FO INTEL

@903124S If by non-magnus you want to measure ssw, you won’t capture it like this bc ssw induced movement can actually be exactly in the direction of magnus

We know for baseball pitches that higher speed lead to better result so does higher break, but how does it individual component matter? How much does non-Magnus movement also contribute. This thread attempt to quantify them bits by bits 1/n 🧵

@TJStats GRIFFIN 1 LOL

My Top 100 Prospects

@DocEisenhauer Going to keep checking in on this one throughout the year, Doc. Try to stay about your wits, pal.

I’m seeing Bobby Witt Jr./Shohei Ohtani/Juan Soto as the undisputed top-3 dynasty assets via startup drafts. Konnor Griffin might be at #4 by the end of March if today’s moonshot clinic was any indication. #HitTool

Someone get this guy a glass of water

@tangotiger @drivelinekyle Tom, please implement a better way to visualize this data. There are many.

Statcast: Tatsuya Imai Mollweide Projection + Spin Axis Report His famous Fadeout-Slider is indeed spinning opposite of gyro-slider. Ive marked where gyro-slider spins for everyone His splitter/changeup are 1 seamers, but his splitter closer to his slider in seam orientation

Seen some stuff about cortisonal left disk blowback and angular femoral acceleration too afaik

@matthewk36711 Kahnle

Keaton Winn continues to test the upper bound of single-inning splitter usage, throwing 10/13 last night to strike out the side (full inning ⬇️). The pitch induced five whiffs on six swings and averaged 90.4 mph (career-high). That velocity with a -8.3 VAA is just ridiculous. On 27 splitters (66% usage) this year, he's thrown 19 strikes for 12 swing-and-misses. Just outrageous stuff.

@srbrown70 Actual question: Why would a stuff model be handedness neutral?

before evaluating the performance of a stuff model you've got to understand some of the choices that went into it. how arsenal aware is it (IE what primary fastball features does it include?)? is it count-aware or count-neutral? is it handedness-aware or neutral? etc.

@xwOBA_enjoyer Really good stuff Nathaniel

Top 10 pitchers by RV/100 swings in the model This shouldn't be interpreted as a strict ranking of pitchers, since some of these guys lose a lot of value on taken pitches. Not necessarily any of these pitchers, but a bit below them you have Dominguezes and Littles and so forth

@hankyklocko @LanceBroz Who was talking about MIL? No bias I’m sure 👍

@MLBFrOfcTakes @LanceBroz Comparing those guys to the brewers developing woodruff burnes peralta, now Miz. Hader, Devin Williams, Abner Uribe, Megill. Laughable

I sent a survey on pitching development and acquisition to 68 #MLB coaches and executives. Who is the best? Who is the worst? Who will rise? Results are below. What stands out? More in thread... 👀

@hankyklocko @LanceBroz You’re cooked bud

@MLBFrOfcTakes @LanceBroz Didn’t develop fried. Weavers a fine reliever, let’s give cam Schlitler more than a few starts. Warren isn’t that good. Nestor’s a bum lol.

@RBunikiewicz Shannon entropy

@MLBFrOfcTakes Could you please elaborate?

The Most Diverse Pitch Usage in a Start? I looked at the most expansive arsenal in the MLB over the past 2 years, Seth Lugo, to see how diverse his pitch mix got. I generated an 'evenness' score to see at what points it was its most diverse and unpredictable.

@hankyklocko @LanceBroz Fried, Weaver, Schlit, ERC, Warren, King, Nestor the list goes on buddy

@MLBFrOfcTakes @LanceBroz Who exactly have they developed? “Luis Gil” or “Will Warren” lol

@DocEisenhauer Ok, “Doc”

@pirates_at448 @Trevor40878381 @LanceBroz @pitchprofiler No, the smart teams either: a) can give almost every pitcher a good/better pitch b) are so good in their analysis they can identify before acquisition what they can change

@MLBFrOfcTakes @Trevor40878381 @LanceBroz @pitchprofiler A trend amongst smart big league to introduce “bad” pitches specifically so the mix is deeper and forcing hitters to consider more pitches

@hankyklocko @LanceBroz 🤣

@MLBFrOfcTakes @LanceBroz Yankees? lol

@903124S I see. No it shouldn’t be regressed to the mean. Like FIP, WOBA, xWOBA, etc. it is, at the end of the day, a descriptive metric that happens to be predictive. If you are using it to project pitch type RV then sure do some regression to mean. Otherwise, no.

@MLBFrOfcTakes For point 4 using statistical trick it's always better if we regress any metrics (i.e. stuff number here) to the mean. While ultimately it doesn't matter too much it could help e.g. stuff leaderboard in spring training or at start of season

There are lots of pitch stuff model out there (e.g. most recently by @RobertStock6 without coding knowledge) and they are using gradient boosting or similar algorithm as bases. But a pitch job is mostly 1. throw harder and 2. more break and we can take advantage of it🧵 1/n

@903124S 4. ? 5. Should be explicitly

4. Should it consider sample size? (also given starter and reliever throw much different number) 5. How the run value (most popular way to calculate stuff implicitly) scale with actual run surrendered by pitcher? 6. How do someone (outside of teams) use stuff number for ??? 11/n

@903124S 1. Doesn’t matter 2. No 3. No Model should see a pitch as appearing out of a pitching machine or another universe/dimension (besides biomechanical effects, most obviously release characteristics, most commonly used: arm angle, though this is not close to the full picture)

Also there are worthwhile questions on how exactly should stuff model develops: 1. How should stuff scale? (50 +/-10 or 100 +/- 20 or run value?) 2. Should stuff be pitch type dependent? 3. Should it consider starter/reliever split? 10/n