Setting of chess program Fritz (Junior, Hiarcs, Shredder)
Oleg Kuryan, Igor Dergilev
- Fritz 8
- Deep Fritz 8
- X3D Fritz
- Hiarcs 9
- Shredder 8
General for all engines:
- (Save) - to memorize the proposed tuning
- (Load) - to load the before stored tuning
- (Default) - to bring an engine over to tuning of "factory"
If the program begins to repeat moves, seeing more advantageous decisions for a competitor. For the exception of it we as though specify the program, that she plays with more weak player, offer the same disrespect.
- At a parameter "+" the program plays more risky, aiming to avoid a draw in game. For example, "+50" shows that the program goes to the variants with a worst estimation in the half of pawn.
- At a parameter "-" the program plays with a supply, searches a draw game.
- Range from +100 to -100 and measured in percents ot the cost of pawn = 100 units.
Very important point. For determination of depth of search, size of search tree.
- A parameter changes from 0 to 6, than higher value of parameter, the deeper there will be a depth of search.
- A parameter strongly influences on tactical force of engine. So at a large value (5-6) tactical continuations (victim, thin motion) can be skipped, if a engine is in the mode "deep analysis" will find more advantageous decisions.
- Setting on "0" (zero) can help in research of positions of tsugtsvanga. I.e. mode "not to threat northing". But not forget to commute it back for normal games. It multiplies practical force of engine considerably.
A parameter is determined by the threshold of address to the base of these end-game, than higher parameter, the more so often (and more early) the program will speak to the database. A parameter changes from 0 to 20, it is not recommended to set a large value without a necessity, as it slows a search substantially.
New parameter which was not in the previous versions of engine. Intended for more aggressive strengthening of depth of search, similarly as well as Selectivity influences on finding of thin motions.
Influence on playing style. Interval:-2 (passive) to +2 (energetic). By default = 0.
A parameter changes from -10 to +10 and determines the preference of estimation of the opened lines.
A parameter changes from -10 to +10 and determines the preference of estimation of rightness of pawn structure.
A parameter changes from -10 to +10 and determines the preference of estimation of protected of king.
A parameter determines, how many weight the program will give material in opposition advantages of positions. A value is given in a percent, "+" a number is multiplied by weight of criteria of positions, "-" a number is a value of material.
Value computer - queen, rook, beshop, knight, pawn. Change of estimation of computer's, opponent's figures.
Mate threat extensions - a parameter in engines for the search of mate, decision of test positions.
Difference of multiprocessor version Deep Fritz 8:
A window of parameters is a complete analogue Fritz 8
- parameter is an amount of processors
- absencejo squares Open files, Pawn levels, King safety, Scale evaluations.
Difference Hiarcs 9 от Fritz 8:
- Selectivity changes from 0 to 7 and determines the width of search tree, instead of depth of search. I.e. as on a picture: most value means the least depth of search.
- Playing style style of game: normal, aggressive, protective.
- Position learning allows an engine to create a file with expansion *.lrn, where he will keep the correction of the playing experience in the process of game or analysis (on the panel of engine, where variants are represented, there is an indicator, which changes the color black, grey, yellow, green and red, at discoloration Hiarcs writes the correction of the playing style in a file).
- Retain hashtable и Smart search influence on the option of cash and functions of search.
- Compent value - size of disrespect. Range: -200 to +200.
- Use Tablebases - end game tables. Threshold of appeal (depth) - like in Fritz8. Range: -3 to +20.
- Threat depth - depth of threat. Range: 0 to +7.
- Huper-modern - hyperactive modern.
- Keep hash tables - to take into account memory of hash tables in an analysis
- Position learning - allows an engine to create a file with expansion *.lrn, where he will keep the correction of the playing experience in the process of game or analysis. (like Hiars9)
- Prefer open positions - similar the parameter of AntiHumanMode - during setting of small flag the program avoids the closed positions, aspiring to the living figured game with the preference of the opened positions.
- Use Tablebase - connection of end game tables
- Combinations - mode of application of combinations: all,most,some,none
- Forward pruning Range: 0-5. Analogue (pruning of uselessness). Intended for more aggressive strengthening of depth of search, similarly as well as Selectivity influences on finding of fine motions.
- Selectivity - For determination of depth of search, size of search tree. Range -3 -:- +99
- Draw Score - similar a parameter "size of disrespect". At "+" to avoid reiterations, at "-" aiming to to nobody. Range -300 -:- +300.
- Style - style of game: balanced, aggressive, active, normal, durable
- Queen, Queen EndGame - value of figure in pawns in middle game, in end game
- Pawn structure: - a preference of rightness of pawn structure in middle game, in end game. +25 to +400
- King safety is a preference of protected of king. +25 to +400
- Center control +25 to +400
- Passed pawn - a preference of communicating pawns in middle game, in end game. +25 to +400.
- Bishop pair - a preference of 2th bishops in middle game, in end game. +25 to +400.
- Time usage - +25 to +400.
- Knowledge or speed -500 to +500.
- Threats - amount of threats.
- Shredder 8(Gambit Shredder 8)
Queen=110 Queen (Endgame)=105 Rook=90 Rook (Endgame)=95 Bishop=90 Bishop (Endgame)=95 Knight=90 Knight (Endgame)=95 Pawn=85 Pawn (Endgame)=95 Draw Score=-10 King Safety=130 Center Control=120 Pawn Structure=95 Passed Pawns=105 Passed Pawns (Endgame)=110 Bishop Pair=105 Bishop Pair (Endgame)=105
Keep Hash Tables=2 Position Learning=2 Prefer Open Positions=1 Use Tablebases=2 Queen=103
Queen (Endgame)=104 Rook=102 Rook (Endgame)=103 Bishop=102 Bishop (Endgame)=103 Knight=104
Knight (Endgame)=104 Pawn=104 Pawn (Endgame)=101 Style=1 Combinations=0 Forward Pruning=2
Selectivity=-1 Draw Score=0 King Safety=115 Center Control=95 Pawn Structure=95
Pawn Structure (Endgame)=100 Passed Pawns=106 Passed Pawns (Endgame)=104 Bishop Pair=105
Bishop Pair (Endgame)=107 Time Usage=100 Knowledge or Speed=0 Threads=1
Keep Hash Tables=2 Position Learning=1 Prefer Open Positions=1 Use Tablebases=2
Queen=102 Queen (Endgame)=104 Rook=103 Rook (Endgame)=102 Bishop=102 Bishop (Endgame)=101
Knight=102 Knight (Endgame)=109 Pawn=103 Pawn (Endgame)=100 Style=0 Combinations=0
Forward Pruning=2 Selectivity=-1 Draw Score=0 King Safety=115 Center Control=95
Pawn Structure=98 Pawn Structure (Endgame)=100 Passed Pawns=104 Passed Pawns (Endgame)=101
Bishop Pair=103 Bishop Pair (Endgame)=104 Time Usage=100 Knowledge or Speed=0 Threads=1
:: Added 06.10.2004