Best depth for blunder checking repertoire games.

[Jonathan003]

I remember watching some YouTube video a long time ago. Where Steve Lopez from Chessbase explained what engine settings are best to use for blunder checking you white repertoire, and what engine settings are best to use for blunder checking your black repertoire.
I remember he was saying that you better use a odd number for depth (like 29 ply), for blunder checking your white repertoire. And to better use a even number for depth (like 30 ply), for blunder checking your black repertoire.
For the computer chess experts here. Does this advice makes sense or not?

[emadsen]

I remember he was saying that you better use a odd number for depth (like 29 ply), for blunder checking your white repertoire. And to better use a even number for depth (like 30 ply), for blunder checking your black repertoire.
For the computer chess experts here. Does this advice makes sense or not?

Time-per-move or time-per-game makes much more sense to me. Depth is an artificial concept. Reduction logic can vary a lot between engines. One engine's search tree is deep and very narrow. Another engine's search tree is shallow and fuller. There's no apples-to-apples comparison of depth across engines.

Modern engine strength is validated with timed games. Depth is irrelevant.

[Jonathan003]

Thanks for the recommendation. If I blunder check a bin book with poly 1.7a, I only have the option depth, and an option max in milliseconds. I can blundercheck by repertoires with Scid vs Pc, and use min seconds per move there.
Wat would be a reasonable setting for min seconds per move? If I want to blunder check my repertoire with Stockfish 12.

This is my hardware:

Windows-versie
Windows 10 Home
© 2018 Microsoft Corporation. Alle rechten voorbehouden.
Systeem
Processor: Intel(R) Core(TM) i7-4770T CPU © 2.50GHz 2.50 G Hz
Installed memory 16,0 GB (15,9 GB available)
(RAM):
Type systeem:
64-bits besturingssysteem, x64-processor

[emadsen]

That's very difficult to answer. It depends on your goals. Are you preparing an opening book for engine-engine competition? That's an arms race. You devote a lot of computing resources to analyzing a repertoire in order to cull inferior moves from your opening book. Your opponent devotes more resources. Now you must devote even more resources.

If you're preparing an opening book to practice your repertoire for over the board competition, then you don't need to devote much computer analysis time at all (per move or per game) because world-class chess engines (Stockfish, Komodo, Ethereal, ~3400 - 3550) are 700 ELO points stronger (likely more) than the best human chess player (Magnus Carlsen, ~2850). Normally one second per move would produce fine analysis. However, chess engines are weaker in the opening than middlegame and endgame, so perhaps more.

Another way to answer the question is to work backwards. How much CPU time and electricity are you willing to devote to strengthening your repertoire / opening book? How many positions does the book contain? Time per move = devotedTime / (positions * degreeOfParallelism). "Degree of parallelism" meaning number of CPU cores analyzing the book concurrently.

Maybe an opening book expert can chime in. I just wanted to make you aware of the fallacy of depth reported by chess engines, and suggest analysis time as a better metric. I can't offer specific recommendations for creating opening books.