Starting This Weekend- Another High-Powered Match!

[shrapnel]

Are both Engines at Default Settings ?

[geots]

Are both Engines at Default Settings ?

Almost. I always run at default settings. I assume the programmer is smarter than me. But with Stockfish development versions, you need to UNTICK the box for "Use Sleeping Threads". Marco says the box is unchecked for all official releases, but not for the dev. versions. If you don't uncheck it, and you try to give the engine all cores available- System Idle Process will eat up part of a core and keep Stockfish from using, in my case the full 6 cores.


Best,

[geots]

I am no one's advertising boy, but it is becoming apparent to me that there is a possibility- I said possibility- that Komodo MP could be the best bang for your buck in the history of commercial computer chess. To illustrate a point, a weaker Komodo beta beat Deep Fritz 13 and Deep Junior 13.3 by a combined score of 40-0. BTW, how much did you pay for Junior and Fritz?

One point that people need to put much more emphasis on is a rating list- or lack of it. You have engines 1 thru 10. Number 1 and Number 4 likely got there with very few games against each other. Possibly none. Number 1 may do much better against the rest of the top 10 than number 4 does. But Number 4 may have a particular style of play that Number 1 cannot deal with. So if 4 beats Number 1- which should be the champion? I am not sure- but it is certainly food for thought. At any rate............................





Alienware AURORA_R4
Intel i7 w/6 True Cores
Factory overclocked
Fritz 11 gui
6 Cores each/64bit
256MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 2012b.ctg w/12-move limit
40/19 Repeating- benchmarked to adapt to 40/40
Match=100 games

Code: Select all

Komodo 5.1r2 64-bit          +5/-4/=25 
Stockfish 290613 64 SSE4.2   +4/-5/=25

More to come-

[geots]

5 more games finished and we are at the 39 game mark. Komodo 5.1r2 has managed to stretch his lead to 3 games, but with 61 games still to play I am not going to over-emphasize the meaning of it. Tho any lead is better than no lead at all. The million dollar question, with this dev. version of Stockfish being between 35 and 50 elo stronger than the last official Stockfish release- is just exactly how high can they go with elo increases. Well, enough talk and now to the update:




Alienware AURORA_R4
Intel i7 w/6 True Cores
Factory overclocked
Fritz 11 gui
6 Cores each/64bit
256MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 2012b.ctg w/12-move limit
40/19 Repeating- benchmarked to adapt to 40/40
Match=100 games

Code: Select all

Komodo 5.1r2 64-bit          +7/-4/=28  
Stockfish 290613 64 SSE4.2   +4/-7/=28

(39 games)

See you after some zzs-

[PaulieD]

5 more games finished and we are at the 39 game mark. Komodo 5.1r2 has managed to stretch his lead to 3 games, but with 61 games still to play I am not going to over-emphasize the meaning of it. Tho any lead is better than no lead at all. The million dollar question, with this dev. version of Stockfish being between 35 and 50 elo stronger than the last official Stockfish release- is just exactly how high can they go with elo increases. Well, enough talk and now to the update:




Alienware AURORA_R4
Intel i7 w/6 True Cores
Factory overclocked
Fritz 11 gui
6 Cores each/64bit
256MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 2012b.ctg w/12-move limit
40/19 Repeating- benchmarked to adapt to 40/40
Match=100 games

Code: Select all

Komodo 5.1r2 64-bit          +7/-4/=28  
Stockfish 290613 64 SSE4.2   +4/-7/=28

(39 games)

See you after some zzs-

It looks like the July versions are even stronger than the June versions...

500 games from recent Stockfish developmental versions

i3 Dual Core @2.53 GHz - 6.0GB RAM
W7 Home Premium X.64 SP-1
GUI: Little Blitzer 2.74
HASH: 64
TABLEBASES: None
BOOK: Nunn 2 Test Suite Reversing Colors
PONDER: OFF
Code:
Games Completed = 500 of 500 (Avg game length = 24.964 sec)
Settings = RR/64MB/17000ms in 200 moves/M 1000cp for 6 moves, D 150 moves/EPD:C:\Users\Paul\Documents\ChessBase\EPD Test Suites\Nunn2 20 position.epd(20)
Time = 13081 sec elapsed, 0 sec remaining
1. Stockfish 062313 122.0/251 50-57-144 (L: m=0 t=0 i=0 a=57) (D: r=120 i=17 f=2 s=3 a=2) (tpm=216.2 d=14.14 nps=1383321)
2. Stockfish 062913 121.5/251 45-53-153 (L: m=4 t=0 i=0 a=49) (D: r=121 i=19 f=8 s=2 a=3) (tpm=203.9 d=14.19 nps=1379037)
3. Stockfish 071113 130.5/250 65-54-131 (L: m=1 t=0 i=0 a=53) (D: r=109 i=14 f=6 s=2 a=0) (tpm=204.8 d=14.25 nps=1394732)
4. Stockfish 071313 126.0/248 52-48-148 (L: m=0 t=0 i=0 a=48) (D: r=114 i=22 f=8 s=1 a=3) (tpm=212.0 d=14.16 nps=1342525)

[geots]

5 more games finished and we are at the 39 game mark. Komodo 5.1r2 has managed to stretch his lead to 3 games, but with 61 games still to play I am not going to over-emphasize the meaning of it. Tho any lead is better than no lead at all. The million dollar question, with this dev. version of Stockfish being between 35 and 50 elo stronger than the last official Stockfish release- is just exactly how high can they go with elo increases. Well, enough talk and now to the update:




Alienware AURORA_R4
Intel i7 w/6 True Cores
Factory overclocked
Fritz 11 gui
6 Cores each/64bit
256MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 2012b.ctg w/12-move limit
40/19 Repeating- benchmarked to adapt to 40/40
Match=100 games

Code: Select all

Komodo 5.1r2 64-bit          +7/-4/=28  
Stockfish 290613 64 SSE4.2   +4/-7/=28

(39 games)

See you after some zzs-

It looks like the July versions are even stronger than the June versions...

500 games from recent Stockfish developmental versions

i3 Dual Core @2.53 GHz - 6.0GB RAM
W7 Home Premium X.64 SP-1
GUI: Little Blitzer 2.74
HASH: 64
TABLEBASES: None
BOOK: Nunn 2 Test Suite Reversing Colors
PONDER: OFF
Code:
Games Completed = 500 of 500 (Avg game length = 24.964 sec)
Settings = RR/64MB/17000ms in 200 moves/M 1000cp for 6 moves, D 150 moves/EPD:C:\Users\Paul\Documents\ChessBase\EPD Test Suites\Nunn2 20 position.epd(20)
Time = 13081 sec elapsed, 0 sec remaining
1. Stockfish 062313 122.0/251 50-57-144 (L: m=0 t=0 i=0 a=57) (D: r=120 i=17 f=2 s=3 a=2) (tpm=216.2 d=14.14 nps=1383321)
2. Stockfish 062913 121.5/251 45-53-153 (L: m=4 t=0 i=0 a=49) (D: r=121 i=19 f=8 s=2 a=3) (tpm=203.9 d=14.19 nps=1379037)
3. Stockfish 071113 130.5/250 65-54-131 (L: m=1 t=0 i=0 a=53) (D: r=109 i=14 f=6 s=2 a=0) (tpm=204.8 d=14.25 nps=1394732)
4. Stockfish 071313 126.0/248 52-48-148 (L: m=0 t=0 i=0 a=48) (D: r=114 i=22 f=8 s=1 a=3) (tpm=212.0 d=14.16 nps=1342525)

Paul, some nice work you have done. I only have one issue with it. Time controls such as the one you used here are fine, and they tell a programmer a lot about his engine during the beta testing process. He needs to get as many games as he possibly can. The problem is, there are a handful of engines that play much, much better at longer controls like 40/40 repeating or even the longer controls that FIDE uses. And Stockfish and Komodo are the 2 leaders in that respect. I know no engines that increase in strength with the longer controls anywhere close to the way Komodo and Stockfish do. The other that would come to mind that might be close to them is Deep Rybka 4.1. It too benefits greatly from the longer controls.

Again, there is not one thing wrong with your controls and your work. It just so happens that you are honing in on Stockfish in your tests- and for the reasons I gave- if you would use longer controls you would get a much better idea of the quality of play that Stockfish possesses. Just a thought you might consider.



All the best to you,

gts

[geots]

One game over the halfway mark, and nothing is coming easy for either of these engines. In these last 12 games, Komodo 5.1r2 has managed to increase his lead by one more game. Obviously there is plenty of time left for both these engines, but Stockfish needs to put together a couple of back-to-backs to even things up here. So here we are................................





Alienware AURORA_R4
Intel i7 w/6 True Cores
Factory overclocked
Fritz 11 gui
6 Cores each/64bit
256MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 2012b.ctg w/12-move limit
40/19 Repeating- benchmarked to adapt to 40/40
Match=100 games

Code: Select all

Komodo 5.1r2 64-bit          +10/-6/=35   
Stockfish 290613 64 SSE4.2   +6/-10/=35

(51 games) 

See you shortly-

[Modern Times]

I predict that Komodo will win this match quite comfortably.

[PaulieD]

5 more games finished and we are at the 39 game mark. Komodo 5.1r2 has managed to stretch his lead to 3 games, but with 61 games still to play I am not going to over-emphasize the meaning of it. Tho any lead is better than no lead at all. The million dollar question, with this dev. version of Stockfish being between 35 and 50 elo stronger than the last official Stockfish release- is just exactly how high can they go with elo increases. Well, enough talk and now to the update:




Alienware AURORA_R4
Intel i7 w/6 True Cores
Factory overclocked
Fritz 11 gui
6 Cores each/64bit
256MB hash
Bases=NONE
Ponder_Learning=OFF
Perfect 2012b.ctg w/12-move limit
40/19 Repeating- benchmarked to adapt to 40/40
Match=100 games

Code: Select all

Komodo 5.1r2 64-bit          +7/-4/=28  
Stockfish 290613 64 SSE4.2   +4/-7/=28

(39 games)

See you after some zzs-

It looks like the July versions are even stronger than the June versions...

500 games from recent Stockfish developmental versions

i3 Dual Core @2.53 GHz - 6.0GB RAM
W7 Home Premium X.64 SP-1
GUI: Little Blitzer 2.74
HASH: 64
TABLEBASES: None
BOOK: Nunn 2 Test Suite Reversing Colors
PONDER: OFF
Code:
Games Completed = 500 of 500 (Avg game length = 24.964 sec)
Settings = RR/64MB/17000ms in 200 moves/M 1000cp for 6 moves, D 150 moves/EPD:C:\Users\Paul\Documents\ChessBase\EPD Test Suites\Nunn2 20 position.epd(20)
Time = 13081 sec elapsed, 0 sec remaining
1. Stockfish 062313 122.0/251 50-57-144 (L: m=0 t=0 i=0 a=57) (D: r=120 i=17 f=2 s=3 a=2) (tpm=216.2 d=14.14 nps=1383321)
2. Stockfish 062913 121.5/251 45-53-153 (L: m=4 t=0 i=0 a=49) (D: r=121 i=19 f=8 s=2 a=3) (tpm=203.9 d=14.19 nps=1379037)
3. Stockfish 071113 130.5/250 65-54-131 (L: m=1 t=0 i=0 a=53) (D: r=109 i=14 f=6 s=2 a=0) (tpm=204.8 d=14.25 nps=1394732)
4. Stockfish 071313 126.0/248 52-48-148 (L: m=0 t=0 i=0 a=48) (D: r=114 i=22 f=8 s=1 a=3) (tpm=212.0 d=14.16 nps=1342525)

Paul, some nice work you have done. I only have one issue with it. Time controls such as the one you used here are fine, and they tell a programmer a lot about his engine during the beta testing process. He needs to get as many games as he possibly can. The problem is, there are a handful of engines that play much, much better at longer controls like 40/40 repeating or even the longer controls that FIDE uses. And Stockfish and Komodo are the 2 leaders in that respect. I know no engines that increase in strength with the longer controls anywhere close to the way Komodo and Stockfish do. The other that would come to mind that might be close to them is Deep Rybka 4.1. It too benefits greatly from the longer controls.

Again, there is not one thing wrong with your controls and your work. It just so happens that you are honing in on Stockfish in your tests- and for the reasons I gave- if you would use longer controls you would get a much better idea of the quality of play that Stockfish possesses. Just a thought you might consider.



All the best to you,

gts

George,
I understand about longer time controls, but in this case it is all the same engine (Stockfish) so ultra fast is fine to test the versions against each other to see which is better for selecting the version to put into a longer time control match like yours.
I do know that both Komodo and Stockfish are not as good at blitz or better when competing against lets say Houdini or other very strong engines.

[geots]

I predict that Komodo will win this match quite comfortably.

Maybe it depends on your definition of "comfortably"- but at this point I would tend to agree.


Best,