Engine UCI parameters

• maxplies (PTC, Bare): Brute-force search depth in plies
• depth (Newt): Maximum brute-force search depth in plies. This can be set quite high, because it will never be reached: for Blitz games, time management will prevent it, while for longer time controls, maxnodes sets an upper limit for computation.
• usebook (Newt, SAE): Use opening book?
• maxnodes (Newt): How many nodes to search at most. Mainly useful for non-Blitz games to limit computation effort.
• qplies: Quiescence search depth in plies
• pstab: Piece-square table factor; 0 = no influence of PST
• matetest: This switch selects whether mates or draws should also be evaluated at maximum search depth, not just the next move as in Turing’s algorithm. It allows PTC to seek out or avoid mates and also avoid draws when it is ahead in material. This also works for Newt and SOMA, which also have a tendency to reeach a draw even when they are ahead in material, because their normal evaluation function does not include any draw rules. In Bare, this will also cause the program to avoid draws when it is ahead and seek immediate checkmate if such a move is available.
• pmtlen (Bernstein): Size of the plausible move table
• pmtstart (Bernstein): First ply where the PMT is used, so e.g. PMTSTART = 2 means that the PMT will not be used during the first two plies.
• EasyLearn (PTC): Learn factor AKA easy play factor; e.g. if EasyLearn = 3, then PTC picks a move randomly from the three best moves. An exponential distribution with λ=EasyLambda/10 is used. With the default value of λ=2, the best move is selected 86% of the time, while for λ=1 and λ=0.5 the probabilities are 63% and 39% respectively. A low value of EasyLearn, e.g. EasyLearn=2, can be used to add occasional randomness without weakening PTC too much.
• PlayerAdvantage (PTC): If not 0, keep the evaluation at least this many decipawns in favor of the player. (Takes precedence over the other easy play parameters below.) Negative values are also allowed. So if you are playing as White, e.g. -30 will keep the eval above -3 and 10 will keep it over +1.
• MoveError (PTC): Choose randomly from moves that are up to MoveError (in decipawns) worse than the best move
• BlunderPercent (PTC): Chance of a blunder in percent
• BlunderError (PTC): If this move is a blunder, choose randomly from moves that are up to BlunderError (in decipawns) worse than the best move
• pawnrule (Shannon): Enable/disable pawn evalution criteria in Shannon. Disabling this makes the engine run faster without massively changing moves.
• nummov (SAE): Number of moves to consider in a position.
• mtime (SAE): Move time in seconds. Starting and stopping Stockfish needs extra time, so in practice it may be about one to three seconds more than mtime per move.
• ev (SAE): Similar to PlayerAdvantage, but in centipawns: target evaluation which the engine tries to maintain. 100 is the default, 300 would mean one bishop advantage for the player. Negative values also work.
• alim (SAE): Maximum deviation from zero allowed for adaptive play in centipawns. Beyond those limits, optimum moves are played.
• lambda (SAE): Similar to EasyLambda, uses an exponential distribution to modify the target evaluation. A lower lambda such as 1 causes more blunders, a high value such as 100 no blunders at all.
• enginepath (SAE): Path to chess engine binary (must support UCI and MultiPV)
• bookpath (SAE): Path to opening book in Polyglot format
• trueval (SAE): If true, use the true evaluation of the board for info score. If false, use evaluation of selected move, which should be close to ev.
• waitbook (SAE): If true, play book moves at about the same speed as regular moves.

Please note that for PTC and Bare, you should use odd numbers for maxplies and qplies (maxplies = 3 equals four plies; maxplies = 1 equals two plies). This is because PTC and Bare do not count the very first ply (i.e. the first own move considered).

Newt on the other hand counts plies normally from the root position, so maxplies and qplies should be even numbers.