I spent a *lot* of time digging into sinkers. Please give this a read! Retweets greatly appreciated as well.

The “reckless close-out” is one of the silliest rules in the NBA. I get it, but there has to be some measure of intent, or making a non basketball play for a play to be deemed a flagrant.

@choice_fielder Hasn’t been a problem for me at all.

Can’t wait. Hope LeBron is back too.

It should be fairly obvious where in the process I implemented momentum.

@RobertStock6 Great question that I hope to be able to answer. Model is still training, working on building the analytical harness that will make this useful.

Here's the same chart, but minimum 75 pitches. I'm giving the network full info on each pitcher's arsenal, so the Imanaga and Stone predictions are really interesting. Again, need to check against the validation data (soon).

RT @srbrown70: How much of the times through the order penalty is familiarity and how much is fatigue? If familiarity, is it familiarity wi…

Hah! forgot to post the chart with Hader's sinker. Here it is:

I'm going to do a "Pitching+" with the base model, and then a version which assumes a natural target point for each pitch type, to simulate a "stuff" score. Hopefully by next weekend. Model is still improving slowly on the validation set, but it's mostly stable at this point.

I'm hoping it will start getting more aggressive predicting HRs for certain pitches, and push their probability up to 5% or so. This has been happening very slowly, but it's happening.

A pretty chart that thinks Jose Hernandez and Isaiah Campbell are interesting. Hernandez was scooped by the Dodgers, so clearly I'm on to something? Super small sample for those guys (bubble size is # of pitches).

Finally, splitters, which include Splinkers. Nice to see the names I expect where I expect them.