CMCanavessi wrote: ↑Thu Aug 16, 2018 10:16 pm
So after the "non-optimal" fix they applied in the 2nd TCEC server (the 2xgpu one) leela is 5.5/6, where's all the people that were saying that the level of play that leela was showing at the beginning was real and the ones that knew that something was wrong were just "fanboys"? I understand variance of outcome, but this is too extreme. From being humiliated and not winning a single game for what, 20 games (?) to getting 5.5/6 is a long way...
There are rumors now that the previous NN games that happened before the ugly fix will be replayed (Leela and DeusX). That's gonna create another flamewar in itself
Fixing the hardware setting for lc0 in division 3 is not fair when tcec did not fix the fact that ivanhoe used only one core in division 4 because of rule that they fix nothing unless the engine crash(even if it is not the exe file but some setting).
Replaying previous NN games is not fair unless you go back to division 4 and replay all Ivanhoe games and if Ivanhoe or Wasp promote instead of Deus as a result also give Ivanhoe or Wasp a chance to compete in division 3 instead of Deus.
It's a completely different thing. The ivanhoe case was an ENGINE problem. The binary used had broken SMP. It was its own fault.
In this case, the problem is the SERVER, and it affects all engines running on it (leela and deusx), and it's a TCEC problem, not engine-related.
If you can't see the difference between those 2 situations, then it's pointless to discuss. Still, I agree that replaying all those games would look "weird" and I personally don't like it.
CMCanavessi wrote: ↑Thu Aug 16, 2018 10:28 pm
And what if
Arasan 0-1 Ethereal
and
lc0 1-0 Arasan
That's the interesting case. Also considering Arasan 0.5-0.5 Ethereal.
I already wrote about that. The potential tie breaks between Arasan and LC0 fall in favour of Arasan.
LC0 falls on the first tie breaker (direct encounters) unless LC0-Arasan is win for LC0, in which case first tiebreak is drawn.
LC0 will almost certainly fail the second tie break condition (win count).
Maybe I should rephrase that: LC0 will probably fail the second tie break condition, more often than not.
Which means LC0 needs 0.5 more pts than Arasan to go through. Tough one.
Last edited by chrisw on Thu Aug 16, 2018 10:59 pm, edited 1 time in total.
Note that I assume that ivanhoe did not use more than one core only because of wrong setting.
If the problem is in the exe and ivanhoe that they got did not run without crashes with more than one core then I do not think TCEC need to repeat the games of it.
CMCanavessi wrote: ↑Thu Aug 16, 2018 10:35 pmIt's a completely different thing. The ivanhoe case was an ENGINE problem. The binary used had broken SMP. It was its own fault.
In this case, the problem is the SERVER, and it affects all engines running on it (leela and deusx), and it's a TCEC problem, not engine-related.
If you can't see the difference between those 2 situations, then it's pointless to discuss. Still, I agree that replaying all those games would look "weird" and I personally don't like it.
You are the one who can see because you are so blinded by your own fanboyism.
Whose fault is Ivanhoe?
Binary was provided by Kranium Fire's author (so totally unrelated person to Ivanhoe engine, who btw. had strong disagreement when Ivanhoe was previously allowed to competed in TCEC) who didn't care to look, or maybe even slipped broken compile intentionally since he openly said he doesn't care for Ivanhoe and want for NN engine(s) to go further.
Then their is incompetent money hungry TCEC owner Anton, who doesn't care about anything but the amount of viewers that bring him earnings and that basically made a charade out of TCEC. Who didn't bother even once to start the binary before tournament started or to actually use the binary that was used previously when Ivanhoe was participating in TCEC.
Mess with Ivanhoe is much more a fault of organizers and handicap was for an order of magnitude worse than with Lc0 and no one even bothered to suggest it was not fair from Lc0 team. So much about objectivity.
Uri Blass wrote: ↑Thu Aug 16, 2018 10:42 pm
Note that I assume that ivanhoe did not use more than one core only because of wrong setting.
That part is wrong, it wasn't a settings issue. Settings were correct for THAT version.
If the problem is in the exe and ivanhoe that they got did not run without crashes with more than one core then I do not think TCEC need to repeat the games of it.
There is a certain irony in leela's race for second place not being helped by a draw with its clone (leela-dx).
On the other hand, now the hardware problem is fixed, it seems to be back to being leela with a much better, and probably anticipated, performance in the latter stages of the competition.
Interested to see how far this experiment runs; and whether we are genuinely witnessing the beginning of the end of the pure AB engine dominance in chess. Surprised we have not seen other NN engines emerging (discounting the leela-dx clone), because I doubt that leela (NN structure, back end 'search' etc) is the optimised approach for chess. Great credit to the leela team and all involved in a terrific project.
Congratulations to John, assuming arasan gets promoted. And of course Ethereal, which seems to be a class above the rest of the field - overall anyway.