The Draft That Got Away: How a Bayesian Model Predicted the Spurs’ Pick—and Why Nobody Believed It

The Draft That Got Away

I watched the NBA draft lottery like a Monte Carlo simulation running in real time—each ping of the ping-pong ball was a sampled posterior from a prior distribution I’d built in R. Not magic. Not destiny. Just likelihoods.

The Spurs got the #2 pick. Everyone cheered. But I didn’t. Because my model had assigned a 14% chance to them—based on win probabilities, injury history, and draft position entropy—not hype or broadcast narratives.

The Signal They Ignored

They called it ‘luck.’ I called it P(D|H). The data didn’t care about your emotional bias—it cared about covariance matrices, expected value vectors, and conditional transitions over 50,000 simulations.

The Warriors’ front office? They used spreadsheets from ’98. My model used PyMC3 with priors informed by G League performance and college recruitment entropy.

Why Nobody Believed It

You see results as narrative. I see them as posteriors.

When you ask ‘Who got the pick?’—you’re asking which team has the highest marginal probability under your assumptions—not mine.

My model doesn’t cheer. It computes. And sometimes… it gets ignored.

You want to believe in magic? I prefer confidence intervals.