LC0 Android RR Tourney 2019..LC0 Disaster

Discussion of computer chess matches and engine tournaments.

Moderators: hgm, Rebel, chrisw

supersharp77
Posts: 1242
Joined: Sat Jul 05, 2014 7:54 am
Location: Southwest USA

LC0 Android RR Tourney 2019..LC0 Disaster

Post by supersharp77 »

Results are in 182 Game LC0 Android tourney completed (Chess For Android) ...After all my extensive LC0 testing not really surprised by the results..."Read Um & Weep" "Live and Learn".. :D 8-) :wink:


Pgn Games file

https://www.4shared.com/archive/uC3Rh00 ... id_RR.html



LCO RR Chess for Android Tournament May 29-30 2019

1 2 3 4 5 6 7 8 9 0 1 2 3 4
1 scorpio-27-ja ** 01 1½ 11 11 11 11 11 11 11 11 11 11 11 24.5/26
2 danasah-488-ja 10 ** ½½ 11 11 11 11 11 11 ½1 11 ½1 11 11 23.0/26
3 wyldchess_v1.5_ 0½ ½½ ** 11 11 01 11 11 11 11 11 11 11 11 22.5/26
4 Typhoon_1.0.r3JA_5 00 00 00 ** 11 10 01 11 11 11 11 11 11 11 18.0/26
5 olithink-531-ja 00 00 00 00 ** 11 11 11 11 01 11 11 01 11 16.0/26
6 Sunsetter_C10.JA_xb 00 00 10 01 00 ** 11 10 11 01 ½½ 11 11 11 15.0/26
7 ZCT-0.3.2500 00 00 00 10 00 00 ** ½1 10 11 1½ 11 11 11 13.0/26
8 sjeng-112-ja 00 00 00 00 00 01 ½0 ** 01 10 11 11 11 11 11.5/26 77.50
9 Galjoen_v0.30.2 00 00 00 00 00 00 01 10 ** 11 ½1 11 11 11 11.5/26 74.75
10 greko-97-ja 00 ½0 00 00 10 10 00 01 00 ** 11 01 01 11 9.5/26
11 Lc0 v0.22.0-dev 00 00 00 00 00 ½½ 0½ 00 ½0 00 ** ½1 11 11 7.5/26
12 Lc0 v0.22.0-dev(full) 00 ½0 00 00 00 00 00 00 00 10 ½0 ** 1½ 11 5.5/26
13 Lc0 v0.22.0-dev(eigen) 00 00 00 00 10 00 00 00 00 10 00 0½ ** 11 4.5/26
14 ifrit-m18-ja 00 00 00 00 00 00 00 00 00 00 00 00 00 ** 0.0/26
User avatar
pedrox
Posts: 1056
Joined: Fri Mar 10, 2006 6:07 am
Location: Basque Country (Spain)

Re: LC0 Android RR Tourney 2019..LC0 Disaster

Post by pedrox »

I do not know details of your games like the time control or the network used. I did the test with a 48x5 network, 8 CPUS, and approximately I see about 500 nps on average.

With a time control of 1 second per move I confronted it to Danasah 7.9 and lost 10-0, showing an ELO of perhaps 2000 ELO points.

But with a time control of 10 seconds per move, the result was matched, 5-5, this can indicate approximately an ELO of about 2500.

Try with more time.
supersharp77
Posts: 1242
Joined: Sat Jul 05, 2014 7:54 am
Location: Southwest USA

Re: LC0 Android RR Tourney 2019..LC0 Disaster

Post by supersharp77 »

pedrox wrote: Fri May 31, 2019 9:47 am I do not know details of your games like the time control or the network used. I did the test with a 48x5 network, 8 CPUS, and approximately I see about 500 nps on average.

With a time control of 1 second per move I confronted it to Danasah 7.9 and lost 10-0, showing an ELO of perhaps 2000 ELO points.

But with a time control of 10 seconds per move, the result was matched, 5-5, this can indicate approximately an ELO of about 2500.

Try with more time.
Time control 2 minutes plus 5 sec per move 3 different LC0's one with the combined weight and engine (LC0-full)
LC0 (eigen) 112 NN and LC0 256 NN :) :wink: