From 31. desember to second january it was a huge improvement to Stockfish. 5 ELO. Impressive.
Tested with Fritz 15 UCI with 4 cpu, no opening book and no sysygy.
30 sekund/game. 256 MByte RAM. i7 cpu 3770 with 16 GByte RAM with windows 10.
1 Stockfish 020117 64 POPCNT +5 +2176/=7905/-1991 50.77% 6128.5/12072
2 Stockfish 311216 64 POPCNT -5 +1991/=7905/-2176 49.23% 5943.5/12072
Stockfish improvement. 5 ELO in 2 days
Moderator: Ras
-
consen
- Posts: 80
- Joined: Tue Mar 11, 2014 6:09 pm
- Location: Norge
-
JJJ
- Posts: 1346
- Joined: Sat Apr 19, 2014 1:47 pm
Re: Stockfish improvement. 5 ELO in 2 days
So far, the official no regression test is showing +5 elo with Stockfish dev against Stockfish 8.
-
ernest
- Posts: 2059
- Joined: Wed Mar 08, 2006 8:30 pm
Re: Stockfish improvement. 5 ELO in 2 days
Well, even with those 12000 games, the error-bar (2-sigma, or 95%) still is +/- 3.7 Elo.consen wrote:From 31. desember to second january it was a huge improvement to Stockfish. 5 ELO. Impressive.
So, improvement yes, but perhaps not 5 Elo...
-
Eelco de Groot
- Posts: 4724
- Joined: Sun Mar 12, 2006 2:40 am
- Full name: Eelco de Groot
Re: Stockfish improvement. 5 ELO in 2 days
It would be nice if Gunnar's test with 4 CPU shows an improvement like that, because there was if I'm not mistaken, only one functional change in that timeperiod? Maybe I did not look right. But it was for use with SMP so it would not show up in the regression test at least not in the single CPU regression test. And also not in the extensive Stockfish test from Adam Hair (also one thread) or the rating list from Stefan.
Last edited by Eelco de Groot on Wed Jan 18, 2017 3:03 am, edited 1 time in total.
Debugging is twice as hard as writing the code in the first
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan
-
Lyudmil Tsvetkov
- Posts: 6052
- Joined: Tue Jun 12, 2012 12:41 pm
Re: Stockfish improvement. 5 ELO in 2 days
yes, it is either SF, or the clones;ernest wrote:Well, even with those 12000 games, the error-bar (2-sigma, or 95%) still is +/- 3.7 Elo.consen wrote:From 31. desember to second january it was a huge improvement to Stockfish. 5 ELO. Impressive.
So, improvement yes, but perhaps not 5 Elo...
if SF does not prosper, the clones will prosper.
If harfd-working people can not work, because all kinds of obstacles are put on their way and their products are stolen, the code-stealers and clone-makers will thrive.
I would not be surprised if in another 3 months' time SF's progress stands at precisely 5 elo again. At the same time, we will have not just 5 or 6 blatant and fully useless clones as appeared very sadly during the last couple of weeks, but 10 or even 20.
anyone can make their own choice whom to support.
-
ernest
- Posts: 2059
- Joined: Wed Mar 08, 2006 8:30 pm
Re: Stockfish improvement. 5 ELO in 2 days
Not that it matters a lot, but what does that "answer" have to do with this thread ???Lyudmil Tsvetkov wrote:yes, it is either SF, or the clones;
if SF does not prosper, the clones will prosper.