Louis Oberdiear

Senior Data Scientist. Sports Analytics enthusiast.

[Louis Oberdiear @STLouisO]

@kurt13warner Here is a write up of the logic:

https://t.co/iwudrbHR1i

[Louis Oberdiear @STLouisO]

@tangotiger can you find defender jump data on baseball savant?

[Louis Oberdiear @STLouisO]

@DownToBlack Where do you get the jump data?

[Louis Oberdiear @STLouisO]

Lamar's performance today

[Kevin Cole @KevinCole___]

In a year with a lot of inexperienced quarterbacks, there is no playoff distribution of efficiencies for single-game starters (Stroud, Love, Purdy). Here are the rest.

Number of games in the samples
Mahomes: 15
Mayfield: 3
Allen: 9
Goff: 7
Jackson: 4

[Louis Oberdiear @STLouisO]

@tangotiger Do you have a blog post or a resource that explains this?

[Louis Oberdiear @STLouisO]

But surely we aren't that confident in the accuracy of the model to be as close to a percentage point to the 'true' win probability.

[Louis Oberdiear @STLouisO]

@bburkeESPN when utilizing win probability models to determine 4th down decisions, should we be using prediction intervals also? If a prediction interval crosses zero then we would call it a toss up? I see people saying +1% and they should go for it

[Louis Oberdiear @STLouisO]

@tangotiger So I think I understand. Throwing a ball on a 0-0 count adds 0.04 runs. Are you saying my model should be closer to 0.04 on average when a ball is thrown on a 0-0 count?

[Louis Oberdiear @STLouisO]

@tangotiger would love to get your thoughts on this new metric I created to try to measure pitch quality

https://t.co/Za0fWVv3gl

[Louis Oberdiear @STLouisO]

@tangotiger Something like this?

[Louis Oberdiear @STLouisO]

@tangotiger Does this model not account for the count? Most of the highest xp_wOBAs are bad pitcher counts like 3-1 and 3-0 and the lowest are good pitcher counts like 0-2 and 1-2. The model basically replicates that first chart but then adds in extra features.

[Louis Oberdiear @STLouisO]

@tangotiger Interesting. The goal was to determine the EV of a single pitch. How could you do that when incorporating pitches that don't have a value?

[Louis Oberdiear @STLouisO]

@tangotiger I wasn't aware of the more detailed zones! Creating custom zones was on the to-do list.

[Louis Oberdiear @STLouisO]

@tangotiger Thanks! I only modeled with event data. So if a batter takes a first pitch then the expected value is in terms of if the PA were to end on that pitch. Basically, were the batter to put it in play, what is the expected value.

[Louis Oberdiear @STLouisO]

⚾ Can we measure the quality of a pitch in the MLB? ⚾

I developed an xgBoost model that attempts to do that.

#RStats
#DataScience

https://t.co/Za0fWVv3gl

[Louis Oberdiear @STLouisO]

Predicting Customer Tenure...with MLB Hitters.

How accurately can we predict how long a MLB hitter will stay in the league? Who are the top young players we expect to be in the league the longest?

#datascience #rstats #mlb

https://t.co/6HRFVNpVjv

[Louis Oberdiear @STLouisO]

⚾️ Customer Segmentation...with MLB Players Pt. 1

What are the different types of hitters in the MLB? I tried to find out here.

https://t.co/y8bvEXLmFb
#datascience #rstats

[Louis Oberdiear @STLouisO]

@jesseagler did definition of shift change at all in those years?

[Louis Oberdiear @STLouisO]

@DSzymborski and you weren't streaming? Shame.

[Louis Oberdiear @STLouisO]

@drewsilv even bad players have had streaks like that or better in 41 plate appearances #sss