mwyoung wrote: ↑Sat Oct 10, 2020 7:23 am
Lasko's Law----What's not clear? 3 doublings in cores mean nowadays at least 2.5 real effective doublings in TC. Each effective doubling in TC in these blitz conditions means at very least 40 Elo points, therefore at very least 80 Elo points 1 core -> 8 cores. In fact more likely 120 - 140 Elo points. That result posted in OP and discrepancy beyond doubt break the Elo model.
It is clear to me that Stockfish NNUE does not obey Lasko's law as stated above. CCRL most likely does not have flawed testing.. And as suspected. The issues is with Stockfish NNUE. It took me many hours to testing to show this result, and the full results will be shown soon. When the testing is completed. The bottom line is the issue is with Stockfish NNUE, and not with CCRL testing. Full results coming soon. As you know testing can take days to answer this kind of anomaly, or false assumption.
All results were tested under the same conditions with a TC = 2m+1s. With the same book, and settings, with Perfect Book 2019. CPU was a 2950x with all cores locked to 4.1 Ghz.
Stockfish 11 with a classical evaluation obeys Lasko's Law. But assuming Stockfish 12 a hybrid with the new NN evaluation will also obey Stockfish's classical pattern was in error. Stockfish 12 does not obey Lasko's Law.
I tested two versions of Stockfish 12, version 12, and version 12 (051020). To make sure this behavior was not with just the original Stockfish 12.
Stockfish 11 1 vs 8 cores +147.2 Elo
Stockfish 12 1 vs 8 cores +77.7 Elo
Stockfish 051020 1 vs 8 cores +54.3 Elo
Code: Select all
Result:
------------------------------------------------------------------------------------------------
# name games wins draws losses score los% elo+/-
1. Stockfish 11 64 POPCNT dup 8 cores 200 81 118 1 140.0 100.0 147.2
2. Stockfish 11 64 POPCNT dup 1 core 200 1 118 81 60.0 0.0 -147.2
Cross table:
------------------------------------------------------------------------------------------------
# name score games 1 2
1. Stockfish 11 64 POPCNT dup 8 cores 140.0 200 x 111===1===111=1==1=1==1===11=1==11=====1====11=11==1=111==111====1111==11===1==11=========1===1====1111=111=1======1=1=1=0===1==1==1====11=11==11=11=1=11=1==1===1===1=11====11=====1==11=1==11==11==1==
2. Stockfish 11 64 POPCNT dup 1 core 60.0 200 000===0===000=0==0=0==0===00=0==00=====0====00=00==0=000==000====0000==00===0==00=========0===0====0000=000=0======0=0=0=1===0==0==0====00=00==00=00=0=00=0==0===0===0=00====00=====0==00=0==00==00==0== x
Tech:
------------------------------------------------------------------------------------------------
Tech (average nodes, depths, time/m per move, others per game), counted for computing moves only, ignored moves with zero nodes:
# name nodes/m NPS depth/m time/m moves time
1. Stockfish 11 64 POPCNT dup 8 cores 35492K 13595007 31.7 2.6 61.0 159.1
2. Stockfish 11 64 POPCNT dup 1 core 4551K 1662287 27.1 2.7 61.1 167.4
all --- 19530K 7478160 29.4 2.7 61.0 163.3
Tournament finished! Elapsed: 18:23:36
Code: Select all
Result:
------------------------------------------------------------------------------------------
# name games wins draws losses score los% elo+/-
1. Stockfish 051020 dup 8 cores 200 31 169 0 115.5 100.0 54.3
2. Stockfish 051020 dup 1 core 200 0 169 31 84.5 0.0 -54.3
Cross table:
------------------------------------------------------------------------------------------
# name score games 1 2
1. Stockfish 051020 dup 8 cores 115.5 200 x =1==1======1========111=====1=====1=================1====1======1===1==1========1=====================1=1======1==========1==1====1========1========1==1=====1============1===1==========1==11====1==1==
2. Stockfish 051020 dup 1 core 84.5 200 =0==0======0========000=====0=====0=================0====0======0===0==0========0=====================0=0======0==========0==0====0========0========0==0=====0============0===0==========0==00====0==0== x
Tech:
------------------------------------------------------------------------------------------
Tech (average nodes, depths, time/m per move, others per game), counted for computing moves only, ignored moves with zero nodes:
# name nodes/m NPS depth/m time/m moves time
1. Stockfish 051020 dup 8 cores 30556K 10784132 37.2 2.8 49.1 139.1
2. Stockfish 051020 dup 1 core 3659K 1282020 29.7 2.9 49.2 140.4
all --- 16695K 6012180 33.4 2.8 49.1 139.8
Tournament finished! Elapsed: 15:50:53
Code: Select all
Result:
--------------------------------------------------------------------------------------
# name games wins draws losses score los% elo+/-
1. Stockfish 12 dup 8 cores 200 44 156 0 122.0 100.0 77.7
2. Stockfish 12 dup 1 core 200 0 156 44 78.0 0.0 -77.7
Cross table:
--------------------------------------------------------------------------------------
# name score games 1 2
1. Stockfish 12 dup 8 cores 122.0 200 x 1===1===============1======11======1==1===========1=1=====1===11===1===11==1===1========1=1=1==1=1========1===1=========1==1===========1===11======1====1====1==1====1=====1======1====111==1=11===1===1
2. Stockfish 12 dup 1 core 78.0 200 0===0===============0======00======0==0===========0=0=====0===00===0===00==0===0========0=0=0==0=0========0===0=========0==0===========0===00======0====0====0==0====0=====0======0====000==0=00===0===0 x
Tech:
--------------------------------------------------------------------------------------
Tech (average nodes, depths, time/m per move, others per game), counted for computing moves only, ignored moves with zero nodes:
# name nodes/m NPS depth/m time/m moves time
1. Stockfish 12 dup 8 cores 28475K 9905390 35.2 2.9 48.0 137.8
2. Stockfish 12 dup 1 core 3474K 1180298 29.1 2.9 48.1 141.4
all --- 15585K 5486571 32.1 2.9 48.0 139.6
Tournament finished! Elapsed: 15:46:49
"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.