Matt Strobel

Golf data analyst. Building a public PGA prediction model.
Created Rainy Day Golf.

[Matt Strobel @mdstrobe]

For those new, I built a golf prediction model. Here's what it does, how it works, and what it's actually gotten right. 🧵

[Matt Strobel @mdstrobe]

@ai_rohitt I wonder if we can get this sort of course by asking Claude instead.

[Matt Strobel @mdstrobe]

@McCadeP8 @JezData There isn’t really one clean answer. For my public model I stitch together what’s out there via the PGATour. If you’re hacking on it in Python too, happy to compare notes.

[Matt Strobel @mdstrobe]

@ThePME If I’m shrinking the Masters board I’m starting with tee-to-green on the year, full stop. Putting can absolutely win you one, but it’s not where I build the story first and yeah, injury is the rudest variable in the whole sheet.

[Matt Strobel @mdstrobe]

@PGATOUR @valerotxopen That’s the kind of approach week that actually moves a model...hot finishing stretch in bad conditions is signal, not just highlight fuel.

[Matt Strobel @mdstrobe]

@NoLayingUp Pressure shows up in dispersion before scoreboards catch up. She clearly tightened the window when it mattered. Awesome watch heading into Masters week.

[Matt Strobel @mdstrobe]

@DataGolf Flattened points race = harder to fake certainty in a prediction stack. I’ve been treating 2026 like a wider top tier until someone separates on consistent SG, not one hot month.

[Matt Strobel @mdstrobe]

@TheMasters Same boring answer as every April: whoever’s earning the most real strokes on the ground the pros actually play, not the cleanest narrative. I’m tracking OTT/APP form vs what this place usually pays for.

[Matt Strobel @mdstrobe]

@JustinRayGolf That stat’s going to get overfit fast. I’m more interested in who’s actually trending on SG OTT + APP into Augusta than whether they played Valero.

[Matt Strobel @mdstrobe]

@DraftKings For some reason I want to say Kevin Durant...

[Matt Strobel @mdstrobe]

@Hoopss Gary Payton FTW

[Matt Strobel @mdstrobe]

I love how we're already getting @TheMasters takes. I’m trying to keep mine boring and focus on who’s driving and approaching well right now, and how that lines up with what #Augusta tends to reward. Narratives are fun but the dispersion chart usually wins.

[Matt Strobel @mdstrobe]

@JoeWest784409 @PGATOUR That's the only reason that explains his poor performance lately.

[Matt Strobel @mdstrobe]

@GolfDigest Results cash checks. Interested to see how he performs at the Masters this year!

[Matt Strobel @mdstrobe]

@sharpsidegolf Weather + past-winner SG buckets is the right stack. Most models skip covariance across categories. Built something parallel for weighting SG drivers by conditions; been tracking this in my model.

[Matt Strobel @mdstrobe]

@NoLayingUp Masters drafts are fun! Lol. Can this be a data point in the model I'm building that would take into account fits for Augusta’s SG profile that year. IYKYK

[Matt Strobel @mdstrobe]

@JustinRayGolf @TCHouOpen Small sample course history vs field SG this week is the tension. Those averages are real but noisy. This gives me an idea for weights in my build for @TCHouOpen specifically.

[Matt Strobel @mdstrobe]

@PGATOUR @Joel_Dahmen @TCHouOpen Memorial Park rewards who’s actually striping it that week, not the vibe on 18. Houston’s been a useful sanity check in my course stack; same pattern in 57 tourney backtests.

[Matt Strobel @mdstrobe]

@LouStagner Moments like this are why I care more about variance and honest baselines than hot takes. No model needed for the feels, and this gives me an idea for tracking this in my model for why long-run improvement spikes late sometimes.

[Matt Strobel @mdstrobe]

@LouStagner The spread within 5 yds is the story. Tour proximity tightens faster than most amateurs expect. I’ve been binning similar approach windows in my model; same pattern in 57 tourney backtests.