Discussion about chess-playing software (engines, hosts, opening books, platforms, etc...)
-
kingliveson
- Posts: 1388
- Joined: Thu Jun 10, 2010 1:22 am
- Real Name: Franklin Titus
- Location: 28°32'1"N 81°22'33"W
Post
by kingliveson » Sun Aug 08, 2010 9:34 pm
Dr.Wael Deeb wrote:kingliveson wrote:Final outcome with games played by IvanHoe 9.52b removed due to way too many time losses. 1+0, ponder off, egtb off, quad core.
Third World War 2010
1 Houdini 1.03a x64 POPCNT 4_CP *********** 24.0 - 26.0 27.0 - 23.0 30.0 - 20.0 81.0/150
2 Deep Rybka 4 SSE42 x64 26.0 - 24.0 *********** 24.0 - 26.0 30.5 - 19.5 80.5/150
3 FireBird 1.1 x64 23.0 - 27.0 26.0 - 24.0 *********** 30.5 - 19.5 79.5/150
4 Ivanhoe-BetaWH_33_w64 20.0 - 30.0 19.5 - 30.5 19.5 - 30.5 *********** 59.0/150
game data available upon request.
Yeap,an expected performance even though at long time controls FireBird 1.1 Dr.Deeb I will overtake both Rybka and Houdini by 30-40 Elo margin which is the case in my rating list approaching +400 games....thanks....
Dr.D
Dr.D
My assessment is that clearly these settings are superior to the default. With the meaningless number of games and fire crashing a couple times, they seem to be even. I am now doing another test after correcting IvanHoe 9.52b time manager where it should no longer lose on time.
The new tournament is 2+2 time control. Your settings are used exactly including pawn hash of 64 mb which is half the Hash size used for the tournament. I suspect Fire with these setting could perform even better, but pawn hash might be a problem. Pawn Hash is not one size fit all and should be relative. I posted a small sample test:
http://www.open-chess.org/viewtopic.php?p=4502#p4502
Third World War 2010
1 2 3 4
1 Houdini 1.03a x64 POPCNT 4_CPU ****************** 110½0½½1½01½½½½½1½ 1½½½½0½½½0½1½½1½01 1½½½1½1½½1½½110110 31.5/54
2 Deep Rybka 4 SSE42 x64 001½1½½0½10½½½½½0½ ****************** 0½0½½½½011½0½½1½10 0½0½½½½1½011111½0½ 26.5/54
3 IvanHoe 9.52b x64 0½½½½1½½½1½0½½0½10 1½1½½½½100½1½½0½01 ****************** 1½½0½0½0½½½½0½½½½½ 25.5/54
4 FireBird 1.1 x64 0½½½0½0½½0½½001001 1½1½½½½0½100000½1½ 0½½1½1½1½½½½1½½½½½ ****************** 24.5/54
PAWN : Knight >> Bishop >> Rook >>Queen
-
slobo
- Posts: 36
- Joined: Sun Jun 13, 2010 2:09 am
Post
by slobo » Mon Aug 09, 2010 1:16 am
8/8/2010 20:44:30 :
AFTER 40 GAMES
Individual statistics:
1 IvanHoe 054_x64_4cpu : 2543 40 (+ 10,= 25,- 5), 56.2 %
FB 101_w32_4cpu : 2 (+ 1,= 1,- 0), 75.0 %
IvanHoe 065b_prefetch_w32_4cpu: 2 (+ 0,= 1,- 1), 25.0 %
IvanHoe T-63Mini10_w64_4cpu : 2 (+ 1,= 0,- 1), 50.0 %
IvanHoe 063b_prefetch_w64_4cpu: 2 (+ 1,= 1,- 0), 75.0 %
Fire_13_w64_4cpu : 2 (+ 2,= 0,- 0), 100.0 %
Houdini_101_w64_4cpu : 4 (+ 1,= 3,- 0), 62.5 %
Stockfish-18-ja-w64-4cpu : 2 (+ 1,= 1,- 0), 75.0 %
Rybka 4_w64_4cpu : 2 (+ 1,= 1,- 0), 75.0 %
Fire_131_KLO_popcnt_x64_4cpu : 2 (+ 1,= 1,- 0), 75.0 %
FB_11-in_WD_x64_4cpu : 8 (+ 0,= 7,- 1), 43.8 %
Houdini 103_x64_4CPU : 4 (+ 0,= 4,- 0), 50.0 %
Houdini 103_POPCNT_x64_4CPU : 4 (+ 1,= 2,- 1), 50.0 %
FB_11-in_WD_popcnt_x64_4cpu : 4 (+ 0,= 3,- 1), 37.5 %
Program Elo + - Games Score Av.Op. Draws
1 IvanHoe 054_x64_4cpu : 2543 68 65 40 56.2 % 2499 62.5 %
2 IvanHoe T-63Mini10_w64_4cpu : 2542 66 62 40 57.5 % 2490 65.0 %
3 IvanHoe-BetaWH_24_AMD64_4cpu : 2533 66 62 40 57.5 % 2481 65.0 %
4 FB_11-in_WD_popcnt_x64_4cpu : 2530 70 68 40 55.0 % 2495 60.0 %
5 Houdini 103_POPCNT_x64_4CPU : 2526 67 66 40 53.8 % 2500 62.5 %
6 Houdini_101_w64_4cpu : 2517 73 70 36 55.6 % 2479 61.1 %
7 IvanHoe-T63Mini9_w64_4cpu : 2516 73 69 36 58.3 % 2457 61.1 %
8 FB_11-in_WD_x64_4cpu : 2514 70 68 30 51.7 % 2503 70.0 %
9 IvanHoe 065b_prefetch_w32_4cpu : 2509 54 47 38 55.3 % 2472 78.9 %
10 IvanHoe 063b_prefetch_w64_4cpu : 2507 59 53 30 53.3 % 2484 80.0 %
11 Fire_13_w64_4cpu : 2504 55 54 38 51.3 % 2495 76.3 %
12 Houdini 103_x64_4CPU : 2494 61 61 32 50.0 % 2494 75.0 %
13 IvanHoe 063b_default_w32_4cpu : 2489 62 60 50 58.0 % 2433 60.0 %
14 Rybka 4_w64_4cpu : 2487 84 83 30 51.7 % 2476 56.7 %
15 IvanHoe 054-2_x64_4cpu : 2486 61 61 32 50.0 % 2486 75.0 %
16 IvanHoe 066b_prefetch_w32_4cpu : 2475 88 86 30 56.7 % 2428 53.3 %
17 IvanHoe T-63Mini5_w32_4cpu : 2472 53 53 32 50.0 % 2472 81.2 %
18 IvanHoe-BetaWH_32_AMD64_4cpu : 2468 63 66 30 46.7 % 2491 73.3 %
19 IvanHoe 052a_popcnt_x64_4cpu : 2466 65 65 30 50.0 % 2466 73.3 %
20 IvanHoe 052_alpha_x64_4cpu : 2466 49 53 30 48.3 % 2478 83.3 %
21 Fire_12_new_SMP_KLO_w64_4cpu : 2466 63 66 30 46.7 % 2490 73.3 %
22 FB 12_w32_4cpu : 2463 52 46 34 52.9 % 2442 82.4 %
23 IvanHoe 065.02-JR_w32_4cpu : 2462 62 60 30 51.7 % 2451 76.7 %
24 FB 11-ms_w32_4cpu : 2458 62 60 30 51.7 % 2446 76.7 %
25 Fire_13_POPCNT_w64_4cpu : 2452 73 77 26 46.2 % 2478 69.2 %
26 IvanHoe 063b_default_w64_4cpu : 2451 72 74 30 46.7 % 2474 66.7 %
27 Igorrit 0086v9_w32_2cpu : 2441 74 70 38 60.5 % 2367 57.9 %
28 Stockfish-18-ja-w64-4cpu : 2440 89 90 30 45.0 % 2475 50.0 %
29 Igorrit 0086v9_w32_4cpu : 2438 70 68 30 51.7 % 2427 70.0 %
30 FB 11-NS_w32 : 2434 79 81 30 46.7 % 2457 60.0 %
31 IvanHoe 064b_prefetch_w32_4cpu : 2430 78 78 28 50.0 % 2430 64.3 %
32 Rybka 3 w64_4cpu : 2428 57 62 38 43.4 % 2474 71.1 %
33 IvanHoe 063b_prefetch_w32_4cpu : 2426 57 64 30 45.0 % 2461 76.7 %
34 IvanHoe 065b_default_w32_4cpu : 2425 57 63 28 46.4 % 2450 78.6 %
35 IvanHoe v70-JR_w32_4cpu : 2419 77 76 30 51.7 % 2407 63.3 %
36 Fire_131_KLO_popcnt_x64_4cpu : 2417 73 81 28 39.3 % 2492 64.3 %
37 Rybka 3 w32_4cpu : 2386 46 47 74 45.3 % 2419 66.2 %
38 FB 101_w32_4cpu : 2385 75 80 28 42.9 % 2435 64.3 %
39 Igorrit 0086v7_w32_2cpu : 2375 87 85 28 53.6 % 2350 57.1 %
40 FB 10_w32_4cpu : 2372 69 69 36 48.6 % 2381 63.9 %
-
robbolito
- Posts: 601
- Joined: Thu Jun 10, 2010 3:48 am
Post
by robbolito » Mon Aug 09, 2010 2:54 am
AMD Phenom(tm) II X6 1090T Processor6x @ 3.2 GHz 4,096 MB Memory
Windows 7 Home Premium Edition (Build 7600)
Fritz Benchmark:
Speed: 23.33
KNS: 11196
GUI: CB Rybka 3
Book: HS 7 moves Plus
Hash: 256
RB and TB: ON
Ponder: OFF
Blitz:10' 0
Code: Select all
1234567890
1 FireBird 1.1 x64 3190 10½0½½½½11 5.5/10
2 Deep Rybka 4 x64 3190 01½1½½½½00 4.5/10
For simple curiosity I have done quick 10 games to see what other people are talking about and not to my surprise these 10 games have confirmed that Fire 1.1 with Dr.Ds settings is a very powerful engine.
I have tested the original Fire 1.1 before but this one being the same except the setting has acted as the original and caused exception at the 9th game so I had to restart the match to get the final 10 games.If the problem with the exceptions could be resolved we would have one of the best chess engines in the world.
Again my thanks to Norman and Dr.D.
-
Dr.Wael Deeb
- Posts: 104
- Joined: Thu Jun 10, 2010 8:29 pm
- Real Name: Dr.Wael Deeb
Post
by Dr.Wael Deeb » Mon Aug 09, 2010 8:12 am
kingliveson wrote:Dr.Wael Deeb wrote:kingliveson wrote:Final outcome with games played by IvanHoe 9.52b removed due to way too many time losses. 1+0, ponder off, egtb off, quad core.
Third World War 2010
1 Houdini 1.03a x64 POPCNT 4_CP *********** 24.0 - 26.0 27.0 - 23.0 30.0 - 20.0 81.0/150
2 Deep Rybka 4 SSE42 x64 26.0 - 24.0 *********** 24.0 - 26.0 30.5 - 19.5 80.5/150
3 FireBird 1.1 x64 23.0 - 27.0 26.0 - 24.0 *********** 30.5 - 19.5 79.5/150
4 Ivanhoe-BetaWH_33_w64 20.0 - 30.0 19.5 - 30.5 19.5 - 30.5 *********** 59.0/150
game data available upon request.
Yeap,an expected performance even though at long time controls FireBird 1.1 Dr.Deeb I will overtake both Rybka and Houdini by 30-40 Elo margin which is the case in my rating list approaching +400 games....thanks....
Dr.D
Dr.D
My assessment is that clearly these settings are superior to the default. With the meaningless number of games and fire crashing a couple times, they seem to be even. I am now doing another test after correcting IvanHoe 9.52b time manager where it should no longer lose on time.
The new tournament is 2+2 time control. Your settings are used exactly including pawn hash of 64 mb which is half the Hash size used for the tournament. I suspect Fire with these setting could perform even better, but pawn hash might be a problem. Pawn Hash is not one size fit all and should be relative. I posted a small sample test:
http://www.open-chess.org/viewtopic.php?p=4502#p4502
Third World War 2010
1 2 3 4
1 Houdini 1.03a x64 POPCNT 4_CPU ****************** 110½0½½1½01½½½½½1½ 1½½½½0½½½0½1½½1½01 1½½½1½1½½1½½110110 31.5/54
2 Deep Rybka 4 SSE42 x64 001½1½½0½10½½½½½0½ ****************** 0½0½½½½011½0½½1½10 0½0½½½½1½011111½0½ 26.5/54
3 IvanHoe 9.52b x64 0½½½½1½½½1½0½½0½10 1½1½½½½100½1½½0½01 ****************** 1½½0½0½0½½½½0½½½½½ 25.5/54
4 FireBird 1.1 x64 0½½½0½0½½0½½001001 1½1½½½½0½100000½1½ 0½½1½1½1½½½½1½½½½½ ****************** 24.5/54
I use 64 Mb Pawn Hash with 512 Mb for the program which translates to 1/8 of the hash size....I found this to be the most optimal value....
I don't have any crashes with FireBird 1.1 nor with several IvanHoe compiles I use....
Dr.D
-
Dr.Wael Deeb
- Posts: 104
- Joined: Thu Jun 10, 2010 8:29 pm
- Real Name: Dr.Wael Deeb
Post
by Dr.Wael Deeb » Mon Aug 09, 2010 8:15 am
robbolito wrote:AMD Phenom(tm) II X6 1090T Processor6x @ 3.2 GHz 4,096 MB Memory
Windows 7 Home Premium Edition (Build 7600)
Fritz Benchmark:
Speed: 23.33
KNS: 11196
GUI: CB Rybka 3
Book: HS 7 moves Plus
Hash: 256
RB and TB: ON
Ponder: OFF
Blitz:10' 0
Code: Select all
1234567890
1 FireBird 1.1 x64 3190 10½0½½½½11 5.5/10
2 Deep Rybka 4 x64 3190 01½1½½½½00 4.5/10
For simple curiosity I have done quick 10 games to see what other people are talking about and not to my surprise these 10 games have confirmed that Fire 1.1 with Dr.Ds settings is a very powerful engine.
I have tested the original Fire 1.1 before but this one being the same except the setting has acted as the original and caused exception at the 9th game so I had to restart the match to get the final 10 games.If the problem with the exceptions could be resolved we would have one of the best chess engines in the world.
Again my thanks to Norman and Dr.D.
You welcome and thanks for taking the time to test
Even the number of games is small,it shows to some degree the power of these settings....
Dr.D
-
TPJR
- Posts: 14
- Joined: Thu Jul 08, 2010 1:42 pm
- Location: Switzerland
Post
by TPJR » Mon Aug 09, 2010 12:12 pm
Hi Dr. Deeb,
maybe Pawn Hash should indeed be assigned relatively to Hash as kingliveson suggests. I asked gaard in this treat
http://www.open-chess.org/viewtopic.php ... 5062#p5062
to make his test of your Firebird again, but with PawnHash=1/8 * Hash.
Thomas
-
xshat
- Posts: 98
- Joined: Sat Jun 19, 2010 10:44 am
Post
by xshat » Tue Aug 10, 2010 6:48 am
I still want to test at least 50 more games when I have time, but the first 50 games went like this:
Fire 1.31 Ns compile = 13
Firebird 1.1 DD Settings = 10
Draws = 27
So far its really close but 1.31 has a slight edge. More games should yield more accurate results.
-
Canto
- Posts: 13
- Joined: Mon Jul 12, 2010 12:30 am
Post
by Canto » Wed Aug 11, 2010 2:00 am
TPJR wrote:Hi Dr. Deeb,
maybe Pawn Hash should indeed be assigned relatively to Hash as kingliveson suggests. I asked gaard in this treat
http://www.open-chess.org/viewtopic.php ... 5062#p5062
to make his test of your Firebird again, but with PawnHash=1/8 * Hash.
Thomas
Yeah, I'd like to see some test results for Pawn Hash = 128. I was not all that impressed with Dr Deeb's settings until I set Pawn Hash = 128, which on my humble test set did indeed impress me. With these settings I wouldn't be surprised if Firebird 1.1_Dr_D is tops.
-
kingliveson
- Posts: 1388
- Joined: Thu Jun 10, 2010 1:22 am
- Real Name: Franklin Titus
- Location: 28°32'1"N 81°22'33"W
Post
by kingliveson » Sun Aug 15, 2010 3:46 pm
After 18 rounds of 25 with 84 games remaining. This time, FireBird is using Pawn Hash 1/8th of Hash. Time control is 4+2. Final standing will be posted on tournament sub-forum.
Third World War 2010
1 Houdini 1.03a x64 POPCNT 4_CPU *********** 19.0 - 17.0 20.0 - 16.0 18.5 - 17.5 57.5/108
2 FireBird 1.1 x64 17.0 - 19.0 *********** 17.0 - 19.0 20.0 - 16.0 54.0/108
3 IvanHoe 9.52b x64 16.0 - 20.0 19.0 - 17.0 *********** 18.0 - 18.0 53.0/108
4 Deep Rybka 4 SSE42 x64 17.5 - 18.5 16.0 - 20.0 18.0 - 18.0 *********** 51.5/108
PAWN : Knight >> Bishop >> Rook >>Queen
-
Dr.Wael Deeb
- Posts: 104
- Joined: Thu Jun 10, 2010 8:29 pm
- Real Name: Dr.Wael Deeb
Post
by Dr.Wael Deeb » Sun Aug 15, 2010 4:25 pm
kingliveson wrote:After 18 rounds of 25 with 84 games remaining. This time, FireBird is using Pawn Hash 1/8th of Hash. Time control is 4+2. Final standing will be posted on tournament sub-forum.
Third World War 2010
1 Houdini 1.03a x64 POPCNT 4_CPU *********** 19.0 - 17.0 20.0 - 16.0 18.5 - 17.5 57.5/108
2 FireBird 1.1 x64 17.0 - 19.0 *********** 17.0 - 19.0 20.0 - 16.0 54.0/108
3 IvanHoe 9.52b x64 16.0 - 20.0 19.0 - 17.0 *********** 18.0 - 18.0 53.0/108
4 Deep Rybka 4 SSE42 x64 17.5 - 18.5 16.0 - 20.0 18.0 - 18.0 *********** 51.5/108
Yeap,looking good thus far even though it's not optimized for this blitz time control....
Dr.D
