RESEARCH: How to estimate home run output for young hitters

We've done a lot of research into developing various metrics, and each has its value in evaluating a hitter. What we perhaps haven't done as well is to figure out how to use these metrics. It's great to have a bunch of ways to break down player performance, but the real meat of fantasy baseball (and is in using what we know and what we've learned to guide us in making decisions about future performance.

We have an alphabet soup of metrics that measure power in one way or another: FB%, PX, xPX, HH%, hr/f, and HctX. These all have value in context, but what is the context? We know what each of them means, but how well do they portend future changes in performance?

In thinking about the basic hitting stats, we generally view runs and RBI as situational. Sure, better batting average and more power will help, but only so much. Steals are almost in their own category—while OBA plays a big role, it's really a player's raw speed, ability to read pitchers, and the green light from his manager that are the main factors. So really, the bulk of what we want to measure with our power metrics lies in batting average and home runs.

We'll leave batting average for another day. It's a very complex interaction between power, contact, speed, and batted ball trajectory. No single basic skill metric can capture these disparate parts, which is why we have xBA to guide us.

That leaves home runs.

Let's consider what skills produce home runs (assuming contact is made):

  • Bat speed/strength: specifically, the ability to hit the ball hard. Do we have something that measures this? Yes! Two, actually: hard-hit ball rate (HH%) and expected power index (xPX). Since xPX combines HH% and medium-hit balls into a metric that is correlated to actual power, this is our preferred skill.
  • Batted ball trajectory: most home runs are from fly balls, while a handful are the result of line drives. We use FB% as our primary indicator of potential home run trajectory.

So we need xPX and FB% to understand home run output. Is that it? What about PX? After all, it's a staple. Recall, though, that PX is based on linear-weighted power, which uses doubles, triples, and home runs as inputs. Because hits and hit types are outcomes, they aren't ideal to use as skills. We want to compare outcomes to skills to see where the outliers are.

So what about HctX? That's a measure of skill. Yes, but because it combines two skills in one (contact rate and hard hit ball rate) and includes HH% on ground balls, it's not a perfect fit.

Finally, what about hr/f? Interestingly, hr/f is a sort of intermediate measure in that it is an outcome (home runs) normalized for a skill (FB%). So it actually can be useful because it allows us to understand, for a given level of fly ball trajectory, what was a player's home run output.

So what happens when we plot xPX (our first home run-related skill: strength) with hr/f (our outcome of interest, normalized for our second home run-related skill: trajectory)?

                       hr/f percentiles*
xPX       #     10     25     50     75     90     
===      ===  =====  =====  =====  =====  =====
<=70     204   0.9%   2.0%   3.8%   5.5%   7.4% 
71-80    104   3.3%   5.1%   6.4%   8.1%  10.0%
81-90    111   3.8%   5.4%   7.4%   9.0%  11.0%
91-100   125   4.7%   6.6%   8.9%  11.3%  13.0%
101-110  133   6.6%   8.3%  10.9%  13.0%  16.2%
111-120  138   7.4%   9.8%  11.9%  14.7%  17.1%
121-130  118   8.5%  10.9%  12.8%  15.5%  17.4%
131-140  119   9.7%  11.9%  14.6%  17.1%  20.4%
141-160  116  11.3%  13.1%  16.5%  19.2%  21.5%
161+      55  14.4%  16.5%  19.4%  22.0%  25.8%

*Reflects all hitter seasons from 2010-2014 with 300+ AB.

This is what we'd expect overall, as the hr/f increases at each level of power. We can now use this to validate past performance and better predict future performance, as players who stray too far from the center can be expected to regress given their xPX. Players in the 75th percentile and up will likely see a decline in hr/f and players in the 25th percentile and below will likely see an increase. Here's the evidence:

The average change in hr/f in the following season:

%-ile  Avg. Change
=====  ===========
 10th     2.1  
 25th     0.4
 75th    -1.2
 90th    -4.1

Note that on both the chart and the table, there's some survivor bias in the bottom percentiles. Players who underperform are at greater risk of losing their jobs, and those who experience positive regression are more likely to hold on to their jobs. Still, the results are fairly robust, making them useful for finding outliers.

Application of results

For established players, we already know that they tend to regress toward their three-year average hr/f rate. So where this newly understood relationship between xPX and hr/f may really come in handy is where we need to evaluate hitters with short major league track records who aren't as easy to pin down.

Let's take a look at some younger players whose 2014 hr/f was in the highest or lowest percentiles and compare their actual performance to what it might have been if their hr/f landed in the middle of the pack (the 50th percentile would be our expectation short of additional evidence) based on their xPX level.

Possible Faders

Hitter              xPX  Pct.-ile  HR  xHR*  Diff
==================  ===  ========  ==  ====  ====
Jose Abreu          132      90    36   19   -17
Tyler Flowers        88      90    15    5   -10
Corey Dickerson     135      75    24   18    -6
Kole Calhoun         94      90    17   11    -6
Rougned Odor         68      90     9    4    -5
Jonathan Singleton  105      75    13    9    -4
Kevin Kiermaier      82      90    10    6    -4
Kolten Wong          82      75    12    8    -4
Daniel Santana       63      90     7    3    -4     

Possible Surgers

Hitter              xPX  Pct.-ile  HR  xHR*  Diff
==================  ===  ========  ==  ====  ====
Anthony Rendon      146      10    21   34    13
Nick Castellanos    135      10    11   21    10
Xander Bogaerts     111      10    12   19     7
Matt Adams          130      10    15   22     7
Conor Gillaspie      95      10     7   13     6
Travis D'Arnaud     136      10    13   18     5
Nolan Arenado       134      10    18   23     5
Eduardo Escobar      93      10     6   11     5

*Assumes player is in the 50th percentile based on xPX.

As can be seen, the outliers are a varied bunch. While you should perhaps take with a grain of salt the restated 2014 HR estimates for Jose Abreu (1B, CHW) and Corey Dickerson (OF, COL), as both have significant power and play in hitter's parks, lesser hitters like Rougned Odor (2B, TEX) and Kolten Wong (2B, STL) should be viewed with more suspicion.

On the flip side, it's quite a surprise to see a Rockies' hitter in the 10th hr/f percentile. With park effects factored in, that 23 xHR would be closer to 30. Anthony Rendon (2B/3B, WAS) and Travis D'Arnaud (C, NYM) also look like potential breakouts. In fact, other than the below-average Connor Gillaspie (3B, CHW) and Eduardo Escobar (SS, MIN), all of the surgers look like strong draft targets.


The results here are promising. We would definitely expect a strong relationship between hard-hit balls (represented here by xPX) and hr/f, and the results bore that out. We also see high rates of regression, especially at the extremes.

In the context of young hitters without long track records to rely on, this may be our best method yet to explain and predict their power output. We'll keep a close eye on these players in 2015.

Click here to subscribe

  For more information about the terms used in this article, see our Glossary Primer.