Not logged inRybka Chess Community Forum
- - By Lazy Frank (****) [gb] Date 2018-10-20 07:52
I dont know why but it seems also TCEC now clear a rows of followers.
Nice! Maybe because Leela's dream can't come true?
Parent - By Peter Grayson (****) [gb] Date 2018-10-20 12:06

> Maybe because Leela's dream can't come true?


The dream is distant from the reality no matter how optimistic are its proponents!
- - By Lazy Frank (****) [gb] Date 2018-11-08 10:33
Some statistics about TCEC 13 superfinal (only decisive games):
Parent - - By Labyrinth (*****) [us] Date 2018-11-08 11:04
It's amazing to me that there could be so much variance between two engines of very comparable strength at such a high level.
Parent - By Lazy Frank (****) [gb] Date 2018-11-08 11:14
Next one: longest moves.
Parent - - By Venator (Silver) [nl] Date 2018-11-08 15:58
Thanks for these statistics. Is it also possible for you to extract the following information? See:

For all drawn games: what was the highest eval of SF and Komodo during the game?

IMO both SF and Komodo defended difficult positions extremely well, saving a lot of precarious situations.
Parent - By Lazy Frank (****) [gb] Date 2018-11-08 18:04 Edited 2018-11-08 18:55
I can say highest eval and losing the game.

For Stockfish it's game 75, move 14, Rb1, eval 0,83.

For Komodo it's game 98, move 12, Bf2, eval 1,00.

BTW, as you see g4 really not a Stockfish signature move, rather h4. :twisted:

Also here is castle information:

Stockfish castled short 4x and long 3x, while Komodo 6x castle short and 7x long.

Biggest difference is how rare Stockfish castled at all compared to Komodo (7 vs 13).
Parent - By gsgs (***) [de] Date 2018-11-09 11:18 Edited 2018-11-09 12:38

> For all drawn games: what was the highest eval of SF and Komodo during the game?


highest eval in drawn games was +2.79 in game 43
lowest eval in drawn games was -1.88 in game 70

SF:+2.79(43),-1.62(33)  ;  +1.59(6),-1.88(70)
KO:+1.65(6),-1.10(70)  ;  +1.14(43),-1.80(33)

white-based-evals , SF has White in odd-numbered-games

average-eval-chart : http://magictour.free.fr/tcec13f.GIF
- By Lazy Frank (****) [gb] Date 2018-11-08 11:37
Ponder rate:

Stockfish - 815 out of 1181 (69%)

Komodo - 853 out of 1179 (72,35%)

That gives match average ponder rate - 70,678% precision.

Most predictable game was 95 (over 82%), most unpredictable was game 17 (just 60%) :lol:.

P.S. For me it's well known fact (last 3 years) Komodo always have higher ponder rate than Stockfish.
Reason - Komodo playing style is to limit opponent only to one reasonable move.
- - By Lazy Frank (****) [gb] Date 2018-11-08 12:02
Time management (decisive games):

Stockfish made 1190 moves in 41:24:32,627 time.

Komodo made 1196 moves in 47:16:54,574 time.

Useless time usage:

Stockfish wasted 12,495 % of time, Komodo wasted 28,659 %.

That gives a proportion of useful time management 1,074232 to Stockfish.
Parent - - By Lazy Frank (****) [gb] Date 2018-11-08 12:17
Visited nodes:

Stockfish = 9 265 886 090 863 nodes (average speed 60700,00  kN/sec.)
Komodo = 9 474 178 168 141 nodes (average speed 54355,67 kN/sec.)

Together with TM again visited nodes proportion is SF-K = 1.2-1.0
Parent - - By Labyrinth (*****) [us] Date 2018-11-08 22:08

>Stockfish = 9 265 886 090 863 nodes (average speed 60700,00  kN/sec.)
>Komodo = 9 474 178 168 141 nodes (average speed 54355,67 kN/sec.)


So SF ran faster, but visited less nodes? Is the correct interpretation here that SF revisited less nodes?
Parent - By Lazy Frank (****) [gb] Date 2018-11-08 22:35
Yes, correct. In all match decisive games Stockfish visited less nodes, use less time, but effectively in the exact the same time period outsearch Komodo by 12:10.

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill