TCEC Division 3 results simulator

[chrisw]

Results Simulator, Division 3, TCEC 13

34 rounds so far

Column 'First' is chance of winning, Column 'First two' is chance of being in first two

Code: Select all

Engine,   Tournament Elo, Initial Elo, First, First two
Ethereal       3332         3295     0.959  0.993
Pedone         3132         3092     0.015  0.298
lc0            3177         3200     0.011  0.243
DeusX          3153         3150     0.009  0.228
Arasan         3128         3124     0.004  0.137
Hannibal       3127         3158     0.001  0.051
Nemorino       3099         3099     0.001  0.050
Bobcat         2999         3029     0.000  0.000

Code: Select all

 1 Ethereal 10.81    3176 7.0  8 26.25 0 +88 87.5 ·· =  1  1= 1  1  1  1 
 2 Pedone 1.8        3104 5.5  9 19.75 0 +48 61.1 =  ·· =  1  =  0= 1  =1
 3 lc0 16.10520      3219 4.5  9 14.75 0 -32 50.0 0  =  ·· =  =  =  == 1=
 4 DeusX 1.0         3200 4.0  8 15.25 0 -19 50.0 0= 0  =  ·· 1  =  =  1 
 5 Arasan TCEC13     3142 4.0  8 14.50 0  +2 50.0 0  =  =  0  ·· 1= =  1 
 6 Nemorino 5.01     3104 4.0  9 17.50 2  +3 44.4 0  1= =  =  0= ·· 1  0 
 7 Hannibal 20180806 3193 3.0  8 10.50 0 -46 37.5 0  0  == =  =  0  ·· 1 
 8 Bobcat 8          3072 2.0  9  9.00 0 -44 22.2 0  =0 0= 0  0  1  0  ··

[chrisw]

After game 42, LC0 losing to Nemorino, the Sim rates the chances of going through to the next division as:

Ethereal 100%
Arasan 30%
Pedone 24%
DeusX 22%
Nemorino 15%
LC0 6%

not looking good for Leela, NN programs looking somewhat over-hyped recently

[CMCanavessi]

They messed up the leela time management, and it hurts a lot.

[kranium]

They messed up the leela time management, and it hurts a lot.

c'mon Carlos

DeusX at Andscacs level?
Leela with recent +150 elo improvement?

oh yeah, nevermind...there have been config issues, not enough threads, GPU temps too high ...and now time management
added to the parade of excuses/explanations

It's pretty clear with all the hype these last weeks, we've all been duped

[Ratosh]

Is the same TC issue affecting DeusX?

[CMCanavessi]

They messed up the leela time management, and it hurts a lot.

c'mon Carlos

DeusX at Andscacs level?
Leela with recent +150 elo improvement?

oh yeah, nevermind...there have been config issues, not enough threads, GPU temps too high ...and now time management
added to the parade of excuses/explanations

It's pretty clear with all the hype these last weeks, we've all been duped

I really don't care if it wins TCEC or if it finishes last in div4, but the time management issue is real and the leela team really screwed up. Check my stream, I'm currently doing lc0 (on a gtx 1080 non ti) vs SF8 (on 8 cores) and currently they are 6-6. Why do you think leela went from dominating in div4 to playing awful in div3? It's no magic, but also, as you say, excuses are useless. Leela team will learn the lesson for next season.

[kranium]

They messed up the leela time management, and it hurts a lot.

c'mon Carlos

DeusX at Andscacs level?
Leela with recent +150 elo improvement?

oh yeah, nevermind...there have been config issues, not enough threads, GPU temps too high ...and now time management
added to the parade of excuses/explanations

It's pretty clear with all the hype these last weeks, we've all been duped

I really don't care if it wins TCEC or if it finishes last in div4, but the time management issue is real and the leela team really screwed up. Check my stream, I'm currently doing lc0 (on a gtx 1080 non ti) vs SF8 (on 8 cores) and currently they are 6-6. Why do you think leela went from dominating in div4 to playing awful in div3? It's no magic, but also, as you say, excuses are useless. Leela team will learn the lesson for next season.

Div 3 engines are much stronger...
As far as your test goes, are you saying that LC0 is as strong as SF8 (or even close)?

There's just been an awful lot of undue hype IMO...and that's disappointing.
Consequently, I saw it announced on TCEC chat that Div 4 had a record numbers of viewers.

[Laskos]

They messed up the leela time management, and it hurts a lot.

c'mon Carlos

DeusX at Andscacs level?
Leela with recent +150 elo improvement?

oh yeah, nevermind...there have been config issues, not enough threads, GPU temps too high ...and now time management
added to the parade of excuses/explanations

It's pretty clear with all the hype these last weeks, we've all been duped

I really don't care if it wins TCEC or if it finishes last in div4, but the time management issue is real and the leela team really screwed up. Check my stream, I'm currently doing lc0 (on a gtx 1080 non ti) vs SF8 (on 8 cores) and currently they are 6-6. Why do you think leela went from dominating in div4 to playing awful in div3? It's no magic, but also, as you say, excuses are useless. Leela team will learn the lesson for next season.

Leela did not dominate in div 4, check the score. Div 3 partial results are just a bit worse than expected. Div 4 had 2-3 engines not even running properly, and were mostly weaker even in ideal conditions. TCEC conditions are to blame, as I stated in another thread. GPUs are underpowered price-wise.

[CMCanavessi]

Is the same TC issue affecting DeusX?

No, deusx plays with default settings

[CMCanavessi]

They messed up the leela time management, and it hurts a lot.

c'mon Carlos

DeusX at Andscacs level?
Leela with recent +150 elo improvement?

oh yeah, nevermind...there have been config issues, not enough threads, GPU temps too high ...and now time management
added to the parade of excuses/explanations

It's pretty clear with all the hype these last weeks, we've all been duped

I really don't care if it wins TCEC or if it finishes last in div4, but the time management issue is real and the leela team really screwed up. Check my stream, I'm currently doing lc0 (on a gtx 1080 non ti) vs SF8 (on 8 cores) and currently they are 6-6. Why do you think leela went from dominating in div4 to playing awful in div3? It's no magic, but also, as you say, excuses are useless. Leela team will learn the lesson for next season.

Div 3 engines are much stronger...
As far as your test goes, are you saying that LC0 is as strong as SF8 (or even close)?

There's just been an awful lot of undue hype IMO...and that's disappointing.
Consequently, I saw it announced on TCEC chat that Div 4 had a record numbers of viewers.

No, I'm saying that with my current config (gtx 1080 for leela, 8 cores for SF8, no TBs), default settings for lc0 and the 10666 net, they are even after 12 games with 2 wins each. Making a bold statement as "lc0 is as strong as X" without giving context is impossible.

Anyways, DeusX with default settings and the same old network as in div4 is ahead of lc0 with newer (and stronger) net but stupid TM settings. How can you explain that?