I remember first step to becoming a good tactical player is examining all captures.
So only near the leaves you can be sure bad captures are really bad captures for there is no depth left to examine them.
If you put bad captures at the end of the move list they get reduced too much by LMR. So it may miss good sacrifices.
Or maybe LMR should not reduce any captures?
Also a bad capture might create a passed pawn so it may be difficult to detect bad captures.
Move ordering - what's best?
Moderators: hgm, Rebel, chrisw
-
- Posts: 1563
- Joined: Thu Jul 16, 2009 10:47 am
- Location: Almere, The Netherlands
Re: Move ordering - what's best?
I exclude all tactical moves (captures, promotions, checking moves, evasions and some pawn pushes) from LMR, maybe this is far from optimal, I don't know. There are so many variables to consider, if you all want to check them out by playing engine-engine matches you have to make a full time job of it. In the past I could sit down working for several days in a row on the engine, but that time is long gone.Henk wrote: ↑Thu Feb 21, 2019 12:28 pm I remember first step to becoming a good tactical player is examining all captures.
So only near the leaves you can be sure bad captures are really bad captures for there is no depth left to examine them.
If you put bad captures at the end of the move list they get reduced too much by LMR. So it may miss good sacrifices.
Or maybe LMR should not reduce any captures?
Also a bad capture might create a passed pawn so it may be difficult to detect bad captures.
-
- Posts: 182
- Joined: Sun Jun 12, 2016 5:44 pm
- Location: London
- Full name: Vincent
Re: Move ordering - what's best?
I use:
- Hash
Good captures (SEE)
Killers (primary, secondary, and primary from ply - 2)
Bad captures (SEE)
Quiets, sorted by history heuristic
-
- Posts: 1871
- Joined: Sat Nov 25, 2017 2:28 pm
- Location: France
Re: Move ordering - what's best?
How small ?I saw a small elo gain after moving bad captures before the quiets, possibly because SEE is often tactically imperfect.