Neural Futsal

Deep MLP football agents trained with Backpropagation (Adam) & Policy Fine-Tuning

Players per team Hidden layers Neurons per layer Population size Games per individual Match duration (min) Goal width (m) Speed multiplier Learning rate (Adam) Exploration noise (σ) Auto-save every (h)

Auto-save to file is disabled.

Fitness Weights

Goal scored Goal conceded Own goal

Changes apply immediately and are saved with the generation.

Neural Network Training Stage Exciting Moments (Highlights)

⚡ Smart Broadcast (1× on goal attacks & shots, 2.5× in midfield) 🔥 1× Goal Attack

Show time indicator Show pitch stripes

Match Reward Breakdown

Component	A	B

Status

Training Epoch: 500

Match: —

Games/s (10s): 0.00

Team Rating: 112.50

Mode: Smart Broadcast

Rendering: On

Match Reward Breakdown

Status

Lineage Across Generations

Neural Network Evolution