Derek Grifka

Data Scientist @ChicagoFire. Check out my baseball bot:

[Derek Grifka @DerekGrifka]

Wanted to put my MLB data statistical chops to the test and create an account that simulates previous game outcomes (including exit velo and launch angle) and compare it to what actually happened. Code is in my repo, which can be found in the account’s bio. Pls reach out with feedback!

[MLB Deserve-To-Win-O-Meter @mlb_simulator]

Simulations for Twins at Astros on 06/02/2024. Actual score: Twins: 4 / Astros: 3. Simulation win percentages: Twins: 41% / Astros: 48% #MLB #Stacast #MLBSim #Twins #Astros

[Derek Grifka @DerekGrifka]

@YungGiamatti King

[Derek Grifka @DerekGrifka]

@srbrown70 @RosenJordanBlum 💪

[Derek Grifka @DerekGrifka]

The irony of a Chris Sale jersey patch card for the White Sox is insane

[Derek Grifka @DerekGrifka]

Catfish Hunter should be on this mural next to Tim Salmon and Mike Trout

[Derek Grifka @DerekGrifka]

@Chandler_Rome Can he pitch

[Derek Grifka @DerekGrifka]

@jacklambert__ Is this from a Bloomberg terminal

[Derek Grifka @DerekGrifka]

@AidanResnick Do NOT underestimate the Tacoma / Sugar Land rivalry 😤

[Derek Grifka @DerekGrifka]

@srbrown70 Thinking about old projects you coded w/o Claude

[Derek Grifka @DerekGrifka]

@RosenJordanBlum Is Oliver aware that RHPs are averaging over 95mph on their fastballs

[Derek Grifka @DerekGrifka]

@srbrown70 If it uses JAX then it might not matter? I haven’t tested nutpie and I probably wasn’t using pymc efficiently but I’ve found numpyro to be much faster (esp with @ JIT decorator)

[Derek Grifka @DerekGrifka]

@srbrown70 Agree. Bambi was helpful pre-Claude (P.C.) lol. I love pymc but admittedly been a numpyro boi recently 🫣

[Derek Grifka @DerekGrifka]

@srbrown70 Yes my b the pitching dist - that makes sense. This is sick! Have you used the pymc skills with Claude code yet? Super helpful

[Derek Grifka @DerekGrifka]

@srbrown70 Right but seems like Shohei’s modeled dist doesn’t follow his underlying pattern and is heavily influenced by the league dist. Or is the “modeled” dist the league dist and Shohei is just there for reference?

[Derek Grifka @DerekGrifka]

@EDLAnalytics It’s been cool seeing your work evolve, Evan. Nice work

[Derek Grifka @DerekGrifka]

@astro_numbers I’m genuinely curious haha

[Derek Grifka @DerekGrifka]

Been on a heater recently 😮‍💨

Vamos Fire btw

[MLB Deserve-To-Win-O-Meter @mlb_simulator]

What does Est. Bases per Plate App (EB/PA) = 0.75 actually mean?

Example Aaron Judge (2025). Each plate appearance produces 0 to 4 estimated bases:

Strikeouts score 0
Walks/HBP score 1
Batted balls range from ~0-4 (using the in-game model)

[MLB Deserve-To-Win-O-Meter @mlb_simulator]

How random is baseball scoring?

I tested two probability distributions against every 2025 MLB game. The Poisson distribution gets the general shape right but badly underestimates shutouts and blowouts.

The Negative Binomial instead captures this variance well, including the extremes.

#MLB #Probability