FG on Fox: Breakout Sluggers, Predicted by Fastballs
You can tell a lot about a hitter by how many fastballs pitchers are willing to throw him. The bigger the bat, the more likely it is to see junk. Turns out, small changes in the number of fastballs a hitter sees can help us project that hitter better.
Sort the leaderboard for lowest fastball percentage, and you’ll see it immediately. It’s full of sluggers at the top. Reverse the filter and it’s mostly slappy speedsters. Rob Arthur took a more scientific approach and showed that isolated slugging and fastball percentage are indeed correlated negatively — sluggers see fewer fastballs.
Rookies see more fastballs when they come into the league. Over the last five years, the league saw 57.5% fastballs, and rookies saw 58% fastballs. That’s not a large difference, but it comes in a large sample. Then again, it’s not a large difference, period. Over the course of a season, a rookie with 600 plate appearances would be expected to see 12 or so extra fastballs.
In any case, even if this effect is small when you zoom out, it seems that individual differences in fastball percentage are predictive of future strong work. As Arthur said when he did the gory math behind this statement, “Fastball frequency normally varies according to the pop of the batter, so that when it changes, it may be indicating a change in the underlying skill level of the same batter.”
Read the rest on Just a Bit Outside.
With a phone full of pictures of pitchers' fingers, strange beers, and his two toddler sons, Eno Sarris can be found at the ballpark or a brewery most days. Read him here, writing about the A's or Giants at The Athletic, or about beer at October. Follow him on Twitter @enosarris if you can handle the sandwiches and inanity.
Close call. I thought this was going to be another list of players rumored to have bacne.
Hmm any examples of this being predictive for breakouts in past seasons?
Similar criteria (<800 Career PA, 150+ PA in 1st & 2nd Half)
Minimum 2% decrease in FB%
I've excluded players that did not accumulate at least 200 PA in subsequent season
First line shows cumulative change wRC+ from group of non-excluded players
2013 89 wRC+ 2014 99 wRC+ (9 players)
Improvements in wRC+: Mesoraco 147/74, Hechaverria 82/53, Gillaspie 108/85, Gattis 125/109, Marte 132/122, Barnes 83/75
Declines: Middlebrooks 44/83, Dominguez 63/88, Mercer 91/114
2012 115 wRC+ 2013 130 wRC+ (6 players)
Improvements: Goldschmidt 156/124, Carpenter 146/124, Trout 176/167
Declines: Cozart 79/83, J.Arias, 76/90, Lombardozzi 66/83
2011 Only 3 Players met criteria
Declines: Thole 61/94, Espinosa 94/103, Freeman 115/120
2010 96 wRC+ 2011 101 wRC+ (5 players)
Improvements: Avila 140/81, Boesch 116/96, LaPorta 96/86
Declines: Colvin 27/113, Au.Jackson 87/101
Awesome.
Looking over the list of least FB’s thrown to, I wonder if it has to do with protection (aka, lack of fear of the player hitting behind each player in their lineup). For example, seeing VMart and Castellanos but not Miggy; Brandon Moss but not Donaldson; Kyle Seager but not Cano.
I would say (anecdotally) a low FB% seems to come from a combination of power potential (fear), lineup construction and players who struggle with pitch recognition.
Incidentally, I wonder why are there 4 Royals in the top 21 when no other team has more than 2 in the top 30? Is there a book that says go offspeed against KC?
I’m obviously not grasping the concept because it seems backwards to suggest that pitch changes forecast a performance increase, as why would pitchers change their style based on something that hasn’t happened yet? Furthermore, wouldn’t a decrease in fastballs seen typically be bad news for future performance given that they’re typically hit more authoritatively than other pitches?
The idea is that pitchers are showing more fear of your ability to slug fastballs, ahead of even your actual results dictating that fear. Could mean the game plan for you as a hitter is changing.
Might be a back-end way of looking at data that the teams have that we don’t. In that, they see something that we couldn’t see, change their approach, then we see it.
As for the future, it’s a small effect, but it did have an effect on future power.
Why would they show fear of something that has yet to occur? That’s what I’m not getting. The premise appears to make no sense whatsoever given that would involve a total guess, and as such would be very unlikely to have consistent predictive value. Now if someone wants to argue that teams are sensitive to changes in performance and may adjust their scouting report before the general public realizes a performance improvement, I can get behind that, but even in that case we should be able to identify such an improvement more directly by looking at that recent performance (line drive rate, batted ball distance, ISO change, etc.) than by analyzing a secondary reaction to that change.
Not if teams have data that we don’t have, which they do. Say they see that dude has great exit velocity and angle on fastballs, something that they’ve correlated to success in the past — they’d tell their pitcher to throw fewer fastballs.
Or even simpler: we don’t know what goes into the heat maps and recommendations that teams give to their pitchers. They probably have more than us. If pitchers are throwing the hitters fewer fastballs, they are probably doing so for a reason.
Anyway, Arthur did the work, I was riffing on it. He showed it improved pecota despite its low r-squared.
“Now if someone wants to argue that teams are sensitive to changes in performance and may adjust their scouting report before the general public realizes a performance improvement, I can get behind that,”
That’s exactly what the approach argues, so you do understand it.
“but even in that case we should be able to identify such an improvement more directly by looking at that recent performance (line drive rate, batted ball distance, ISO change, etc.) than by analyzing a secondary reaction to that change.”
No, because short-term (half-year) changes in LD%, batted ball distance, etc. are much more likely to be noise than genuine performance improvements. That’s why using them doesn’t get you very far. On the other hand, the teams are better able to tell whose talent has changed (using the richer datasets they have access to like HitF/X), and will instruct their pitchers to compensate if player X made a mechanical tweak that makes them much more powerful.
Regardless of WHY it works, it does seem to work, and you can quibble with the statistics underlying it here:
http://www.baseballprospectus.com/article.php?articleid=25413
om, you undermined your own argument by pointing out that such small sample sizes would be overcome with noise. Small samples don’t suddenly become more predictive when analyzed by teams than by us. Moreover, you still have to deal with the problem of it being second-order information. Watching how pitchers treat a hitter will give you an inferior understanding of that hitter than evaluating the hitter directly, just as reading articles about a game will always give you an inferior understanding of that game than watching it directly. I think Eno just needed something to write about this week and couldn’t come up with any better ideas.
OK, jdbolick, you can win the argument, I’ll take the breakout predictor that works in a statistically robust way.
Because it does, regardless of whether it should or shouldn’t.
Shouldn’t we also look at how these players are hitting offspeed pitches? Pitchers may be afraid of their HR ability, or they might have figured out they’ll swing at every breaking pitch thrown in their general direction