#
Understanding Evaluation Score
#
What is Evaluation Score?
Evaluation Score (Eval) is the metric FPL Review's solvers use to rank different plans. It combines Expected Value (EV) with your solver settings.
Important
Eval ≠ Team EV. Raw team EV is just the expected total of projected player points. Eval incorporates your solver settings to generate the best plan based on what the settings are optimised for.
#
Full vs Linear Evaluation
#
Full Evaluation (Transfer Solver)
Full evaluation holds the capability to include non-linear effects in the assessment:
- Uses full probability calculations recognising complexities around autosubs contributions
- Distinguishes between 11 nailed players vs a team with several players who may not play
- Calculates actual substitution probabilities based on xMins, availability and player usage
- Naturally handles same team keeper modelling
- More realistic but computationally intensive
- Does not guarantee optimality
#
Linear Evaluation (Linear Optimiser)
- Linearised simplification of evaluation for mathematical optimisation
- Fixed sub & vice-captain contribution assumptions rather than dynamic probability-based calculations
- Will not detect interchangeability of same team keepers
- Enables mathematically guaranteed optimal solutions within the linear framework
Note
Optimality in the linear framework does not guarantee global optimality when translated back into the full (non-linear) framework.
Warning
Different solver settings/frameworks/chip plans create incomparable evaluation scores.
#
Literal Evaluation
In cases where we wish to compare the expected outcomes of chip plans, the solution is to change to literal solver settings. This is particularly useful when comparing different chip strategies or validating transfer plans. This is achieved by:
- FT Value, Burn FT Value, Future Info Value, Bank Value (per £1M) => set to 0.00 (remove these effects)
- Time Decay => set to 1.01 (cancelling out xMin decay effects)
- Transfer Depth => maximised (allow for full transfer plans)