Suisho 5 YaneuraOu NNUE - Parameters

Download File Pdf …

Importance	Parameter	Default Value	Explanation	Beginner
AAA	USI_Hash	16	Amount of memory (in megabytes) the engine uses to “remember” positions already analyzed. The higher it is, the stronger the engine becomes.	4
AAA	USI_Ponder	False	If True, the engine keeps thinking even while you’re moving. This makes it stronger. False makes it more “human,” because it doesn’t analyze during the opponent’s turn.	False
AAA	Threads	4	Number of “parallel processes” the engine uses to think. More threads = faster calculations and greater strength.	1
	Stochastic_Ponder	False	Adds a bit of randomness during the “anticipatory” (ponder) calculation. If True, the engine may choose less predictable moves. It doesn’t always mean it becomes weaker, but it can make its behavior less “linear.”	False
AAA	MultiPV	1	How many main lines (variations) are analyzed simultaneously. 1 = focuses on a single line; higher values = broader analysis.	1
	NetworkDelay	120	A fictitious delay (in milliseconds), usually used in online matches to simulate network latency. It doesn’t significantly affect playing strength.	120
	NetworkDelay2	1120	A second “virtual” network delay. This also does not directly affect the engine’s strength.	1120
AAA 🙁	MinimumThinkingTime	2000	The minimum time (in milliseconds) the engine waits before moving. High values make it more accurate (because it calculates more), while low values make it faster but less precise.	500
AA	SlowMover	100	A “slowdown” percentage for calculations. The higher it is, the more time the engine spends thinking, often finding stronger moves.	200
	MaxMovesToDraw	0	If > 0, after a certain number of moves, the game is declared a draw. With 0, it’s disabled and the game continues normally.	0
AAA	DepthLimit	0	The maximum depth of analysis (0 = unlimited). Lowering it (e.g., to 5) reduces how far ahead the engine looks, making it more “human” and weaker.	5
AAA	NodesLimit	0	The limit on the number of positions analyzed (0 = no limit). A lower value (e.g., 3000) makes the engine weaker.	3000
	EvalDir		Indicates the folder or file for evaluation data. “” means the engine uses built-in data. Normally, you don’t need to change this.
	WriteDebugLog	False	If True, creates log files to analyze the engine’s behavior (useful for developers). It doesn’t affect gameplay.	False
	GenerateAllLegalMoves	False	If True, it generates all legal moves (a debug option). Normally not needed.	False
	EnteringKingRule	CSARule27	Determines how to handle the special rule of the “King entering the opponent’s area” (nyuugyoku).	CSARule27 (standard) o NoEnteringKing (più semplice)
			- NoEnteringKing: No special rule is applied.
			- CSARule24 / CSARule24H: Variations of rule 24.
			- CSARule27 / CSARule27H: More recent variations (27).
			- TryRule: Another particular mode.
	ThreadIdOffset	0	A technical parameter for assigning thread IDs. It doesn’t affect playing strength.	0
BBB	LargePageEnable	True	If True, uses a “special” large-page memory mode to speed up calculations. Not all PCs support it. If disabled, the engine is slightly slower (thus weaker).	False
AAA	USI_OwnBook	True	If True, the engine uses its own opening book (pre-studied initial moves). This makes it stronger in the opening phase.	False
A	NarrowBook	False	If True, the opening book is “narrow,” meaning fewer variations. A narrow book makes the opening less strong and more predictable.	True
A	BookMoves	16	The maximum number of moves the engine will follow from its opening book. The higher it is, the more “predefined” moves are used.	8
B	BookIgnoreRate	0	The percentage of times the engine deliberately ignores the opening book. 0 = it always follows the book.	30
	BookFile	no_book	Name of the external opening book file. “no_book” means there is no dedicated file. If USI_OwnBook is True, there may still be an internal book.	no_book
	BookDir	book	The folder where the opening book file is searched for (if present).	book
BBB	BookEvalDiff	30	The threshold of evaluation difference among book moves. The higher it is, the more the engine may choose moves with different (and sometimes less optimal) evaluations.	50
	BookEvalBlackLimit	0	The minimum evaluation value (for Black) to use a book move. If negative, the engine accepts moves even if they are less favorable.	0
	BookEvalWhiteLimit	-140	The minimum evaluation value (for White) to use a book move. The higher it is, the more selective the engine becomes.	-140
A	BookDepthLimit	16	The maximum depth of book moves (how many “predefined” moves it can follow).	8
	BookOnTheFly	False	If True, the engine updates/creates the book while playing (self-learning). Usually not needed for casual play.	False
	ConsiderBookMoveCount	False	Previously called “ConsiderBookMoveCour.” If True, the engine pays more attention to certain book moves. Not always necessary for normal play.	False
A	BookPvMoves	8	How many “principal variations” (PV) the engine considers from the book. The higher it is, the more opening choices.	4
	IgnoreBookPly	False	If True, the engine stops using the book after a certain number of half-moves (ply).	False
A	SkillLevel	20	Overall skill level (the higher, the stronger). Not all engines handle this the same way, but lowering it usually weakens the engine.	10
	DrawValueBlack	-2	The value the engine assigns to a draw from Black’s perspective. Negative values mean the engine prefers winning over drawing.	-2
	DrawValueWhite	-2	The value the engine assigns to a draw from White’s perspective.	-2
	PvInterval	300	How many milliseconds pass before the engine updates its “main move” (PV) while calculating. It doesn’t change strength, only how often analysis is shown.	300
BBB	ResignValue	99999	The threshold above which the engine resigns. Such a high value means it almost never resigns. If you want it to resign when heavily behind, lower this value.	30000
	ConsiderationMode	False	If True, the engine keeps analyzing even when it’s paused. It doesn’t greatly affect strength but uses more resources.	False
CCC	OutputFailLHPV	True	If True, also shows analysis lines that are not the main line (a debug option). It doesn’t change strength but can be confusing if you don’t know how to interpret them.	False
	FV_SCALE	24	The scale of evaluation scores (doesn’t affect strength, only how the values are “displayed”).	24