Fig 1.
The three identified transitions correlate with periods of intensive warfare.
Fig 2.
Size and casualty ratio by battle.
The total number of soldiers involved in each battle is defined in the Y axis while the size of each point shows the casualty ratio of the battle.
Fig 3.
Casualties of two forces as computed by the four competing models with P = 1.5, Bt = 0 = 15000 and Rt = 0 = 10000.
Fig 4.
Flowchart for the ABC framework.
Example for a experiment using 1000 runs and tolerance τ = 0.01. Left side illustrates the rejection algorithm while the green panel details the simulation of the Lanchester model.
Fig 5.
Model selection for different tolerance levels.
Proportion of the models used in the best runs for the four historical periods and three τ values (corresponding to the selection of left: 500, centre: 5000 and right: 50000 best runs).
Fig 6.
Distance from fatigue model to evidence with τ = 0.0005.
Absolute distances (Y axis) of the best 500 runs ordered by rank (X axis, being 1 the best one), model (colour) and historical period (left to right).
Fig 7.
Results for the fatigue model and τ = 0.0005.
Fig 8.
Results for the fatigue model and τ = 0.0005.
Fig 9.
Relation between μ and σ values with τ = 0.0005.
Large σ values are only selected when μ value is also large.