The Quest to Predict HR/FB Rate, Part 5

Before diving back into this, after our long weekend away, I want to let you know what to expect this (hopefully this) week: Part 5 is clearly here right now, and Mike and I are working on parts 6 and 7 as we speak. In Part 6, I’ll be looking at the equations Mike came up with the other day along with one other equation that smarter statistical minds than mine are currently developing. Ideally, part 6 will be something of a wrap on our attempt to find an equation for expected HR/FB within a season. In Part 7, Mike will take a deeper look at a handful of players, and see how useful this new equation really is.

Today, I’ll be looking at another piece of the puzzle – what happens in future years to guys who show breakout distance on their fly balls.

First, for those interested:

Part 1
Part 2
Part 3
Part 4

Now, what we basically found through that analysis was that you can do a pretty good job predicting HR/FB rate based on distance, and a slightly better job if you include batted ball angle. But the question I have still wanted to dive into is what happens when a player’s distance (or perhaps angle) suddenly changes?

Throughout this process, Mike and I have been creating new data sets to fit our needs, depending on the analysis we were doing. Over the last few days, we have compiled a master set that we will use for all analysis moving forward. This way we know we are comparing apples to apples. Using this data set, I created a list of 807 instances of a single player hitting at least 25 fly balls in three consecutive seasons between 2008 and 2012. Keep in mind, that a player or a season can be on this list multiple times. Dustin Pedroia, for example, makes the cut in all five years, and so he appears on the list thrice – 2008-09-10, 2009-10-11, and 2010-11-12.

Let’s start with a smattering of correlations to get acquainted with the new data:

Variable 1	Variable 2	Correlation
Y1 Distance	Y1 Angle	-0.01
Y1 Distance	Y1 HR/FB	0.71
Y1 Angle	Y1 HR/FB	0.12
Y1 Distance	Y2 Distance	0.61
Y1 Angle	Y2 Angle	0.12
Y1 HR/FB	Y2 HR/FB	0.67

A lot of this repeats what we saw before, which is good – means we are seeing the same results with this data set we did before. We already saw that fly ball distance correlates reasonably well year to year, but now we can see clearly that angle does not. This tells us a player who pulls the ball a ton one year, might not do that again the next, but a guy who hits the ball 300 feet one year is a decent bet to do it again.

So what happens when a player goes from hitting the ball 280 feet to 300 feet (or vice versa)? The table below shows the 807 instances above divided into cohorts based on the change in their FB distance between years one and two.

Year 2 Delta	Number	Y1 Avg Dist	Y2 Avg Dist	Y3 Avg Dist	Y1 HR/FB	Y2 HR/FB	Y3 HR/FB	Y2 Dist Gain	Y3 Dist Gain	Y2 HR/FB Gain	Y3 HR/FB Gain
Lost 15+	114	294.3	274.0	281.2	15.71%	10.81%	14.12%	-20.3	-13.2	-4.90%	-1.59%
Lost 10-15	86	289.1	277.6	278.2	13.77%	10.86%	12.52%	-11.6	-10.9	-2.91%	-1.25%
Lost 5-10	138	287.7	280.4	278.0	14.02%	12.05%	12.37%	-7.3	-9.7	-1.98%	-1.65%
Lost 0-5	144	288.5	285.9	283.3	14.37%	13.54%	13.62%	-2.6	-5.2	-0.83%	-0.74%
Gained 0-5	115	284.6	287.0	282.7	12.75%	14.21%	13.03%	2.4	-1.9	1.47%	0.28%
Gained 5-10	99	281.9	288.7	282.3	11.92%	14.38%	12.84%	6.8	0.4	2.46%	0.92%
Gained 10-15	66	280.3	292.2	280.6	11.10%	15.74%	12.31%	11.9	0.3	4.63%	1.20%
Gained 15+	45	275.9	295.2	285.0	9.29%	16.58%	13.44%	19.3	9.1	7.29%	4.15%
Total	807	286.7	284.1	281.3	13.43%	13.20%	13.07%	-2.6	-5.4	-0.23%	-0.37%

First a note that the Y3 Gain columns represent the difference between Year 3 and Year 1 – so the 9.1 ft gain for the bottom cohort means that their Year 3 Distance was 9.1 ft longer than their Year 1 Distance. It’s also important to note that average fly ball distance decreased in subsequent years. The average fly ball distance for year one of these instances was 286.7. In year two is was 284.9 and by year 3 it was 281.3. So when you see that the players who gained 5-10 feet in year two kept only .41 feet by year three, keep in mind this is compared to an expected drop of over 5 feet.

The reason for this likely ties to something Eno Sarris wrote about just over a year ago – power seems to peak around age 24. The average ago of players in year one was 28.4, meaning most of the players in our sample should be expected to see some decline in their power over the subsequent years, when they turn 29 and 30, on average.

The columns to really focus on are the last four. There you can see that the group of players who gained 15+ feet in year two maintained almost half the distance they gained in Year 3. Similarly, they saw their HR/FB rate jump by 7.3 percentage points in year two and kept just over half of that (4.1 percentage points) in Year 3.

The key things that jump out at me are as follows:

1) We should expect most players to see a decline in distance year over year. Nothing huge, but two to three feet can make a difference.
2) A player who gains a big chunk of distance will likely keep about half that change, but this really only holds if the distance gained is 15 feet or more.
3) Players who gain distance at a lower rate in year two will like fall back to something close to their year one distance in year three.
4) Players who gain less than five feet in year two (or who lose distance in year two) are likely going to put up less distance in year three than they did in year two.
5) Players who take a big hit on distance (10+ feet) tend to get a bit better the next year – but they are the only ones who don’t see a drop from year two to year three.

The summary – players tend to lose distance and when players have a big change, lost distance is more likely to stick than gained distance.

Not surprisingly, the HR/FB rates generally follow the same pattern, although players seem to hold MORE of their gains in HR/FB than they do in distance. For example, the 15+ cohort kept 47% of their distance gains and 57% of their HR/FB gains.

For prognostication purposes, this tells me to do the following when you see a player suddenly make a big HR/FB gain:

First, see if the player saw a big increase in distance. For example, Chase Headley increased his HR/FB from 4.3% in 2011 to 21.4% in 2012, and increased his FB distance by more 16.9 feet (from 282.2 to 299.1), putting him in our elite class of gainers. Billy Butler, on the other hand, increased his HR/FB from 10.4% to 19.9% (a gain of 9.5 percentage points) but actually lost distance, going from 297.6 to 297.2.

Next, I would assume the player will lose some of their distance. How much they lose depends on two things – how much they gained in the past year and how old they are.

Age	Y1	Y2	Diff
<24	287.9673542	286.8532084	-1.114145822
25-28	287.2298092	284.821436	-2.408373166
29-32	286.9610965	283.5338895	-3.427206983
33+	284.027566	280.0966305	-3.930935448

As you can see, older players lose distance faster (shocking, right?), but as we saw above, the biggest gainers (like Headley) are the only ones who keep much of their distance gains, if any. Headley and Butler are 28 and 27, respectively, so there is no reason to think their age will have a major impact. So maybe the Headley will end up closer to 290 next year, about an eight foot increase over year one, while Butler will end up close to 294 feet – about the 3 foot drop we’d expect from a player year over year.

Finally, I’d assume their HR/FB rates will see similar regression – with Headley settling in around 14-15 and Butler barely ticking up above the 10% he had in 2011.

The piece still missing is using the xHR/FB equation to see if their HR/FB rates were even reasonable in those years. For example, if Headley’s 2011 xHR/FB turns out to be 10%, you might expect him to stay a bit higher in 2013, keeping more of his gains from 2011. If, on the other hand, his 2012 xHR/FB turns out to be 15%, you might expect him to drop even lower, keeping less of his 2012 gains.

But we’ll get into more of that in parts 6 and 7.