darts.mattalldian.com · github.com/mattalldianhr/darts-cricket-research
Extends Andrew Frongello's 2018 analysis with realistic skill profiles, bull-accuracy modeling, and four new strategies.
The interesting finding: there's no single best strategy. The optimum shifts with your skill, your opponent's skill, and how hard the bullseye is on the board you're playing — and the strategy that wins at MPR 4.0 is dead last when the bull gets harder.
A reinforcement-learning agent, trained from scratch against thirteen hard bots, independently rediscovered Frongello's "never chase, score before cover" principles. It also surfaced one move no human strategy makes: open with the 20 if you're going first, the 18 if you're going second.