Alayan wrote: ↑Thu Jan 30, 2020 1:31 pm
For such comparisons, the TC is critical. If Chess Tiger 14 is 8x faster on your hardware, then dividing the TC by 8 and giving SF 8knps is the thing to do, but I don't think 40 minutes/game was the original TC at the events where Chess Tiger beat those human GMs.
if it was something like classical with 2h for the first 40 moves then some additional time ; this could change the result by a fair amount.
Where do the 10knps and 20knps results come from ?
In my tests the performance of Stockfish at 5 minutes/game and 10 minutes/game is very close.
I don' think long time tournament control change the result more than 50 elo. And I guess that much time will be in favor of Stockfish.
Alayan wrote: ↑Thu Jan 30, 2020 6:01 pm
More time is better for Stockfish, yes, because at this nps it's starved for nodes and still in the zone of explosive elo growth.
mehmet123 wrote: ↑Wed Jan 29, 2020 9:08 pm
According to the my tests.
The human elo of Stockfish with 1 kn/s is 2612.
The human elo of Stockfish with 10 kn/s is 2950.
The human elo of Stockfish with 20 kn/s is 3035.
The speed of Stockfish on a Pentium-75MHz is 6 kn/s : http://talkchess.com/forum3/viewtopic.p ... 51#p821251
may be a bit more with optimization.
So we could expect SF beat Kasparov and Kramnik on a CPU from 1994 with an engine from year 2020
mehmet123 wrote: ↑Wed Jan 29, 2020 9:08 pm
According to the my tests.
The human elo of Stockfish with 1 kn/s is 2612.
The human elo of Stockfish with 10 kn/s is 2950.
The human elo of Stockfish with 20 kn/s is 3035.
The speed of Stockfish on a Pentium-75MHz is 6 kn/s : http://talkchess.com/forum3/viewtopic.p ... 51#p821251
may be a bit more with optimization.
So we could expect SF beat Kasparov and Kramnik on a CPU from 1994 with an engine from year 2020
All Humans are really bad at chess!. It makes it funny to hear from humans about how chess is played out.
"The worst thing that can happen to a forum is a running wild attacking moderator(HGM) who is not corrected by the community." - Ed Schröder
But my words like silent raindrops fell. And echoed in the wells of silence.
mehmet123 wrote: ↑Wed Jan 29, 2020 9:08 pm
According to the my tests.
The human elo of Stockfish with 1 kn/s is 2612.
The human elo of Stockfish with 10 kn/s is 2950.
The human elo of Stockfish with 20 kn/s is 3035.
The speed of Stockfish on a Pentium-75MHz is 6 kn/s : http://talkchess.com/forum3/viewtopic.p ... 51#p821251
may be a bit more with optimization.
So we could expect SF beat Kasparov and Kramnik on a CPU from 1994 with an engine from year 2020
No surprise.
1994 was a year when Genius3 with P90 beat Kasparov.
It was 25 minutes per game but it is obvious that software made a big improvement from that time.
mehmet123 wrote: ↑Wed Jan 29, 2020 9:08 pm
According to the my tests.
The human elo of Stockfish with 1 kn/s is 2612.
The human elo of Stockfish with 10 kn/s is 2950.
The human elo of Stockfish with 20 kn/s is 3035.
The speed of Stockfish on a Pentium-75MHz is 6 kn/s : http://talkchess.com/forum3/viewtopic.p ... 51#p821251
may be a bit more with optimization.
So we could expect SF beat Kasparov and Kramnik on a CPU from 1994 with an engine from year 2020
No surprise.
1994 was a year when Genius3 with P90 beat Kasparov.
It was 25 minutes per game but it is obvious that software made a big improvement from that time.
I mean "beat Kasparov and Kramnik in long games and in a match".
In 1993 and 1994 Chess Genius made a perf around 2535 : https://www.chessprogramming.org/Auberv ... _Open#1993
That means it had to improve more than 400 Elo to reach the expected level.
In my test Junior 6 % 60 faster than Deep Junior 6 (Dortmund 2000)
The human elo of Stockfish with 10 kn/s is ~2950 according to this test
I'm guessing you don't use increment because the old engines didn't support it perhaps? If that's the reason, then I would recommend you run at something like 40/5 minutes followed by 40/2 minutes repeating. That way you won't get nonsense results, but also won't waste huge amounts of time playing out drawn endgames until the 50 moves rule hits. Sudden death with no increment may be useless if the time management is poor, but with either increment or repeating but faster secondary time controls it plays only a minor role.
Hi Mehmet,
Would you mind redoing some matches with SF 15.1 (speed 5 kn/s) vs Fritz in Bahrain and Shredder 9 ?
I would be interesting to have a recent update.
