Our research method.
Course fit is not a footnote. It sits alongside player ability, recent form, and measured conditions in our independent PGA GOD V3.5 research framework.
A disciplined lens on each field.
The weights are defined as standardised contributions to projected 72-hole player-strength variance before stochastic noise, not direct shares of outright win probability. Historical strokes-gained ability, course-specific specialisation, and measured form residuals are treated separately to limit double-counting.
Research, not bookmaker-following.
Model ratings are built independently of market prices. We later compare the estimated probability with observed odds. An apparent overlay is a hypothesis for testing, never a promise of return.
Version control & the Monday audit.
Our original 17 November RSM reference run remains frozen. Different experimental methods, including Round 1 three-balls, retain their own trials and settlement rules rather than rewriting past predictions. We record the pre-event time, market terms, estimated prices and outcome when available.
Model limits.
50,000 simulations improve Monte Carlo precision, not predictive validity by themselves. The Round 1 scoring-noise assumptions are not yet externally calibrated; wind-specific player skill should not be inferred unless measured. The 2025 reported record is not a substitute for a full independent audit.
See this week's research ↗