Skip to main content

Creating *.efd

warning

This article is under construction

TheParaziT
TheParaziT
Anomaly Modding Book maintainer

About​

*.efd Anomaly Modding Book files are required for simplified probability calculations in Offline

OfflineTechnical Terms
Simulation of life in any given radius from the player will be calculated by a simplified formula.
mode. For example, encounters of mutants and stalkers with each other, anomalies, etc.

Get program​

To create an *.efd file, we need the Evaluation Function Constructor program. To obtain it, you will need to compile it from sources or download the precompiled version from the latest release on GitHub GitHub (Utilities). We only need EFC.exe.

Start​

After you have obtained the program, you need to create the main configuration *.ini file. This file contains global settings, variables, and functions.

Evaluation Function Constructor

List of all parameters.

View Details →
warning

Be sure to name the file "efc.ini"!

efc.ini
;File Names
  LogData = %s\efc.log
  TextData = %s\examples.txt
  BinaryData = %s\binary.dat
  ConfigData = %s\configs.dat
  PatternData = %s\patterns.dat
  CoreData = %s\core.dat
  ParametersData = %s\params.dat
  EFData = %s.efd

;Weight Fitting Parameters
  Epsilon = 0.00001
  Alpha = 1.0
  Beta = 0.01
  MaxIterationCount = 10000

;Probabilistic Weight Fitting Parameters
  RandomFactor = 4
  RandomProbability = 1
  RandomUpdate = 1
  RandomStartSeed = 0

;Configuration Generation Parameters
  MatchThreshold = 1
  MaxCardinality = 30

;Patterns Generation Parameters
  PatternExistanceCoefficient = 1.0

;Function Types

  ;Primary Functions
    Distance = 0
    GraphPointType0 = 1
    EquipmentType = 2
    ItemDeterioration = 3
    EquipmentPreference = 4
    MainWeaponType = 5
    MainWeaponPreference = 6
    ItemValue = 7
    WeaponAmmo = 8
    DetectorType = 9
  
    PersonalHealth = 21
    PersonalMorale = 22
    PersonalCreatureType = 23
    PersonalWeaponType = 24
    PersonalAccuracy = 25
    PersonalIntelligence = 26
    PersonalRelation = 27
    PersonalGreed = 28
    PersonalAggressiveness = 29
    PersonalEyeRange = 30
    PersonalMaxHealth = 31

    EnemyHealth = 41
    EnemyCreatureType = 42
    EnemyWeaponType = 43
    EnemyEquipmentCost = 44
    EnemyRukzakWeight = 45
    EnemyAnomality = 46
    EnemyEyeRange = 47
    EnemyMaxHealth = 48
    EnemyAnomalyType = 49
    EnemyDistanceToGraphPoint = 50

  ;Complex Functions
    WeaponEffectiveness = 61
    CreatureEffectiveness = 62
    IntCreatureEffectiveness = 63
    AccWeaponEffectiveness = 64
    FinCreatureEffectiveness = 65
    VictoryProbability = 66
    EntityCost = 67
    Expediency = 68
    SurgeDeathProbability = 69
    EquipmentValue = 70
    MainWeaponValue = 71
    SmallWeaponValue = 72
    TerrainType = 73
    WeaponAttackTimes = 74
    WeaponSuccessProbability = 75
    EnemyDetectability = 76
    EnemyDetectProbability = 77
    EnemyRetreatProbability = 78
    AnomalyDetectProbability = 79
    AnomalyInteractProbability = 80
    AnomalyRetreatProbability = 81
    BirthPercentage = 82	
    BirthProbability = 83	
    BirthSpeed = 84

Creating project​

The Evaluation Function Constructor operates based on a project design, which means you need "projects" in the form of folders containing files.

info

Create a folder named project1 inside the data folder—this will be the name of the project.

Each project must contain a *.txt file with the project parameters.

warning

The name of the project's *.txt file must be "examples.txt"!

info

Directory structure example:

data/
├── project1/
│   ├── examples.txt      (input)
│   ├── binary.dat        (stage 1)
│   ├── configs.dat       (stage 2)
│   ├── patterns.dat      (stage 3)
│   ├── core.dat          (stage 4)
│   ├── params.dat        (stage 5)
│   ├── ef.dat            (stage 6)
│   └── efc.log           (logging)
├── project2/
│   └── ...
└── projectN/
    └── ...

Project config​

This text-based file contains training data in the form of test examples with associated expected results that the system uses to discover patterns and optimize weights. Let's break it down line by line.

info

example.txt example:

<variable_count>
<function_type>
<value1_min> <value1_max>
<value2_min> <value2_max>
...
<example1_var1> <example1_var2> ... <example1_result>
<example2_var1> <example2_var2> ... <example2_result>
...

Header Line 1: Variable Ranges​

The first line contains space-separated integers, each representing the number of possible values for one variable (range_1 range_2 range_3 ... range_N).

Header Line 2: Variable Type Names​

warning

Must be defined in efc.ini!

The second line contains space-separated strings identifying the semantic type of each variable (type_1 type_2 type_3 ... type_N).

Header Line 3: Function Type Name​

warning

Must be defined in efc.ini!

The third line contains a single string identifying the type of evaluation function (function_type_name).

Test Example Lines: Training Data​

Each subsequent line (lines 4 through end-of-file) represents one test example, containing variable values followed by the expected evaluation result (value_1 value_2 value_3 ... value_N expected_result). This is are the ground truth examples that teach the system how to evaluate different variable combinations

Example​

Now that we know the purpose of each line, we can create our own file or explore existing ones.

Example
5 3 4 2  
Distance PersonalHealth PersonalMorale PersonalWeaponType 
1 2 1 1 12.5
2 1 3 2 8.3  
3 3 2 1 15.7  
4 1 1 2 6.2  
5 2 3 1 11.9  
1 2 1 1 13.1  
2 1 3 2 7.9

For example, the first training line means: When Distance = 1, PersonalHealth = 2, PersonalMorale = 1, PersonalWeaponType = 1, the evaluation should be 12.5

Seven-Stage Pipeline Execution​

The complete pipeline executes seven sequential stages. Each stage can be run independently if prerequisite files exist.

Stage 1: Convert Text to Binary​

efc.exe -p project_name -c

  • Inputs:

    • data/project_name/examples.txt

  • Outputs:

    • data/project_name/binary.dat - Binary test data
    • data/project_name/patterns.dat - Initial single-variable patterns

  • Optional flag:

    • Add -cd to display duplicates found during conversion.

Stage 2: Generate Configurations​

efc.exe -p project_name -gc

  • Inputs:

    • data/project_name/binary.dat

  • Outputs:

    • data/project_name/configs.dat - Configuration bit masks by cardinality

Stage 3: Generate Patterns​

efc.exe -p project_name -gp

  • Inputs:

    • data/project_name/binary.dat
    • data/project_name/configs.dat

  • Outputs:

    • data/project_name/patterns.dat - Filtered, non-redundant patterns

  • Optional flag:

    • Add -gps to show all configurations during generation.

Stage 4: Generate Pattern Basis​

efc.exe -p project_name -gb

  • Inputs:

    • data/project_name/binary.dat
    • data/project_name/patterns.dat

  • Outputs:

    • data/project_name/core.dat - Combined test and pattern data

Stage 5: Optimize Parameters​

efc.exe -p project_name -f

  • Inputs:

    • data/project_name/core.dat

  • Outputs:

    • data/project_name/params.dat - Optimized pattern weights

  • Optional flags:

    • -fp - Force probabilistic weight fitting (random perturbations)
    • -fu - Use previous parameters as starting point if they exist

Stage 6: Build Evaluation Function​

efc.exe -p project_name -b

  • Inputs:

    • data/project_name/core.dat
    • data/project_name/params.dat

  • Outputs:

    • data/project_name.dat - Final evaluation function (note: no subdirectory)

  • Optional flag:

    • Add -bf to save parameters as floats instead of doubles.

Stage 7: Validate Evaluation Function​

efc.exe -p project_name -v

  • Inputs:

    • data/project_name.dat

  • Outputs:

    • Interactive console validation interface

Common Workflows​

Complete Pipeline Execution​

Run all stages in a single command
efc.exe -p project_name -c -gc -gp -gb -f -b -v

This executes all seven stages sequentially using the same project data.

Multi-Project Batch Processing​

efc.exe -pa -c -gc -gp -gb -f -b

Sources​

DeepWiki External link