Do hitters decline after the Home Run Derby?

image
After hoisting the hardware at last year’s HR Derby, Justin Morneau underperformed in the second half. Is he the exception or the rule? (Icon/SMI)

For years now, we’ve heard how players who participate in the Home Run Derby screw up their swing or tire more easily in the second half of the year. It’s gotten to the point where players are declining invitations to the Home Run Derby in droves. Major League Baseball seemed to have a particularly tough time filling out the American League side this year. To my knowledge, however, no one has actually tested this theory. Today, I’d like to do just that.

Parameters

What I’ve done is compare second half performance with preseason Marcel projections for every Home Run Derby participant since 2001 (excluding Evan Longoria in 2008, whose Marcel projection would have been league average as a rookie). I’ve also adjusted the second half numbers to account for the fact that the league as a whole hits home runs at a slightly higher rate after the All-Star Break.

You may be wondering why I’m using projections instead of comparing the first half to the second half. This is because, had I done this, I’d be inviting a whole deal of bias into the equation, the biggest being selection bias.

If a player overperforms his true talent level in the first half, he stands a better chance of being selected to the Derby. Because he overperformed, though, he’s bound to play worse in the second half. A great example of this is Alex Rios in 2007 (Marcels AB/HR: 38; first half AB/HR: 21; second half AB/HR: 42). While it may have looked like he declined, he actually just regressed back to his true talent level.

To help solve this problem, I’m using projections to estimate true talent level and then seeing if the player underperforms this level in the second half. Ideally, I’d be using mid-season projections to account for the undoubtedly good first halves of these players, but this isn’t readily available and would take a long time to calculate.

Results

Here are the aggregate results for every year since 2001 (the first year Tom Tango published Marcel projections) as well as the combined results. Remember that for AB/HR, lower is better (it tells us the average number of at-bats a hitter takes in-between home runs).

+---------+---------------+----------+
| Year    | Marcels AB/HR | 2H AB/HR |
+---------+---------------+----------+
| 2008    |          20.7 |     25.5 |
| 2007    |          18.9 |     17.2 |
| 2006    |          19.7 |     15.2 |
| 2005    |          19.9 |     17.7 |
| 2004    |          15.4 |     16.0 |
| 2003    |          18.8 |     16.7 |
| 2002    |          15.2 |     15.6 |
| 2001    |          15.7 |     11.0 |
+---------+---------------+----------+
| Overall |          17.7 |     16.3 |
+---------+---------------+----------+

As you can see, the Home Run Derby hitters seemed to outperform their preseason Marcels every year except 2008, 2004, and 2002 (though the latter two only showed small differences). Despite conventional wisdom, it doesn’t look like derby participants play any worse in the second half of the season (on the whole). If you’re looking for the results in terms of percentages, 57 percent of derby participants outperform their projections in the second half.

Of course, this shouldn’t be a huge surprise since a hitter who is invited to the Derby likely will have improved his preseason projection by the All-Star Break, but even if we accounted for this, it’s very doubtful the results would swing so far in the other direction that it would confirm the conventional wisdom.

Another theory might be that players who last longer in the Derby or hit more home runs during it are more likely to decline.

+---------+--------+---------------+----------+
| Round   | Sample | Marcels AB/HR | 2H AB/HR |
+---------+--------+---------------+----------+
| 1st Rnd |     63 |          17.7 |     16.3 |
| Semis   |     32 |          17.3 |     16.3 |
| Finals  |     16 |          18.8 |     17.6 |
| Champ   |      8 |          20.1 |     17.6 |
| 20+ HR  |     14 |          19.2 |     17.7 |
+---------+--------+---------------+----------+

Nope, doesn’t seem to be the case. No matter how long a hitter lasts or how many home runs he hits, we still don’t see any signs of a second-half decline.

So where has this theory come from?

While the theory doesn’t appear to be true, we’re still likely to hear about it from the mainstream media over the next few hours and days. Why does the media seem to believe this, though? Here are a few possible reasons:

Last year: 2008 seemed to prove the theory in a big way, so it’s fresh in everyone’s mind.
The selection bias I mentioned earlier: Those selected likely overperformed in the first half, so second-half regression to the mean is viewed by the uninformed as a decline and not normalization.
Raw totals: Because the 50 percent mark often occurs a couple weeks before the All-Star Break, “first half” totals can look inflated if compared directly to “second half” totals.
Outspoken players: Media is a lot more likely to listen to players than numbers, and when players start blaming the derby for second-half struggles, it’s an easy story to run with.
Snowball effect: Once players start talking and complaining, it makes other players less likely to want to participate and draws more attention to the situation, creating a snowball effect.

Study caveats

There are a few caveats to this study.

Use of preseason projections: I mentioned this earlier, and it likely wouldn’t have changed the conclusions, but it warrants mentioning again.
Generalizing to all players: This study looks at the participants on the whole. We are dealing with human beings, though, each having their own unique swings and physiologies. It’s entirely possible some players are affected by the Derby, even if the overall effect is small.
Derby participants: There might be some additional selection bias in who participants in the Derby. If a player is legitimately affected by the Derby, he is less likely to participate in future years and thus will only be included in the study once.
Steroids: A study like this necessitates using many years since we only have eight sample points per year, but in doing so we look at years when guys like Barry Bonds, Sammy Sosa, and Jason Giambi were playing. Can we really say that the effects in these years will be the same as those in 2009? (if we only use 2005-2008, however, we still see a 19.7 Marcel AB/HR to 18.0 second half AB/HR)
Small sample: Because we only see eight hitters participate per year, there’s no choice but to try and draw conclusions from a small-ish sample size.

2009 participants

So what does this mean for the participants in tonight’s 2009 Home Run Derby?

Joe Mauer
Brandon Inge
Nelson Cruz
Carlos Pena
Albert Pujols
Adrian Gonzalez
Prince Fielder
Ryan Howard

While you likely don’t have to worry about any of these guys falling off a cliff in the second half, there is an opportunity to be had for fantasy owners. If the owner of any of these players is worried, you might be able to acquire him at a discount, especially if someone puts on a Josh Hamilton-esque show tonight.


6 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Derek Carty
14 years ago

Interesting article here, titled “Past participants deny Derby curse”: http://mlb.mlb.com/news/article.jsp?ymd=20090712&content_id=5852516&vkey=allstar2009&fext;=.jsp

Dan Turkenkopf
14 years ago

Nice article Derek.

In fact, I was going to write a very similar one for Wednesday.

Instead of projections, I was going to look at fly ball percentage and home run / fly ball percentage to see if those fluctuated at all.

Derek Carty
14 years ago

Thanks, Dan.  That would be interesting to see.  I’d guess that HR/FB would drop from the 1H to the 2H because of the selection bias issue, although it’s possible FB% might not drop as much.

Philip Christy
14 years ago

What you clearly need here is a control group. You could do something like find an equal number of players from each season who “could have” played in the derby (“could have” being subjective of course), finding guys who had comparative numbers to the Home Run Derby participants in terms of home runs, and look at how their second-halves compare to this group.

Alan
9 years ago
Reply to  Philip Christy

As Philip Christy said, you should use a control group. As you mentioned in the article, “a hitter who is invited to the Derby likely will have improved his preseason projection by the All-Star Break.” It is the ultimate selection bias when you use home run derby contestants as your selection criteria without providing a control group.
I further postulate that home run derby contestants tend to be break-out players (thus out-performing their Marcels), because you see many new faces on that list every year. At that point, comparing their numbers to a preseason projection is not very useful.

Dan Turkenkopf
14 years ago

Yeah, I was thinking that was pretty much how it was going to go too.

I’m interested in how those numbers compare to career rates though.  I would guess that the first half numbers are higher (more conducive to homers) and the second half numbers are along the lines of the career.

But I’ve scrapped that avenue for now to work on a new topic for this week.  If I get a chance, I’ll run the analysis the second half of the week.