Please could you tell me whether this statement is true or false:
"For Lc0 to analyse any position well it needs a move history of at least one ply. So if you have a position in mind you should set up the board one ply earlier."
True or false?
Lc0 Question
Moderator: Ras
-
zullil
- Posts: 6442
- Joined: Tue Jan 09, 2007 12:31 am
- Location: PA USA
- Full name: Louis Zulli
Re: Lc0 Question
False in that generality. Lc0 analyzes startpos quite well, for example.
-
MOBMAT
- Posts: 412
- Joined: Sat Feb 04, 2017 11:57 pm
- Location: USA
Re: Lc0 Question
It would still need a correct FEN in order to know the castling permissions, the EP square, and the half-move count.
i7-6700K @ 4.00Ghz 32Gb, Win 10 Home, EGTBs on PCI SSD
Benchmark: Stockfish15.1 NNUE x64 bmi2 (nps): 1277K
Benchmark: Stockfish15.1 NNUE x64 bmi2 (nps): 1277K
-
Ovyron
- Posts: 4562
- Joined: Tue Jul 03, 2007 4:30 am
Re: Lc0 Question
A more interesting question is if Lc0 will analyze a position with a history *better*.
-
Werewolf
- Posts: 2080
- Joined: Thu Sep 18, 2008 10:24 pm
- Full name: Carl Bicknell
-
smatovic
- Posts: 3532
- Joined: Wed Mar 10, 2010 10:18 pm
- Location: Hamburg, Germany
- Full name: Srdja Matovic
Re: Lc0 Question
This was at least true for the past, you had to feed LC0 with the move history, not sure when and how they changed it.
--
Srdja
--
Srdja
-
jp
- Posts: 1488
- Joined: Mon Apr 23, 2018 7:54 am
Re: Lc0 Question
The claim that a few plies of move history were needed was certainly made on the Leela blog, which is supposed to be semi-official, in the past.