back


29498ken
2019-06-20 02:50:51
CCC 9 game #291, Laser vs Lc0

There are 7 pieces left on the board.

Laser is using 7-man endgame tablebases but lost against Lc0.

Myabe these 6-man and 7-man endgame tablebases have problems and is unreliable. Like what happened to stockfish.
29497Hovhannes
2019-06-18 19:20:42
Stockfish 1902810 is match stronger then Stockfish 19061419 in 5+6 blitz game
29496ken
2019-06-18 18:11:11
@glitzerpony, yes I would like to meet with you.
29495glitzerpony
2019-06-18 17:55:20
@ken, I'll give you right. I've gotten a lot of dog food, and I feel strong. Shall we meet? You seem to be a very hot guy. I would like to meet with you. I like men.
29494ken
2019-06-18 16:36:01
continuation to #29493

It is somewhat very similar to what happened in a bonus game before the TCEC season 15 superfinal where stockfish (or maybe bluefish) failed to execute its final 50th move to force a 50-move rule draw even though it is equipped with 7-men tablebases which resulted to a lose against lc0.
29493ken
2019-06-18 16:26:55
Is this another bug in stockfish engine?

CCC 9 is ongoing and stockfish (played as black pieces) lost against lc0 in this endgame position, FEN: 1QK5/6k1/1qB1R3/8/8/8/6p1/8 b - - 0 1

Stockfish is equipped with 7-men tablebases in CCC 9 and the losing move for the stockfish (157...g1=Q) is not in its principal variation.

My homefish (equipped with 3, 4, and 5-men syzygy tablebases) found the drawing move (157...Qxb8) faster than the blink of the human eye.
29492WCN
2019-06-15 14:16:32
Proof of the Pudding as they say that LCZero is about the same True! rating as Stockfish 8 where it only scored 3 more Points than Stcokfish 8 :)

http://legacy-tcec.chessdom.com/live.php


1 Sufi14 Stockfish 190203 37.5
2 Sufi13 Stockfish 18102108 37.5
3 Sufi12 Stockfish 180614 34.0
4 Sufi11 Stockfish 260318 32.0
5 Sufi15 LCZero v0.21.1-nT40.T8.610 31.0
6 Sufi9 Stockfish 8 28.0





WCN
29491Nobody
2019-06-14 18:34:02
Be careful what you wish for.
29487DICK-HEAD-WATCH! :(
2019-06-13 09:19:03
29486 sf-fan
2019-06-13 07:57:43

29485 glitzerpony
2019-06-12 21:25:19
29486sf-fan
2019-06-13 07:57:43
@29485, :(
29484WCN
2019-06-12 10:25:47
TCEC SCAM! WATCH - LCZero True Rating

https://groups.google.com/forum/#!topic/fishcooking/qGDQYwgEV9g




WCN
29483Lc0 SCAM WATCH!
2019-06-11 19:50:24
Lc0's True rating is about the same as Stockfish 8

https://tcec.chessdom.com


5 Sufi15 LCZero v0.21.1-nT40.T8.610 43 27.5 0 15 [9/6] 3 [0/3] 508.00 3589 -152 [3437]


7 Sufi9 Stockfish 8 43 25.0 0 16 [8/8] 9 [4/5] 444.75 3228 153 [3381]
29481Mrert
2019-06-10 14:36:00
Tell your stockfish to git gud. should work 5/7
29480Salazari
2019-06-10 13:49:28
Hi
in this position 5rk1/pp3p1p/7b/2pR4/8/2P4P/P1PNr1P1/2K4R b - - 0 1
Latest stockfish can not find 1...Rxd2
Why?
29479WIZE-GUY! :)
2019-06-10 06:20:48
29478 Lolz
2019-06-07 20:10:44

So, what now? Fan speed/GPU temperature?


NO! The normal thing TCEC SCAM! :)
29478Lolz
2019-06-07 20:10:44
A theory that Lc0 does worse against weaker oppo seems to be popular these days. Here standings with only top 5 round robin:
Sufi 14: +2
Sufi 12: +1
Sufi 13: 0
Sufi 11: -1
Lc0 : -2
So, what now? Fan speed/GPU temperature?
29477Mrert
2019-06-07 04:08:36
Many pending pull requests with passed patches. Anyone knows what's happening ?
29476Mrert
2019-06-06 21:34:47
Well it's human thinking who developped powerful GPUs so that NN algorithms could be used efficiently.
29475Fahimar
2019-06-06 21:08:31
Sufi15, a story that the stockfishfan :(, will not forget, I am very pleased that Lc0 has shown that it is superior to the human "developers." It is the end of human thinking.I am looking forward to Sufi16, and Another humiliating defeat for Stockfish.
29474Mrert
2019-06-06 20:53:49
Good point.
Funny given Leela's (slightly) inflated evals ^^
Still does not explain its loss against Houdini
29473Nobody
2019-06-06 20:25:30
@Mrert @SF_Forever
I think we have seen before that LC0 has a tendency to score worse against lesser opponents than say SF, kind of the same way of SF before the new contempt.
LC0 needs contempt :)
29472Mrert
2019-06-06 18:09:54
Just my opinion, i'm no expert
29471SF_Forever
2019-06-06 14:38:56
Thanks for the explanation, kind Sir
29470Mrert
2019-06-05 23:04:34
pretty sure LC0 suffers from the shorter TC. combined with its tendency to mess up in endgames, it can lead to terrible blunders.
29469SF_Forever
2019-06-05 21:36:00
So, why is Leela not so dominant in the new TCEC event?

I find it odd given that it was so dominant in the head-to-head against Stockfish

next page >
single page