Repost: Pro Deo 1.5 in Arasan 7.epd

[Eelco de Groot]

I have been experimenting a bit with the new Pro Deo and I think that, even without the EOCs, Ed has improved the engine and Pro Deo 1.5 should show some good results! This is just a first impression, but the engine seems faster and has a better time to plydepth performance, better at what is usually called the branching factor. Of course I had to try some personality settings for Pro Deo 1.5 as well so I can't show you many games of Pro Deo 1.5 itself with the standard settings.

But here is the result for the Arasan 7.epd test, the version by Jon Dart dated feb 27 2007:

Pro Deo default scored 152 found positions from 227 at 20 seconds per position, 128 Mb hash, single Athlon 2009 MHz. Mot bad!

The resultmatrix:

Code: Select all

 
Final result of EPD analysis, 128Mb hastables, 20 seconds per position, sol. times longer than 20 s. are in reality less than 20. 

Engine: Pro Deo 1.5 - prodeo 
Analyzed 227 positions of which 152 were solved. 
Total time for the solved positions 1200 seconds. 

Time to solve matrix (all times in seconds). 
   1   2   3   4   5   6   7   8   9   0 
  0   20   -   1   5   21   4   4   20   10   3 
 10   0   -   20   0   20   1   12   0   4   19 
 20   20   -   5   7   -   -   13   4   6   - 
 30   4   4   4   5   14   3   7   12   0   - 
 40   6   -   4   4   5   -   0   -   10   0 
 50   3   24   6   -   21   -   4   -   10   1 
 60   12   2   18   -   -   4   4   4   6   6 
 70   7   -   -   4   5   4   -   12   -   - 
 80   -   -   -   11   4   20   0   4   4   5 
 90   -   3   -   2   3   5   -   5   0   5 
100   4   3   19   2   21   -   4   0   11   - 
110   -   7   18   16   23   -   13   0   13   4 
120   24   -   -   -   -   -   4   0   5   - 
130   -   6   1   3   4   1   -   17   -   24 
140   6   6   -   6   4   4   -   10   -   - 
150   -   5   8   4   -   1   3   -   9   - 
160   5   -   10   -   5   -   13   4   -   2 
170   3   21   6   -   5   -   -   7   19   - 
180   15   3   -   20   -   15   10   -   -   - 
190   9   -   16   -   11   -   10   5   -   21 
200   -   13   1   -   10   4   18   7   -   - 
210   -   11   7   -   11   1   -   4   1   - 
220   4   6   -   -   -   14   12 
 

At first I had wanted to test this with Pro Deo as UCI engine in the Shredder 9.1 GUI but Prodeo can't analyze exactly 20 seconds there, at least not with rebeluci.exe translating the UCI commands. But Pro Deo 1.5 can play games as an UCI engine and normal analysis is also going well. The Arasan test I did in the ChessPartner GUI, with Pro Deo as a Winboard engine, that also shows the EOC screens etc. and is kind of Pro Deo's "home base".

I also tried some of my older Pro Deo settings, here a variation of the "Excalibur" settings, in testing Arasan 7.epd The result for Pro Deo Excalibur I was 158 found from 227 positions:

Code: Select all

 
Final result of EPD analysis 128Mb hastables 20 seconds per position, sol. times longer than 20 s. are in reality less than 20. 

Engine: Pro Deo 1.5 - Excalibur Xa 
Analyzed 227 positions of which 158 were solved. 
Total time for the solved positions 1317 seconds. 

Time to solve matrix (all times in seconds). 
   1   2   3   4   5   6   7   8   9   0 
  0   21   -   1   9   -   -   11   20   8   6 
 10   0   14   20   -   20   -   -   1   4   - 
 20   20   -   4   7   -   -   1   -   6   16 
 30   4   4   9   5   5   1   4   20   1   4 
 40   6   -   5   4   6   -   1   20   -   1 
 50   4   13   5   7   15   8   10   -   16   8 
 60   -   4   -   -   6   6   -   4   -   6 
 70   13   3   -   4   5   4   -   11   -   - 
 80   -   -   7   21   5   20   1   19   4   7 
 90   5   8   -   -   7   15   -   6   1   4 
100   4   4   -   -   8   -   4   1   4   4 
110   20   17   10   12   22   11   13   1   4   4 
120   24   -   16   -   -   10   4   0   5   - 
130   -   14   1   6   4   3   -   5   15   23 
140   14   4   24   9   4   4   -   20   2   - 
150   -   4   3   7   7   3   6   -   3   - 
160   -   10   14   -   2   -   2   4   -   - 
170   -   20   6   -   3   7   -   11   13   - 
180   -   4   3   2   20   -   5   -   -   - 
190   11   -   -   13   5   18   6   5   20   20 
200   -   -   1   17   3   9   8   4   -   - 
210   23   16   1   -   -   0   -   4   0   - 
220   5   12   -   13   -   22   3 

Not a bad result but I think it can be improved because several lines in the Excalibur Xa.eng file are outdated and are either not working anymore or may even hurt Pro Deo 1.5 performance. For instance I had a switch

[Pruning ? MISC_13] * Extra bonus Bishop Pair

but looking in prodeo.eng I see I have to replace that with

[Bishop Pair Extra = 100] * Tune Strong Bishop Pair evaluation

Another example, there was

[Rook 7th Rank (low) ? 60] *
[Rook 7th Rank (high) ? 60] *

which I think should be replaced with

[Rook 7th Rank (1) = 60] * When opponent king is on the 8th rank
[Rook 7th Rank (1) = 60] * Tune caught opponent king (endgame)

I am not even very sure about the second line, but that was how the latest update Pro Deo 1.2 was doing it.

So now with just these first changes I made an new Excalibur II, but it still has many settings that I am not sure are working properly. But this version already can solve one position at least better than the old Excalibur, Arasan position No. 6 goes from 25 seconds to now 8 seconds!


Excalibur II

[D]4k3/5ppQ/1n3b2/3bp3/1B6/1P1R2NP/5PPK/1qr5 w - - bm Ba5; id "arasan4.9";

4k3/5ppQ/1n3b2/3bp3/1B6/1P1R2NP/5PPK/1qr5 w - -

00:00:00 1.00 -9.26 1.Rxd5 Qxh7
00:00:00 1.03 -2.84 1.Bd6
00:00:00 1.17 -2.60 1.f3
00:00:00 1.20 0.95 1.Qg8 Kd7 2.Qxf7 Kc8 3.Rxd5 Nxd5 4.Qxd5
00:00:00 2.00 0.95 1.Qg8 Kd7 2.Qxf7 Kc8 3.Rxd5 Nxd5 4.Qxd5
00:00:00 3.00 0.75 1.Qg8 Kd7 2.Qxf7 Kc6 3.Rxd5 Nxd5 4.Qe6 Kb7
5.Qxd5 Ka7 6.Qd7 Ka8
00:00:00 4.00 0.75 1.Qg8 Kd7 2.Qxf7 Kc6 3.Rxd5 Nxd5 4.Qe6 Kb7
5.Qxd5 Ka7 6.Qd7 Ka8
00:00:00 5.00 1.18 1.Qg8 Kd7 2.Qxf7 Kc8 3.Rxd5 Nxd5 4.Qxd5 Qa2
5.Bc5 Rb1 6.Qc6 Kb8
00:00:00 6.00 0.82 1.Qg8 Kd7 2.Qxf7 Kc6 3.Rxd5 Nxd5 4.Qe6 Kb5
5.Ba3 Nb6 6.Bxc1 Qxc1 7.Qe8 Ka6 8.b4 Qf4 9.b5
Ka7 10.Qf8 Qxf2 11.Qb4
00:00:00 7.00 0.27 1.Qg8 Kd7 2.Qxf7 Kc6 3.Rxd5 Nxd5 4.Qe6 Kb5
5.Bd2 Nf4 6.Qd6 Rc6 7.Qb8 Ka6 8.Bxf4 exf4 9.Qxf4
Qxb3
00:00:00 8.00 0.86 1.Qg8 Kd7 2.Qxf7 Kc6 3.Rxd5 Nxd5 4.Qe6 Kb5
5.Bd2 Nf4 6.Qd7 Rc6 7.Nf5 Kb6 8.Be3 Ka5 9.Qxc6
Qxf5
00:00:00 9.00 0.86 1.Qg8 Kd7 2.Qxf7 Kc6 3.Rxd5 Nxd5 4.Qe6 Kb5
5.Bd2 Nf4 6.Qd7 Rc6 7.Nf5 Kb6 8.Be3 Ka5 9.Qxc6
Qxf5
00:00:02 10.00 -0.01 1.Qg8 Kd7 2.Qxf7 Kc6 3.Rxd5 Nxd5 4.Qe6 Kb5
5.Bd2 Nc7 6.Qc8 Kb6 7.Be3 Ka5 8.Bd2 Kb6 9.Be3
Ka5 10.Bd2
00:00:06 11.00 0.14 1.Qg8 Kd7 2.Qxf7 Kc6 3.Rxd5 Nxd5 4.Qe6 Kb5
5.Bd2 Nc7 6.Qc8 Kb6 7.Be3 Kc6 8.f3 Rc2
00:00:08 11.10 0.14 1.Ba5
00:00:09 11.10 0.41 1.Ba5 Nd7 2.Qg8 Nf8 3.Rxd5 Qxb3 4.Rd2
00:00:11 12.00 0.13 1.Ba5 Nd7 2.Qg8 Nf8 3.Rxd5 Qxb3 4.Rd2
00:00:26 13.00 0.13 1.Ba5 Nd7 2.Qg8 Nf8 3.Rxd5 Qxb3 4.Rd2
00:01:24 14.00 0.85 1.Ba5 Nd7 2.Qg8 Nf8 3.Rxd5 Qxb3 4.Rd6 Be7 5.Rb6 Qd1 6.Rb8 Kd7 7.Rb7
00:03:26 15.00 1.05 1.Ba5 Nd7 2.Qh8 Nf8 3.Rxd5 Qh7 4.Qxh7 Nxh7
5.Ne4 Ke7 6.Bc3 Ng5

Another nice result is Aljechin - Sterk, one of the last positions in Arasan 7.epd

Pro Deo 1.5 has trouble finding Bf6, at least testing it as UCI engine with 128 MB:

[D]r4rk1/p4ppp/qp2p3/b5B1/n1R5/5N2/PP2QPPP/1R4K1 w - - bm Bf6; id "Alekine-Sterk, Budapest 1921";
r4rk1/p4ppp/qp2p3/b5B1/n1R5/5N2/PP2QPPP/1R4K1 w - -

Engine: ProDeo 1.5 (128 MB)
gemaakt door Ed Schroeder

8.00 0:00 +0.34 1.Pd4 Pc5 2.b4 e5 3.Pc6 f6 4.bxc5 fxg5 5.Pxe5 bxc5 6.De4 (167.763) 167763

8.00 0:00 +0.34 1.Df1 (324.991) 324991

8.00 0:00 +1.10 1.Df1 Pc5 2.b4 Lxb4 3.Tbxb4 Dxa2 4.Db1 Dxb1+ 5.Txb1 (336.218) 336218

9.00 0:00 +1.29 1.Df1 Pc5 2.b4 Lxb4 3.Tcxb4 Dxa2 4.Ta1 Dd5 5.Db1 (434.619) 434619

10.00 0:00 +1.23 1.Df1 Pc5 2.b4 Lxb4 3.Tcxb4 f6 4.Le3 Dxa2 5.Lxc5 bxc5 6.T1b2 Dd5 (703.292) 703292

11.00 0:00 +1.23 1.Df1 Pc5 2.b4 Lxb4 3.Tcxb4 f6 4.Le3 Dxa2 5.Lxc5 bxc5 6.T1b2 Dd5 (1.142.597) 1142597

12.01 0:01 +1.33 1.Df1 Pc5 2.b4 Lxb4 3.Tcxb4 Dxa2 4.Ta1 Dc2 5.Tc4 Db2 6.Le7 (2.213.897) 2213

13.01 0:04 +1.04 1.Df1 Pc5 2.b4 Lxb4 3.Tcxb4 Dxa2 4.Ta1 Dc2 5.Tc4 Db2 6.Dc1 Dxc1+ 7.Taxc1 f6 8.Le3 (5.632.011) 1408

13.02 0:05 +1.25 1.Pd4 Pc5 2.b4 Pe4 3.Le7 Tfe8 4.bxa5 Pd2 5.Dxd2 Dxc4 6.Tc1 Dd5 (7.878.268) 1575

14.01 0:20 +1.58 1.Pd4 Pc5 2.b4 e5 3.Pf5 Pe6 4.Le7 b5 5.Tc3 Lb6 6.Lxf8 Txf8 7.Td1 Ld4 8.Tcc1 (29.197.786) 1459

15.01 0:40 +1.64 1.Pd4 Pc5 2.b4 Pe4 3.Le7 Tfe8 4.Tbc1 b5 5.Tc6 Lb6 6.Dxe4 Txe7 (58.973.817) 1474

16.01 1:53 +1.28 1.Pd4 Pc5 2.b4 Pe4 3.Lf4 Tad8 4.Pxe6 fxe6 5.Dxe4 b5 6.Tc6 (165.660.014) 1466

17.01 12:37 +1.44 1.Pd4 Pc5 2.b4 Pe4 3.Lf4 Tad8 4.Dc2 e5 5.Lxe5 Pd2 6.Dxd2 Dxc4 7.bxa5 Tfe8 8.Tc1 Da4 (1.092.795.493) 1443

17.19 123:01 +2.71 1.Lf6 Tfc8 2.De5 Tc5 3.Dg3 g6 4.Txa4 (1.648.525.801) 223

After 123 minutes Pro Deo 1.5, UCI in Shredder GUI, finally completed its (Fail-high?) search of 1.Bf6.

But Excalibur II is really very fast! 256 Mb, Winboard in ChessPartner but the UCI version with 128 Mb also goes well and is just as fast:

r4rk1/p4ppp/qp2p3/b5B1/n1R5/5N2/PP2QPPP/1R4K1 w - -

00:00:00 8.00 0.97 1.Nd4 Nc5 2.b4 h6 3.Be7 Rfe8 4.Bxc5 bxc5 5.bxc5
00:00:00 9.00 0.97 1.Nd4 Nc5 2.b4 f6 3.Bc1 e5 4.bxc5 exd4 5.Qe6 Kh8 6.Rxd4
00:00:00 10.00 0.48 1.Nd4 Nc5 2.b4 e5 3.Nf5 f6 4.bxc5 fxg5 5.Qxe5
00:00:01 10.01 0.54 1.Qc2 b5 2.Rc6 Qb7 3.Rc1 f6 4.Be3 Rae8
00:00:01 10.13 0.54 1.Qf1
00:00:01 10.13 1.19 1.Qf1 f6 2.Be3 Qb5 3.b4 Rac8 4.Rxc8 Qxf1 5.Kxf1 Rxc8 6.bxa5
00:00:02 11.00 1.19 1.Qf1 f6 2.Be3 Qb5 3.b4 Rac8 4.Rxc8 Qxf1 5.Kxf1 Rxc8 6.bxa5
00:00:02 11.02 1.22 1.Nd4 Nc5 2.b4 e5 3.Nf5 Ne6 4.bxa5 Nxg5 5.Qxe5 Qxa5 6.Qxg7
00:00:03 12.00 1.17 1.Nd4 Nc5 2.b4 e5 3.Nf5 Ne6 4.b5 Qb7 5.Ne7 Kh8 6.Qxe5 Rae8
00:00:08 13.00 1.48 1.Nd4 Nc5 2.b4 e5 3.Nc6 Ne6 4.Bd2 b5 5.Ne7 Kh8 6.Rh4 Nd4 7.Rxh7 Kxh7 8.Qh5 Qh6 9.Bxh6 gxh6
00:00:13 14.00 1.29 1.Nd4 Nc5 2.b4 Ne4 3.Be7 Rfe8 4.bxa5 Nd2 5.Qxd2 Qxc4 6.Rc1
00:00:45 15.00 1.33 1.Nd4 Nc5 2.b4 e5 3.Nc6 Ne6 4.Be3 b5 5.Ne7 Kh8 6.Rh4 g5 7.Rxh7 Kxh7 8.Qh5 Kg7 9.bxa5
00:01:12 15.04 1.36 1.Bh6 Nc5 2.Ne5 f5 3.b4 gxh6 4.bxc5 b5 5.Rxb5 Qxb5 6.Rg4 fxg4 7.Qxb5
00:01:56 15.09 1.41 1.Bf6 Nc5 2.Ne5 gxf6 3.Rg4 Kh8 4.Qe3 Bd2 5.Qxd2 Qxa2 6.Rc1 Rac8 7.Qh6
00:07:32 16.00 2.66 1.Bf6 Rfc8 2.Qe5 Rc5 3.Rxc5 gxf6 4.Qg3 Kh8 5.Rc7

A nice game from the first Excalibur version against Colossus 2007b:

[Event "6 minuten per partij + 6 seconden per zet"]
[Site "Engine Match"]
[Date "2007.06.15"]
[Round "1"]
[White "ProDeo 1.5 Excalibur I"]
[Black "Colossus 2007b"]
[Result "1-0"]

1. d4 {Boek 2:18m} Nf6 {Boek 0s} 2. c4 {Boek 0s} c5 {Boek
0s} 3. d5 {Boek 0s} e6 {Boek 0s} 4. Nc3 {Boek 0s} exd5
{Boek 0s} 5. cxd5 {Boek 0s} d6 {Boek 0s} 6. e4 {Boek 0s} g6
{Boek 0s} 7. h3 {Boek 0s} Bg7 {-0.26/12 19s} 8. Bd3 {Boek
0s} O-O {-0.26/12 18s} 9. Nf3 {Boek 0s} Re8 {-0.40/12 24s}
10. O-O {Boek 0s} c4 {Boek 0s} 11. Bc2 {Boek 0s} b5 {Boek
0s} 12. Nxb5 {Boek 0s} Nxe4 {-0.42/13 18s} 13. Bxe4 {Boek
0s} Rxe4 {-0.38/13 19s} 14. Nd2 {Boek 0s} Rh4 {0.00/13 22s}
15. Re1 {Boek 0s} Bxh3 {+0.18/12 12s} 16. gxh3 {+0.64/8 1s}
Qd7 {+0.32/12 24s} 17. Qf3 {+0.35/12 15s} Qxb5 {+0.74/13
11s} 18. Re7 {+0.40/13 18s} f5 {+1.04/13 14s} 19. a4
{+0.80/12 23s} Qc5 {+0.52/13 17s} 20. Re8+ {+0.66/12 17s}
Kf7 {+0.48/14 12s} 21. Ne4 {+0.84/14 19s} Qb4 {+0.08/13
17s} 22. Bd2 {+1.40/12 9s} Qb6 {-0.06/13 14s} 23. a5
{+2.08/12 28s} Qa6 {-1.08/13 19s} 24. Re1 {+2.44/12 18s}
Rxe4 {-1.48/13 11s} 25. R1xe4 {+4.57/12 10s} Be5 {-1.56/14
18s} 26. Rd8 {+5.18/12 11s} Bf6 {-0.30/11 19s} 27. Ree8
{+9.42/13 11s} Nd7 {-9.82/11 16s} 28. Rxa8 {+12.10/12 15s}
Ne5 {-12.08/10 16s} 29. Rf8+ {+16.66/10 12s} 1-0

To be sure, Colossus won the return game but that did not go so fast! I think Pro Deo 1.5 was using its own book here and Colossus was using the Shredder 9 book, so in the two games the opening was not completely repeated. I had not disabled Pro Deo's own book in wb2uci.eng here as I should have to let Pro Deo use the GUI's shredder 9 book.

P.S.

[Rook 7th Rank (1) = 60] * When opponent king is on the 8th rank
[Rook 7th Rank (1) = 60] * Tune caught opponent king (endgame)

should really be

[Rook 7th Rank (1) = 60] * When opponent king is on the 8th rank
[Rook 7th Rank (2) = 60] * Tune caught opponent king (endgame)

to activate the second line, which means that in the current Excalibur II it is not yet working. I'm a bit more confident about the first line having a reasonable value, maybe I'll leave out the second.

[Eelco de Groot]

Kaj Soderberg wrote:

Good to hear that PD15 is a slight improvement. Being a personalities fan myself, it would be good to know that the parameters of PD15 correspond with PD12, before starting o fiddle around with them and of course use current good personalities for comparison. It should be all unchanged, but you never now. Maybe the master himself will speak, Ed?

Best regards,

Kaj

Hello Kaj, I found some of the older posts from Ed back and a few of my own .eng files, with that I hope I can eliminate the most offending lines from my old settings. It doesn't hurt to try some of the old settings though in Pro Deo 1.5. For instance I thought that most old [Pruning = MISC] settings would simply have no effect anymore unless Ed specified them as still working. But I found several that still have an effect, although I'm not sure they are doing the same thing as before. I had for instance the following line

[Pruning = MISC_12] * Double Lazy Eval (safety)

which I thought I could safely scrap but then the result for Aljechin - Sterk became much worse. It should not really have much effect because I am also using [Chess Knowledge = 300] which means Pro Deo is using very little Lazy Eval. But maybe the switch is doing something else, I think it is worth trying it in the default parameters as well, at least this switch is now still in Excalibur!

With some settings called Test3 I could improve the Arasan result to 161, almost as good as Glaurung 1.2.1!

Code: Select all

Final result of EPD analysis

Engine: Pro Deo 1.5 - Test3
Analyzed 227 positions of which 161 were solved.
Total time for the solved positions 1193 seconds.

Time to solve matrix (all times in seconds).
	1	2	3	4	5	6	7	8	9	0
  0	20	-	4	12	-	-	1	23	4	9
 10	4	10	24	-	24	-	-	4	1	-
 20	23	-	8	4	-	-	5	-	3	18
 30	0	1	5	1	2	4	1	9	4	23
 40	2	-	1	0	2	-	5	-	-	4
 50	1	9	1	8	11	4	6	-	13	11
 60	-	7	-	-	3	3	-	0	-	2
 70	15	13	-	1	8	0	-	7	-	-
 80	-	-	4	17	2	24	5	15	0	4
 90	1	10	-	-	8	11	-	3	4	0
100	1	0	-	9	5	-	0	4	1	0
110	17	13	7	9	18	7	10	4	8	1
120	20	-	6	-	-	6	1	3	1	-
130	24	10	5	2	1	7	-	2	20	-
140	17	1	20	5	1	0	17	16	5	12
150	-	0	0	3	3	7	2	-	5	-
160	-	13	17	-	5	-	5	1	-	23
170	-	24	3	-	6	3	-	8	10	-
180	-	7	0	5	24	-	1	-	-	-
190	7	-	-	9	2	15	6	1	24	24
200	-	-	4	20	7	5	4	0	-	-
210	20	12	5	-	-	4	-	0	4	-
220	1	8	-	9	-	18	0

The Arasan No.6 position is still not solved with 128 Mb hashtables though and these settings have the old Pro Deo 1.1
[Pruning = MISC_37] * Hash Table extension (-18%) DEFAULT
which I'm not 100% sure is still good. I see I still have left this in Test 10 though.

I would also like to try the Botwinnik- Markov extensions
[Pruning = MISC_11] * Allow Botvinnik-Markoff Extension
again which also are not in Test3

With the latest Test10 settings I have not yet tried Arasan7.epd but two testgames ProDeo playing White against Colossus went well. I think Pro Deo was just pretty lucky But in both games Pro Deo managed to win out of even looking middlegames, it won in the endings! Maybe just luck and Pro Deo Default would play this much better but I still liked watching these two games!

[Event "6 minuten per partij + 6 seconden per zet"]
[Site "Engine Match"]
[Date "2007.06.18"]
[Round "1"]
[White "ProDeo 1.5 Excalibur Test10"]
[Black "Colossus 2007b"]
[Result "1-0"]

1. d4 {Boek 0s} Nf6 {Boek 0s} 2. c4 {Boek 0s} e6 {Boek 0s}
3. Nc3 {Boek 0s} Bb4 {Boek 0s} 4. e3 {Boek 0s} c5 {Boek 0s}
5. Nge2 {Boek 0s} b6 {Boek 0s} 6. a3 {Boek 0s} Ba5 {Boek
0s} 7. g3 {Boek 0s} Bb7 {+0.60/13 22s} 8. d5 {Boek 0s} exd5
{Boek 0s} 9. Bg2 {+0.04/14 13s} Ba6 {Boek 0s} 10. b4
{+0.88/14 14s} cxb4 {-1.44/14 21s} 11. axb4 {+1.10/14 21s}
Bxb4 {-1.44/15 20s} 12. Bxd5 {+1.26/14 21s} Nxd5 {-1.30/14
20s} 13. Qxd5 {+1.08/11 2s} O-O {-0.54/14 20s} 14. Bd2
{+0.30/13 12s} Qc8 {+0.04/14 13s} 15. Qxa8 {+0.12/12 3s}
Bb7 {+0.12/15 18s} 16. Qxa7 {+0.60/15 12s} Bxh1 {+0.08/14
11s} 17. Nd5 {+0.09/14 22s} Bc5 {+0.08/13 12s} 18. Bc3
{+0.30/13 17s} b5 {+0.24/12 11s} 19. Qa2 {+0.37/13 13s}
Bxd5 {+0.16/13 22s} 20. cxd5 {+0.25/12 3s} b4 {+0.10/14
17s} 21. Bb2 {+0.05/15 21s} d6 {+0.20/14 18s} 22. Nf4
{+0.05/13 11s} Qf5 {+0.16/14 18s} 23. Qc4 {+0.09/12 9s} Nd7
{+0.16/14 19s} 24. e4 {+0.10/12 10s} Qg5 {+0.24/13 14s}
25. Qe2 {+0.04/12 11s} b3 {+0.22/13 17s} 26. Nh5 {+0.33/12
24s} f6 {+0.26/12 10s} 27. h4 {+0.59/13 14s} Qg6 {+0.26/14
14s} 28. Nf4 {+0.42/13 15s} Qf7 {+0.22/14 14s} 29. Ne6
{+0.44/12 13s} Rb8 {+0.16/13 16s} 30. f4 {+0.37/13 16s} Bb6
{+0.32/11 10s} 31. h5 {+0.37/12 9s} Nc5 {+0.36/12 16s}
32. f5 {+0.22/12 12s} h6 {+0.28/13 13s} 33. Kf1 {+0.28/12
11s} Kh8 {+0.26/12 15s} 34. Ra3 {+0.31/12 12s} Qd7
{+0.22/13 14s} 35. Kg2 {+0.36/12 7s} Qe8 {+0.24/13 10s}
36. Qg4 {+0.30/11 8s} Qe7 {+0.24/14 10s} 37. Qf3 {+0.28/12
12s} Kg8 {0.00/14 12s} 38. Ra1 {+0.26/12 8s} Qb7 {+0.18/13
13s} 39. Qg4 {+0.32/12 12s} Kh8 {0.00/12 10s} 40. Qe2
{+0.18/12 14s} Kh7 {+0.18/12 12s} 41. Ra3 {+0.28/12 13s}
Kg8 {+0.18/14 10s} 42. Qg4 {+0.31/12 13s} Kh8 {0.00/13 10s}
43. Qf3 {+0.31/12 8s} Qe7 {+0.18/13 10s} 44. Kh1 {+0.19/13
19s} Ba7 {+0.22/11 1s} 45. Kh2 {+0.25/12 8s} Qf7 {+0.10/13
11s} 46. Kg2 {+0.32/12 11s} Qb7 {0.00/13 2s} 47. Ra1
{+0.30/12 10s} Rc8 {+0.18/12 10s} 48. Kh3 {+0.28/12 8s} Qf7
{+0.24/11 3s} 49. Bd4 {+0.21/12 11s} Rb8 {+0.24/11 3s}
50. Rb1 {+0.16/12 11s} Qe8 {+0.32/12 8s} 51. Bxc5 {+0.06/12
12s} dxc5 {+0.16/12 2s} 52. e5 {+0.30/13 11s} b2 {-0.30/13
10s} 53. exf6 {+0.56/12 11s} gxf6 {-0.88/12 2s} 54. Nf4
{+0.62/13 12s} Qd7 {-0.12/13 10s} 55. Qg4 {+0.45/11 5s} Qe8
{-0.34/11 2s}
{[D]1r2q2k/b7/5p1p/2pP1P1P/5NQ1/6PK/1p6/1R6 w - -}
56. d6 { Of course! This is very strong now, maybe d6 was already possible one move earlier but Pro Deo did not see it yet then. +1.14/12 8s} Rb4 {-1.16/12 9s}
57. Qg6 {+0.98/11 9s} Qxg6 {-2.04/13 4s} 58. hxg6 {+1.30/5
0s} Bb8 {-2.46/15 9s} 59. d7 {+1.97/14 10s} Rd4 {-2.30/16
2s} 60. Nh5 {+1.98/15 9s} Be5 {-2.32/17 9s} 61. Rxb2
{+1.98/15 6s} Rxd7 {-2.32/15 1s} 62. Rb6 {+2.22/15 9s} c4
{-2.40/15 11s} 63. Rc6 {+2.22/14 9s} Rc7 {-2.40/15 4s}
64. g7+ {+2.32/14 9s} Rxg7 {-3.18/18 4s} 65. Nxg7 {+1.94/5
0s} Kxg7 {-3.20/18 1s} 66. Rxc4 {+2.40/16 9s} Kf7 {-4.16/19
9s} 67. Kh4 {+2.60/16 9s} Bd6 {-4.70/19 5s} 68. g4
{+3.08/16 9s} Ke7 {-5.60/17 4s} 69. Re4+ {+3.48/16 14s} Kf7
{-5.60/22 9s} 70. Kh5 {+3.48/16 9s} 1-0

[Event "6 minuten per partij + 6 seconden per zet"]
[Site "Engine Match"]
[Date "2007.06.18"]
[Round "2"]
[White "ProDeo 1.5 Excalibur Test10"]
[Black "Colossus 2007b"]
[Result "1-0"]

1. d4 {Boek 0s} Nf6 {Boek 0s} 2. c4 {Boek 0s} e6 {Boek 0s}
3. Nc3 {Boek 0s} Bb4 {Boek 0s} 4. e3 {Boek 0s} c5 {Boek 0s}
5. Nge2 {Boek 0s} b6 {Boek 0s} 6. a3 {Boek 0s} Ba5 {Boek
0s} 7. Rb1 {Boek 0s} Na6 {Boek 0s} 8. Qa4 {Boek 0s} Bb7
{Boek 0s} 9. Bd2 {Boek 0s} Bc6 {Boek 0s} 10. Qc2 {Boek 0s}
O-O {Boek 0s} 11. Ng3 {Boek 0s} Bb7 {Boek 0s} 12. b4
{+0.28/12 19s} cxb4 {+0.16/14 19s} 13. Na2 {+0.29/14 20s}
b3 {+0.14/14 20s} 14. Rxb3 {+0.22/14 25s} Bxd2+ {+0.14/13
25s} 15. Qxd2 {+0.23/11 3s} Rc8 {+0.14/13 17s} 16. f3
{+0.30/13 15s} Nb8 {+0.02/13 21s} 17. e4 {+0.23/14 21s} Ba6
{+0.04/12 17s} 18. Rc3 {+0.35/14 21s} Ne8 {-0.06/13 13s}
19. Bd3 {+0.32/12 12s} d6 {0.00/12 24s} 20. O-O {+0.48/13
15s} Qh4 {-0.04/13 12s} 21. Nb4 {+0.34/12 11s} Bb7
{-0.12/13 15s} 22. Bc2 {+0.39/12 16s} Nf6 {-0.02/13 19s}
23. Rb1 {+0.36/12 19s} Rfd8 {-0.02/13 20s} 24. Ba4
{+0.33/12 15s} Nc6 {-0.06/12 21s} 25. Qe3 {+0.36/12 14s}
Na5 {-0.04/11 12s} 26. Rbc1 {+0.36/12 9s} Ba8 {-0.10/12
20s} 27. R3c2 {+0.38/12 17s} Bc6 {-0.10/12 19s} 28. Nxc6
{+0.45/12 13s} Nxc6 {-0.22/14 12s} 29. Bb5 {+0.49/12 9s}
Kh8 {-0.10/12 18s} 30. Rb1 {+0.51/12 15s} Nd7 {-0.10/12
17s} 31. Rd2 {+0.53/11 12s} Na5 {0.00/13 14s} 32. Qd3
{+0.48/13 17s} Nb8 {0.00/13 11s} 33. Rc2 {+0.35/13 12s} Rc7
{-0.02/13 10s} 34. Nf1 {+0.45/13 14s} Nbc6 {-0.02/12 9s}
35. Ne3 {+0.39/12 11s} f5 {-0.18/12 14s} 36. g3 {+0.46/13
14s} Qf6 {-0.20/13 15s} 37. Rd2 {+0.37/13 14s} f4 {-0.04/12
9s} 38. gxf4 {+0.23/11 8s} Qxf4 {+0.20/12 9s} 39. Ng2
{+0.18/11 15s} Qh6 {+0.36/13 14s} 40. f4 {+0.08/12 13s} Nb8
{+0.36/13 11s} 41. Rc2 {+0.22/13 11s} Rdc8 {+0.36/13 11s}
42. Rbc1 {0.00/13 12s} a6 {+0.28/12 13s} 43. Ba4 {-0.01/14
9s} Kg8 {+0.20/11 11s} 44. Rc3 {+0.23/13 12s} Qg6 {+0.12/12
10s} 45. Kh1 {+0.17/12 12s} Kh8 {+0.16/11 10s} 46. R3c2
{+0.27/12 11s} h6 {+0.12/12 11s} 47. Ne3 {+0.26/12 8s} Qh5
{0.00/12 9s} 48. Rf1 {+0.31/12 9s} Qh3 {+0.24/13 9s}
49. Qe2 {+0.16/12 6s} Nbc6 {+0.10/11 2s} 50. Bxc6 {+0.07/12
11s} Nxc6 {0.00/13 9s} 51. Rc3 {+0.10/13 6s} Qh4 {-0.02/12
5s} 52. d5 {+0.16/13 11s} Qf6 {-0.04/11 4s} 53. Qc2
{+0.43/13 7s} Na5 {+0.42/10 2s} 54. e5 {+0.26/12 5s} dxe5
{+0.70/13 9s} 55. d6 {-0.29/12 14s} Rc5 {+0.96/12 4s}
56. Ng4 {-0.07/11 6s} Qf7 {+0.36/12 11s} 57. Qe4 {+0.03/11
7s} exf4 {+0.26/11 3s} 58. Rxf4 {-0.05/11 12s} Qd7
{+0.26/11 5s} 59. Ne5 {+0.85/10 6s} Rxe5 {+0.12/13 3s}
60. Qxe5 {+1.04/5 0s} Nc6 {0.00/13 9s} 61. Qe4 {+0.64/13
5s} Qxd6 {0.00/12 2s} 62. Rd3 {+0.34/12 10s} Qe7 {-0.20/13
10s}
{[D]2r4k/4q1p1/ppn1p2p/8/2P1QR2/P2R4/7P/7K w - -}
63. Rg3 { Continues the attack and I think it might be winning in all lines. Colossus does not see 63.Rg3 until it is too late. Colossus eval at 15 ply:
15.01 5:25 +0.80 63.Tg3 Tc7 64.Tfg4 Dd8 65.Dxe6 Pa5
66.Te3 Df8 67.De8 Dxe8 68.Txe8+ Kh7
69.Tee4 g5 70.a4 Pc6 71.Te8 (191.217.460) 604 edb:35

Pro Deo Excalibur at 15 ply:
15.01 3:48 +1.07 63.Tg3 Pa5 64.Tfg4 Db7 65.Dxb7 Pxb7
66.Txg7 Pd6 67.T3g6 Pf5 68.Ta7 Txc4
69.Txe6 Tc1+ 70.Kg2 (164.142.608) 719

+0.66/11 7s} e5 {-0.38/11 4s} 64. Rfg4
{+1.61/12 5s} Rc7 {-0.84/11 2s} 65. Rxg7 {+1.80/11 5s} Qxg7
{-3.74/18 9s} 66. Rxg7 {+1.13/5 0s} Kxg7 {-4.00/15 1s}
67. Qd5 {+2.53/12 7s} b5 {-4.36/16 10s} 68. Qd6 {+3.17/13
9s} Re7 {-4.36/14 2s} 69. cxb5 {+3.88/14 9s} axb5 {-4.74/16
8s} 70. Qxc6 {+3.40/11 1s} e4 {-4.74/14 2s} 71. Kg1
{+4.07/12 5s} e3 {-4.70/13 10s} 72. Kf1 {+4.23/12 10s} Re5
{-4.92/12 2s} 73. Ke2 {+4.58/11 6s} Rg5 {-5.42/12 11s}
74. Kxe3 {+5.42/10 5s} Rh5 {-5.82/11 3s} 75. Qd7+ {+5.41/10
10s} 1-0

The two openings are identical for the first six moves. I did not disable Pro Deo's book, but as the pgn shows {Boek 0s} for Pro Deo's moves as well as for Colossus I think the built in book was not used but I'm only 50% sure about this. Anyway the opening was not decisive in these games. Two times White is a bit unfair though...

Eelco

[Kaj Soderberg]

Thanks for the repost. Very intriguing that the [Pruning = MISC] stuff at least partly still works. I did not dear to use it anymore when fiddling with the settings. But anyway, it is clear that the status of some parameters is unclear. Ed?

BTW, setting Chess Knowledge at higher values seems an improvement on longer time controls and certainly at analysis level. I am playing around with the values 300 and 400 right now in PD15. I could not resist but to try something in Rebel 12 too, and now i have found a nasty engine in the October 2003 version with Chess Knowledge at 400.

Best regards,

Kaj