Some fun with Komodo 8

[Laskos]

Another interesting point would be to compare with a top dog of 2002 on SSDF, Deep Shredder 6.

Deep Shredder 6 (2002 IIRC)

Code: Select all

  TotTime: 115:59m    SolTime: 115:59m
  Ply: 0   Positions:150   Avg Nodes:       0   Branching = 0.00
  Ply: 1   Positions:150   Avg Nodes:      42   Branching = 0.00
  Ply: 2   Positions:150   Avg Nodes:     265   Branching = 6.31
  Ply: 3   Positions:150   Avg Nodes:    1207   Branching = 4.55
  Ply: 4   Positions:150   Avg Nodes:    4065   Branching = 3.37
  Ply: 5   Positions:150   Avg Nodes:   12228   Branching = 3.01
  Ply: 6   Positions:150   Avg Nodes:   44043   Branching = 3.60
  Ply: 7   Positions:150   Avg Nodes:  123814   Branching = 2.81
  Ply: 8   Positions:150   Avg Nodes:  354372   Branching = 2.86
  Ply: 9   Positions:150   Avg Nodes:  884464   Branching = 2.50
  Ply:10   Positions:150   Avg Nodes: 2334303   Branching = 2.64

EBF last 5 plies is 2.86, higher than Komodo 8 without LMR and Null Move.
__________________________
Score of Komodo 8 depth 10 vs Shredder 6 depth 10: 37 - 41 - 22 [0.48] 100
ELO difference: -14

Approximately equal to the same depth.
Komodo 8 average nodes on depth 10: 64,039
Shredder 6 average nodes on depth 10: 2,334,303

1/ The value as strength goes of a ply didn't change much since 2002, but present engines go ~2 times deeper on the same TC and hardware, and prune MUCH more.
2/ With x40 more nodes at fixed depth 10, Shredder 6 of 2002 is on a par with Komodo 8 of 2014. The nodes became much more valuable to the present time.
3/ The node in Shredder 6 is ~2 times slower than in Komodo 8, so total time advantage to the same strength of Komodo 8 is a factor of ~80 compared to Shredder 6.

[bob]

Another interesting point would be to compare with a top dog of 2002 on SSDF, Deep Shredder 6.

Deep Shredder 6 (2002 IIRC)

Code: Select all

  TotTime: 115:59m    SolTime: 115:59m
  Ply: 0   Positions:150   Avg Nodes:       0   Branching = 0.00
  Ply: 1   Positions:150   Avg Nodes:      42   Branching = 0.00
  Ply: 2   Positions:150   Avg Nodes:     265   Branching = 6.31
  Ply: 3   Positions:150   Avg Nodes:    1207   Branching = 4.55
  Ply: 4   Positions:150   Avg Nodes:    4065   Branching = 3.37
  Ply: 5   Positions:150   Avg Nodes:   12228   Branching = 3.01
  Ply: 6   Positions:150   Avg Nodes:   44043   Branching = 3.60
  Ply: 7   Positions:150   Avg Nodes:  123814   Branching = 2.81
  Ply: 8   Positions:150   Avg Nodes:  354372   Branching = 2.86
  Ply: 9   Positions:150   Avg Nodes:  884464   Branching = 2.50
  Ply:10   Positions:150   Avg Nodes: 2334303   Branching = 2.64

EBF last 5 plies is 2.86, higher than Komodo 8 without LMR and Null Move.
__________________________
Score of Komodo 8 depth 10 vs Shredder 6 depth 10: 37 - 41 - 22 [0.48] 100
ELO difference: -14

Approximately equal to the same depth.
Komodo 8 average nodes on depth 10: 64,039
Shredder 6 average nodes on depth 10: 2,334,303

1/ The value as strength goes of a ply didn't change much since 2002, but present engines go ~2 times deeper on the same TC and hardware, and prune MUCH more.
2/ With x40 more nodes at fixed depth 10, Shredder 6 of 2002 is on a par with Komodo 8 of 2014. The nodes became much more valuable to the present time.
3/ The node in Shredder 6 is ~2 times slower than in Komodo 8, so total time advantage to the same strength of Komodo 8 is a factor of ~80 compared to Shredder 6.

Don't forget futility pruning which has also been ramped up by everybody as the depths have climbed. I do it in the last 6 plies, where 10 years ago most were doing it in last 2-3 plies ala' Heinz's paper.

[Laskos]

And some fun with direct strength, at 60'' + 0.15'', showing some 920 Elo points difference, or, alternatively, Komodo 8 and a recent Stockfish as perfect players

Code: Select all

Games Completed = 100 of 100 (Avg game length = 115.240 sec)
Settings = Gauntlet/32MB/60000ms+150ms/M 400000cp for 1000 moves, D 120000 moves/PGN:C:\Users\Kai\Downloads\LittleBlitzer\swcr.pgn(5120)
Time = 11770 sec elapsed, 0 sec remaining
 1.  Komodo 8                 	99.5/100	99-0-1  	(L: m= 0 t=0 i=0 a=0)	(D: r=1 i=0 f=0 s=0 a=0)	(tpm=1061.8 d=13.27 nps=370196)
 2.  Shredder 6               	 0.5/100	0-99-1  	(L: m=99 t=0 i=0 a=0)	(D: r=1 i=0 f=0 s=0 a=0)	(tpm=1285.6 d= 8.47 nps=308841)

Games Completed = 100 of 100 (Avg game length = 113.095 sec)
Settings = Gauntlet/32MB/60000ms+150ms/M 400000cp for 1000 moves, D 120000 moves/PGN:C:\Users\Kai\Downloads\LittleBlitzer\swcr.pgn(5120)
Time = 11559 sec elapsed, 0 sec remaining
 1.  SF 210914                	99.5/100	99-0-1  	(L: m= 0 t=0 i=0 a=0)	(D: r=1 i=0 f=0 s=0 a=0)	(tpm=1009.8 d=22.69 nps=436169)
 2.  Shredder 6               	 0.5/100	0-99-1  	(L: m=99 t=0 i=0 a=0)	(D: r=1 i=0 f=0 s=0 a=0)	(tpm=1291.2 d= 8.46 nps=316436)

One ca see everything here, the depths are now ~2 times higher, the hardware or time control cannot be accused. Great progress.