Freyr 0.960b released

[Andrei Fortuna]

Geschrieben von: / Posted by: Andrei Fortuna at 18 March 2000 22:58:30:
Hi
Because Freyr 0.960 included a piece of debugging code which slowed it down tremendously, I'm releasing an update (0.960b) with that piece of code removed, using a new search algorithm and tablebases.
Hope it's bug-free this time, but since it wasn't tested too much please let me know if you have any problems with it.
URL: http://www.geocities.com/SiliconValley/Lab/3716/

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 20 March 2000 21:23:53:
Als Antwort auf: / As an answer to: Freyr 0.960b released geschrieben von: / posted by: Andrei Fortuna at 18 March 2000 22:58:30:

Hi
Because Freyr 0.960 included a piece of debugging code which slowed it down tremendously, I'm releasing an update (0.960b) with that piece of code removed, using a new search algorithm and tablebases.
Hope it's bug-free this time, but since it wasn't tested too much please let me know if you have any problems with it.
URL: http://www.geocities.com/SiliconValley/Lab/3716/

Perhaps that explains the poor tournament performance that we have been seeing. It is strange that Freyr seems to do very well at EPD test sets and not well at some recent tournaments. I will play around with it a bit and see what I can find out.


My FTP site

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 20 March 2000 22:03:14:
Als Antwort auf: / As an answer to: Freyr 0.960b released geschrieben von: / posted by: Andrei Fortuna at 18 March 2000 22:58:30:

Hi
Because Freyr 0.960 included a piece of debugging code which slowed it down tremendously, I'm releasing an update (0.960b) with that piece of code removed, using a new search algorithm and tablebases.
Hope it's bug-free this time, but since it wasn't tested too much please let me know if you have any problems with it.
URL: http://www.geocities.com/SiliconValley/Lab/3716/

This one looks very strong. It scored 225 out of 300 on a lowly 300 MHz PII.
There is a bug in your program processing EPD files. It repeats the last row twice for some reason. For instance, here are the last two rows in WAC:
1n2rr2/1pk3pp/pNn2p2/2N1p3/8/6P1/PP2PPKP/2RR4 w - - bm Nca4; id "WAC.299";
b2b1r1k/3R1ppp/4qP2/4p1PQ/4P3/5B2/4N1K1/8 w - - bm g6; id "WAC.300";
But the output from your program is this:
Analysing position : b2b1r1k/3R1ppp/4qP2/4p1PQ/4P3/5B2/4N1K1/8 w - - bm g6; id "
b2b1r1k/3R1ppp/4qP2/4p1PQ/4P3/5B2/4N1K1/8 w - -
+---+---+---+---+---+---+---+---+
| b | | | b | | r | | k |
+---+---+---+---+---+---+---+---+
| | | | R | | p | p | p |
+---+---+---+---+---+---+---+---+
| | | | | q | P | | |
+---+---+---+---+---+---+---+---+
| | | | | p | | P | Q |
+---+---+---+---+---+---+---+---+
| | | | | P | | | |
+---+---+---+---+---+---+---+---+
| | | | | | B | | |
+---+---+---+---+---+---+---+---+
| | | | | N | | K | |
+---+---+---+---+---+---+---+---+
| | | | | | | | |
+---+---+---+---+---+---+---+---+
2. -75 0 1468 Ra7
3. -70 0 3675 Ra7 Bc6 fxg7 Kxg7 Kf2
4. 3 0 11873 Ra7 Bxf6 gxf6 Qxf6 Kh1 Kg8
5. 197 1 62757 g6 h6 Qxh6 gxh6 g7 Kg8 gxf8Q Kxf8 Rxd8 Qe8 Rxe8
Time to analyse position : 2.09 seconds
Best move found !
pawn hashtable full : 554 (0.845337%)
Analysing position : b2b1r1k/3R1ppp/4qP2/4p1PQ/4P3/5B2/4N1K1/8 w - - bm g6; id "
b2b1r1k/3R1ppp/4qP2/4p1PQ/4P3/5B2/4N1K1/8 w - -
+---+---+---+---+---+---+---+---+
| b | | | b | | r | | k |
+---+---+---+---+---+---+---+---+
| | | | R | | p | p | p |
+---+---+---+---+---+---+---+---+
| | | | | q | P | | |
+---+---+---+---+---+---+---+---+
| | | | | p | | P | Q |
+---+---+---+---+---+---+---+---+
| | | | | P | | | |
+---+---+---+---+---+---+---+---+
| | | | | | B | | |
+---+---+---+---+---+---+---+---+
| | | | | N | | K | |
+---+---+---+---+---+---+---+---+
| | | | | | | | |
+---+---+---+---+---+---+---+---+
2. -75 0 1471 Ra7
3. -70 0 3678 Ra7 Bc6 fxg7 Kxg7 Kf2
4. 3 0 11876 Ra7 Bxf6 gxf6 Qxf6 Kh1 Kg8
5. 197 1 62760 g6 h6 Qxh6 gxh6 g7 Kg8 gxf8Q Kxf8 Rxd8 Qe8 Rxe8
Time to analyse position : 3.07 seconds
Best move found !
pawn hashtable full : 645 (0.984192%)

RESULTS :
nodes = 19616394 (24.6175%)
qnodes = 60068366 (75.3825%)
positiones searched = 301
time taken = 661.82 seconds
best moves identified = 301
of which were not recognized = 0
from the rest, best moves found = 226 (75.0831%)
average nps = 120402
hash probes = 130873
hash hits = 33908 (25.9091%)
pawn hash replaced = 13994 (3.94623%)
As you see, the statistics also report 301 rows when actually, there are only 300.

My FTP site

[Volker Pittlik]

Geschrieben von: / Posted by: Volker Pittlik at 20 March 2000 22:07:25:
Als Antwort auf: / As an answer to: Re: Freyr 0.960b released geschrieben von: / posted by: Dann Corbit at 20 March 2000 22:03:14:

Hi
Because Freyr 0.960 included a piece of debugging code which slowed it down tremendously, I'm releasing an update (0.960b) with that piece of code removed, using a new search algorithm and tablebases.
Hope it's bug-free this time, but since it wasn't tested too much please let me know if you have any problems with it.
URL: http://www.geocities.com/SiliconValley/Lab/3716/

This one looks very strong. It scored 225 out of 300 on a lowly 300 MHz PII.
There is a bug in your program processing EPD files. It repeats the last row twice for some reason. For instance, here are the last two rows in WAC:
1n2rr2/1pk3pp/pNn2p2/2N1p3/8/6P1/PP2PPKP/2RR4 w - - bm Nca4; id "WAC.299";
b2b1r1k/3R1ppp/4qP2/4p1PQ/4P3/5B2/4N1K1/8 w - - bm g6; id "WAC.300";
But the output from your program is this:
Analysing position : b2b1r1k/3R1ppp/4qP2/4p1PQ/4P3/5B2/4N1K1/8 w - - bm g6; id "
b2b1r1k/3R1ppp/4qP2/4p1PQ/4P3/5B2/4N1K1/8 w - -
+---+---+---+---+---+---+---+---+
| b | | | b | | r | | k |
+---+---+---+---+---+---+---+---+
| | | | R | | p | p | p |
+---+---+---+---+---+---+---+---+
| | | | | q | P | | |
+---+---+---+---+---+---+---+---+
| | | | | p | | P | Q |
+---+---+---+---+---+---+---+---+
| | | | | P | | | |
+---+---+---+---+---+---+---+---+
| | | | | | B | | |
+---+---+---+---+---+---+---+---+
| | | | | N | | K | |
+---+---+---+---+---+---+---+---+
| | | | | | | | |
+---+---+---+---+---+---+---+---+
2. -75 0 1468 Ra7
3. -70 0 3675 Ra7 Bc6 fxg7 Kxg7 Kf2
4. 3 0 11873 Ra7 Bxf6 gxf6 Qxf6 Kh1 Kg8
5. 197 1 62757 g6 h6 Qxh6 gxh6 g7 Kg8 gxf8Q Kxf8 Rxd8 Qe8 Rxe8
Time to analyse position : 2.09 seconds
Best move found !
pawn hashtable full : 554 (0.845337%)
Analysing position : b2b1r1k/3R1ppp/4qP2/4p1PQ/4P3/5B2/4N1K1/8 w - - bm g6; id "
b2b1r1k/3R1ppp/4qP2/4p1PQ/4P3/5B2/4N1K1/8 w - -
+---+---+---+---+---+---+---+---+
| b | | | b | | r | | k |
+---+---+---+---+---+---+---+---+
| | | | R | | p | p | p |
+---+---+---+---+---+---+---+---+
| | | | | q | P | | |
+---+---+---+---+---+---+---+---+
| | | | | p | | P | Q |
+---+---+---+---+---+---+---+---+
| | | | | P | | | |
+---+---+---+---+---+---+---+---+
| | | | | | B | | |
+---+---+---+---+---+---+---+---+
| | | | | N | | K | |
+---+---+---+---+---+---+---+---+
| | | | | | | | |
+---+---+---+---+---+---+---+---+
2. -75 0 1471 Ra7
3. -70 0 3678 Ra7 Bc6 fxg7 Kxg7 Kf2
4. 3 0 11876 Ra7 Bxf6 gxf6 Qxf6 Kh1 Kg8
5. 197 1 62760 g6 h6 Qxh6 gxh6 g7 Kg8 gxf8Q Kxf8 Rxd8 Qe8 Rxe8
Time to analyse position : 3.07 seconds
Best move found !
pawn hashtable full : 645 (0.984192%)

RESULTS :
nodes = 19616394 (24.6175%)
qnodes = 60068366 (75.3825%)
positiones searched = 301
time taken = 661.82 seconds
best moves identified = 301
of which were not recognized = 0
from the rest, best moves found = 226 (75.0831%)
average nps = 120402
hash probes = 130873
hash hits = 33908 (25.9091%)
pawn hash replaced = 13994 (3.94623%)
As you see, the statistics also report 301 rows when actually, there are only 300.

[Andrei Fortuna]

Geschrieben von: / Posted by: Andrei Fortuna at 21 March 2000 00:22:48:
Als Antwort auf: / As an answer to: Re: Freyr 0.960b released geschrieben von: / posted by: Dann Corbit at 20 March 2000 21:23:53:

Hi
Because Freyr 0.960 included a piece of debugging code which slowed it down tremendously, I'm releasing an update (0.960b) with that piece of code removed, using a new search algorithm and tablebases.
Hope it's bug-free this time, but since it wasn't tested too much please let me know if you have any problems with it.
URL: http://www.geocities.com/SiliconValley/Lab/3716/

Perhaps that explains the poor tournament performance that we have been seeing. It is strange that Freyr seems to do very well at EPD test sets and not well at some recent tournaments.

It certainly doesn't have the positional function calibrated. Perhaps this has to do with it, or I should set thread priority for the analyze thread higher than for the reading the command line thread, or when out of time it might return a garbage move sometimes (although I think I took special measures against this case not to save the pv if out of time, besides this would affect analysis too).

[Andrei Fortuna]

Geschrieben von: / Posted by: Andrei Fortuna at 21 March 2000 00:25:25:
Als Antwort auf: / As an answer to: Re: Freyr 0.960b released geschrieben von: / posted by: Dann Corbit at 20 March 2000 22:03:14:

Hi
Because Freyr 0.960 included a piece of debugging code which slowed it down tremendously, I'm releasing an update (0.960b) with that piece of code removed, using a new search algorithm and tablebases.
Hope it's bug-free this time, but since it wasn't tested too much please let me know if you have any problems with it.
URL: http://www.geocities.com/SiliconValley/Lab/3716/

This one looks very strong. It scored 225 out of 300 on a lowly 300 MHz PII.
There is a bug in your program processing EPD files. It repeats the last row twice for some reason.

One of my first releases scored 229/300 as you told me, wasn't on the same 300 Mhz PII ?
I know of this since day 1, I was just too lazy to fix it.

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 21 March 2000 00:38:15:
Als Antwort auf: / As an answer to: Re: Freyr 0.960b released geschrieben von: / posted by: Andrei Fortuna at 21 March 2000 00:25:25:

Hi
Because Freyr 0.960 included a piece of debugging code which slowed it down tremendously, I'm releasing an update (0.960b) with that piece of code removed, using a new search algorithm and tablebases.
Hope it's bug-free this time, but since it wasn't tested too much please let me know if you have any problems with it.
URL: http://www.geocities.com/SiliconValley/Lab/3716/

This one looks very strong. It scored 225 out of 300 on a lowly 300 MHz PII.
There is a bug in your program processing EPD files. It repeats the last row twice for some reason.

One of my first releases scored 229/300 as you told me, wasn't on the same 300 Mhz PII ?
I know of this since day 1, I was just too lazy to fix it.

I don't know. If I said it was a 300 MHz machine, it was probably the same one. I also run program benchmarks on a 500 MHz PIII. The figures on the 300 MHz machine are rough ballparks anyway, since I run a lot of programs on it and test in the background.


My FTP site

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 21 March 2000 05:22:55:
Als Antwort auf: / As an answer to: Re: Freyr 0.960b released geschrieben von: / posted by: Dann Corbit at 21 March 2000 00:38:15:

Hi
Because Freyr 0.960 included a piece of debugging code which slowed it down tremendously, I'm releasing an update (0.960b) with that piece of code removed, using a new search algorithm and tablebases.
Hope it's bug-free this time, but since it wasn't tested too much please let me know if you have any problems with it.
URL: http://www.geocities.com/SiliconValley/Lab/3716/

This one looks very strong. It scored 225 out of 300 on a lowly 300 MHz PII.

One of my first releases scored 229/300 as you told me, wasn't on the same 300 Mhz PII ?

I don't know. If I said it was a 300 MHz machine, it was probably the same one. I also run program benchmarks on a 500 MHz PIII. The figures on the 300 MHz machine are rough ballparks anyway, since I run a lot of programs on it and test in the background.

I think probably it was the 500 MHz machine that got 229, since that is what this version just scored on that machine (230 - 1 since the last one is a "bounce"). Here is the crafty report of the output file from Freyr:
Crafty v17.10
White(1): epdpfdr wac.out wac.res
White(1): epdscor wac.res
total: 301
total acd: 1831 mean total acd: 6.08
total acn: 28824200 mean total acn: 95761.46
total acs: 437 mean total acs: 1.45
total mean node frequency: 65959.27 Hz
solve: 230 solve percent: 76.41
solve acd: 1370 mean solve acd: 5.96
solve acn: 20611850 mean solve acn: 89616.74
solve acs: 297 mean solve acs: 1.29
solve mean node frequency: 69400.17 Hz
unsol: 71 unsol percent: 23.59
unsol acd: 461 mean unsol acd: 6.49
unsol acn: 8212350 mean unsol acn: 115666.90
unsol acs: 140 mean unsol acs: 1.97
unsol mean node frequency: 58659.64 Hz
White(1):