Skip to main content

Setting analysis parameters

Since version 2.04, you can change the parameters used by the program to determine whether a move is good or bad and also the elo calculated for an analyzed game.

The parameters are as follows:

limit_score

Default value: 2000
Minimum value: 1000
Maximum value: 4000
Data type: integer
Determines the score in centipawns from which the position is considered to be totally won, that the probability of winning is 100%.

curve_degree

Default value: 50
Minimum value: 1
Maximum value: 100
Data type: decimal percentage

Expresses the correlation between the score in centipawns and the probability of winning/losing. It associates -limit_score with 0 and +limit_score with 100.
Curve formula created based in  python-chess/Engine/Score/wdl  

difmate_inaccuracy

Default value: 3
Minimum value: 1
Maximum value: 99
Data type: integer

If the difference between the best move and the move made means going from mate to mate-3 or higher, it is considered an inaccuracy. To cancel it, just give it a value of 99 for example.


difmate_mistake

Default value: 12
Minimum value: 1
Maximum value: 99
Data type: integer

If the difference between the best move and the move made is from mate to mate-12 or higher, it is considered an error. To cancel it, just give it a value of 99 for example.

difmate_blunder

Default value: 20
Minimum value: 1
Maximum value: 99
Data type: integer

If the difference between the best move and the move made means going from mate to mate-20 or higher, it is considered a blunder. To cancel it, just give it a value of 99 for example.


mate_human

Default value: 15
Minimum value: 10
Maximum value: 99
Data type: integer

The maximum mate that on average can be calculated by a human.


blunder

Default value: 15.5
Minimum value: 1.0
Maximum value: 99.0
Data type: decimal

If the difference between the best movement and the realized movement (scaled to the correlation curve) is greater than this value, it is considered a blunder.


mistake

Default value: 7.5
Minimum value: 1.0
Maximum value: 99.0
Data type: decimal

If the difference between the best movement and the realized movement (scaled to the correlation curve)  is greater than this value, it is considered a mistake.


inaccuracy

Default value: 3.3
Minimum value: 1.0
Maximum value: 99.0
Data type: decimal

If the difference between the best movement and the realized movement (scaled to the correlation curve) is greater than this value, it is considered to be an inaccuracy.


very_good_depth

Default value: 8
Minimum value: 1
Maximum value: 128
Data type: integer

For the determination that a movement is very good, the depth to which the motor had to reach, to realize that this is the best movement, is considered.

good_depth

Default value: 5
Minimum value: 1
Maximum value: 128
Data type: integer

For the determination that a movement is good, the depth to which the motor had to reach, to realize that this is the best movement, is considered.


speculative_depth

Default value: 3
Minimum value: 1
Maximum value: 128
Data type: integer

For the determination that a movement is speculative, the depth to which the motor had to reach, to realize that this is the best movement, is considered. 


max_elo

Default value: 3300
Minimum value: 2000
Maximum value: 4000
Data type: integer

Maximum elo that can be assigned to a game.


min_elo

Default value: 200
Minimum value: 0
Maximum value: 2000
Data type: integer

Minimum elo assignable to a game.

elo_blunder_factor

Default value: 12.0
Minimum value: 1
Maximum value: 99
Data type: decimal

In the calculation of the elo of a game, the blunder has a higher weight than a normal move, and this value is used as a multiplier effect.

elo_mistake_factor

Default value: 6.0
Minimum value: 1
Maximum value: 99
Data type: decimal

In the calculation of the elo of a game, mistakes have a higher weight than a normal move, and this value is used as a multiplier effect.

elo_inaccuracy_factor

Default value: 2.0
Minimum value: 1
Maximum value: 99
Data type: decimal

In the calculation of the elo of a game, inaccuracy has a higher weight than a normal move, and this value is used as a multiplier effect.



To change these parameters, when visualizing any game, Config/Analysis configuration parameters



A window appears in which the parameters can be modified directly, and the changes take effect immediately:




Comments

  1. Thanks for the great update sir. Please add an evaluation bar for chess analysis in the next update. Thanks

    ReplyDelete
  2. Very nice updates. Thanks for your efforts for making my chess good.

    ReplyDelete

Post a Comment

Popular posts from this blog

R 1.06 Maia

 This update contains a major improvement with the inclusion of Maia. Maia is a neural network chess engine, created with the purpose of making her movements as human as possible. The Maia creation team has prepared 9 versions, from 1100 to 1900 of elo, and thus they have been included in Lucas Chess. You have all the information about Maia at: https://maiachess.com/

LINUX - First binary version

The first official binary version of Lucas Chess on Linux has just been released. It has only been made possible by the great work done by Laudecir Daniel , choosing and compiling the 67 engines, as well as laying the groundwork so that the binaries can be more widely used and others such as shortcut creation, installation scripts, ...... In addition,  Graham O'Neill  has provided Linux-specific drivers for several digital boards. This version has two limitations : 64-bit Linux OS only. Minimum GNU libc version: 2.26 (ldd --version shows it) Installation has been made as simple as possible: The installer is a .sh file, generated with the  makeself  tool, and can be downloaded from: SourceForge , GitHub The file must be run from a terminal with user permissions only: sh ./LucasChessR123_LINUX.sh Launches a GUI with very basic options: Install: Installs in LucasChessR folder inside the home user folder. Generates a shortcut in Applications > Games Launch the ap...

Future version 10 step 05: Scanner of positions

Some days ago Juan Carlos Antón, head of UNED chess school , does give me permission to include tactical positions from the PDFs on the downloads of UNED web, which I consider very educational. The problem is there is no possibility of obtaining  original fens. The solution to ease my work, has been to create something like an OCR of image positions, which at this specific case is working very well. Only three concepts are handled : mark, deduce, and learn. First, mark the board, the next step is to deduce, if there are errors the user corrects them manually and asked the program to learn from them. At the beginning there are quite a few errors, but less and less. To mark the Board a transparent window is launched that occupies the desk and lets see the PDF and select a position. It is selected with the mouse by dragging a corner to the other, and it is fitted with the arrow keys. At the end the Enter key is pressed to accept the selected and continue in the program wi...