed Gilbert]

Are you sure that thew value of a man is 100 and not 1000?

Yes, 100.

1. 32-28 {34-30(0/29)* 34-29(0/28)* 31-27(0/31)? 32-27(0/31)? 31-26(0/28)? 32-28(0/26)? 33-29(2/26)? 33-28(2/26)? 35-30(6/26)?}

In the first move the score for 32-28 would be zero?

Yes. The first number in (0/29) is the move score. The second number is the book PV depth along that branch.

-- Ed

[janflorijn]

Rein,

Regarding the output from Kingsrow, are the moves after 'pv' the best calculated moves using the given time?

It should be handy if I could see what the value is of the calculated best move in order to see if I can handle the played move as a good move, bad move(and include the good move), of less better move. Or can give Kingsrow a analyse/remarks in other way?

Kind regards,

Jan FLikweert

[Wieger Wesselink]

Hi Ed,

- In positions where the player has a forced move, kingsrow only searches to a very shallow depth of about 5 plies, so these search scores are not accurate and could be excluded from the graph.

I think it would be useful if these evaluations with shallow depth were replaced by a more accurate evaluation, that is based on the evaluation of the position after the forced moves. Perhaps you can implement this in a future version of Kingsrow?

Regards,

Wieger

[janflorijn]

Hi Ed and Wieger,

It is for me to difficult to overview what is better.

The purpose of all is to create graphs wich give a correct evaluation of the game. That is the purpose of statistics. How we reach that goal , does not matter.

Kind regards,

Jan Flikweert