PHCEL and LC0

[Rubinus]

LCZero 0.10-id323 - Spike 1.4_SpikePircBook_12m+5s +3=6-7
games
Big loss of three losses at the end.
SpikePircBook

[Rubinus]

LCZero 0.10-id330 - Spike 1.4_SpikeD00-D09Book_12m+5s +4=6-6
games
Another small improvement.
SpikeD00-D09Book

[Rubinus]

LCZero 0.10-id359 - Spike 1.4_SpikeC20-C29Book_12m+5s +5=6-5
games
Another small improvement.
SpikeC20-C29Book

All my LC0 games

[Rubinus]

LCZero 0.10-id359 - Spike 1.4_SpikeC40-C49Book_12m+5s +4=7-5
games
Step back. But - the start used by Spike's white here is probably better than the one in the previous match. In addition, the match played three times and it is interesting to make an unplanned restart with the course of the match. The foolish Microshit now overwhelms it so it restarts quietly without the approval of the PC owner in the middle of the job. The first three games won LCZero, and the fourth restart seemed to me that she was well. After the match was made a day later, it looked like two other engines were playing at the same time! And because of the time, it was a matter of two bets earlier, which then played the next day.
SpikeC40-C49Book

All my LC0 games

[Rubinus]

LCZero 0.10-id359 - Spike 1.4_SpikeScotchBook_12m+5s +2=5-9
games
A good shot, probably the optimism of the previous quite balanced matches was not quite right. Tactical holes are still there.SpikeScotchBook
Next C5.

All my LC0 games

[Rubinus]

LCZero 0.10-id437 - Spike 1.4_SpikeC50-C59Book_12m+5s +0=5-11
games

LCZero 0.10-id443 - Spike 1.4_SpikeSizilian_g6Book_12m+5s +0=6-10
games
I think reboot learning was a big mistake, lost 2 months. New networks have a higher rating on sites, but the real power against id 359 is at least -100ELO! The program would need to be upgraded to the programming page. Monte Carlo search could perhaps only be used in a quieter position, so LC0 sees tactics too late - just as well as Komodo and Scorpio if they have MCS on.


SpikeC50-C59Book
SpikeSizilian_g6Book
Next Sizilian_e5.

All my LC0 games

[Rubinus]

LCZero 0.10-id446 - Spike 1.4_SpikeSizilian_e5Book_12m+5s +4=4-8
games
After a while, a little more hope. But the last pre-match networks played even better matches.

SpikeSizilian_e5Book
Next C60-C67.

All my LC0 games

[Rubinus]

LCZero 0.10-id455 - Spike 1.4_SpikeC60-C67Book_12m+5s +4=6-6
games
It starts to look like the wounds after the self-learning restart are healed. Now, the question is whether the hope of the authors is that the program will not become obtrusive as the last one. I think the power gets a little farther, but ...

SpikeC60-C67Book
Next C68-C79.

All my LC0 games

[Rubinus]

LCZero 0.10-id462 - Spike 1.4_SpikeC68-C79Book_12m+5s +2=6-8
games
No, it's not before the restart. My estimate = 3 weeks.
SpikeC68-C79Book
Next C80-C83.

All my LC0 games

[Rubinus]

LCZero gauntlets 2+2
12x (Fritz 16, Nimzo 8, Senpai 2, Colosss 2008b, Hiarcs 14, ProDeo 2.6, Spike 1.4, Fruit 2.3.1, Junior Yokohama, Zappa II Mexico)
id 359 - 51.0/120, ph-elo 2940
id 390 - 48.5/120, ph-elo 2926
games 359, games 390
Next id 483

All my LC0 games