The Cliff
There’s an obvious running theme to the posts of the last several days, mostly inspired by discussions about Frank Thomas’ release and whether he would help the Mariners offense. The underlying question essentially boils down to this: “Is Frank Thomas, at age 40, finally just finished as a major league hitter?” I’ve written on this subject, both at Fangraphs and here, as well as talking about it on KJR yesterday. My conclusion is that Thomas is not washed up, and while he’s never going to be the Big Hurt of the 1990s again, he’s got more to contribute as a major league hitter.
However, the Thomas discussion simply leads to one larger point – how do we know when a player really is done? I wrote up a post on Fangraphs about this topic this morning, but wanted to delve into a little more detail here, since this is an issue that Mariner fans have some experience with and it relates strongly to the 2008 roster. As Derek noted yesterday, there are some signs that perhaps Brad Wilkerson is finished as an everyday player. I’ve made no secret of the fact that I think Jose Vidro’s career is basically over at this point, and we’ve talked recently about Richie Sexson’s old player skills and what that might mean for the length of his career.
So, if we agree that there’s a point in time where a player’s skills just deteriorate to the point that he’s not longer a contributor, how do we know when a player has reached that point? What are the signs, and how do we see it coming?
As I said in the Fangraphs post this morning, I believe the first place you should look when determining how much a hitter has left to offer is his power. The ability to drive the ball is what causes pitchers to not just throw fastballs down the middle on every pitch, and once they’ve ascertained that a hitter can’t drive the ball with authority anymore, he’s not going to contribute for much longer. Now, every hitter has their own established level of power production that works for their particular skill set. We don’t want to look at an arbitrary number for every player, but instead, we’re looking to identify an erosion of that particular player’s established power level.
Thankfully, we have metrics available that measure power, and terrific sites like The Hardball Times and Fangraphs that give us historical data for research purposes. As Mariner fans, we’ve watched over the last decade as the M’s have hosted farewell tours for a lot of past-their-prime players. The M’s are notorious for trying to squeeze the last drops out of a proven veteran’s career, so we’ve had a front row seat to a lot of collapses. Among the notable ones we’ve seen are Bret Boone, Edgar Martinez, John Olerud, Jeff Cirillo, and Carl Everett, with bench players such as Mark McLemore, Eduardo Perez, and Dave Hansen also meeting their end here. We’ve seen a lot of career ending collapses, and we pretty much know what they look like. So, what do all these guys have in common? Take a look at these year to year changes in HR/FB rates:
Bret Boone: 2004 – 14.1%, 2005 – 6.5%
Edgar Martinez: 2003 – 16.6%, 2004 – 9.2%
John Olerud: 2002 – 12.2%, 2003 – 6.1%
Carl Everett: 2005 – 14.4%, 2006 – 9.3%
Mark McLemore: 2002 – 8.1%, 2003 – 2.5%
Eduardo Perez: 2005 – 22.4%, 2006 – 5.3%
Dave Hansen: 2004 – 10.0%, 2005 – 7.5%
All of these guys experienced significant dropoffs from their previously established levels of power. They were still hitting the ball in the air, but it wasn’t clearing the fence anymore, and that switch to having warning track power signaled the end of their playing careers. More than anything else, it’s this loss of power that we can look at if we want to know if a player really has anything left, or if his current performance is just a short term anomaly.
So, using HR/FB% as our window, let’s look at some of the recent subjects we’ve talked about.
Frank Thomas: 2006 – 17.4%, 2007 – 11.1%, 2008 – 11.5%
Jose Vidro: 2005 – 8.0%, 2006 – 5.3%, 2007 – 4.1%, 2008 – 7.7%
Richie Sexson: 2006 – 19.3%, 2007 – 16.7%, 2008 – 21.6%
Brad Wilkerson: 2006- 14.9%, 2007 – 19.4%, 2008 – 0.0%
I want to stress that the sample sizes for ’08 are really small and conclusions shouldn’t be drawn yet. If wind would have knocked down one of Vidro’s two homers, then his HR/FB% would be 4% instead of 8%. You just can’t make any kind of judgment about a power rebound when one fly ball with a different result changes the conclusion.
However, I think we can make some preliminary assessments based on the multi-year trends.
Thomas has clearly lost a good chunk of his power, but because he had such a high platform to fall from, he’s still able to be a decently productive hitter even after a pretty significant decline. Expecting him to turn back into the Frank Thomas of 2006 doesn’t look like a good bet, though.
Vidro’s collapse came in 2005, when he fell from a 12.8% HR/FB rate to the 8.0% you see listed above, and he’s been in slight decline from that rate ever since. His career has been on life support for a couple of years now, and I’d argue that he’s pretty obviously done.
Sexson’s power took a slight dip last year, but nothing even close to being large enough to constitute a collapse. Through this window, it seems pretty clear the he’s not done as a hitter, even though his old player skills suggest that he’s in the last stage of his career. There’s still life left in his bat, though.
Wilkerson is an odd case. His ’08 numbers aren’t particularly instructive because the sample is just so tiny (16 flyballs total), but there are extenuating circumstances with Wilkerson – he’s had old player skills for quite a while, and injuries have taken a toll on his body. If you were going to create a list of potential collapse guys before the season, he’d have been on it. But we’ve also seen his HR/FB rate nosedive and then rebound before (2004 – 16.2%, 2005 – 5.9%, 2006 – 14.9%), so drawing any conclusions based on the fact that he hasn’t hit a home run in the first few weeks of April seems foolish. He might be done, but we don’t have enough information to know for sure yet.
To summarize a post that turned out much longer than I expected, I believe we can usually tell if a player has really fallen off The Cliff by looking at his ability to drive the ball. HR/FB% stands in quite well as a proxy for his power, and significant changes in that can be a red flag that a hitter’s skill set has changed for the worse. It’s not definite, however, and you shouldn’t conclude that every drop in HR/FB rate is signs of imminent retirement. However, if a player is struggling to produce at the plate while posting a stable HR/FB rate, we could say it is unlikely that their career is really teetering on the edge of extinction.
Or, to put it back in player terms, Thomas and Sexson are not done, Vidro is, and we don’t know about Wilkerson just yet.
Old, established player with cobwebs in his joints and a bat made of mulch. We still have some trade bait (Mike Morse) and could make a move.
That was kinda my first reaction, too.
Thanks for this analysis. It’s the kind of simple, straightforward rule of thumb thing that’s extrememly helpful, that I could never, ever come up with.
[gasp] I misspelled Vizquel.
#27 Isn’t there some self-selection going on in that data, though? Good pitchers are going to pitch more often with runners not on base, while bad pitchers will naturally be in more situation with runners on base. I don’t know if it’s enough to significantly affect the data, but it would be nice to control for it. Having said that, I’m certainly much too lazy and incompetent to do it.
#44 While I agree with your general point, the Giants have some hope, but only in the low minors. And I would pay money to watch Lincecum and Cain. (Well, in theory I guess, since I don’t actually do that.)
Dave,
While I think your attempt at finding a metric to identify cliff-ness is good, I think the actual stat you selected is terrible, for the very reason that you brought up when noting why 2008 must be discounted for Vidro.
HR/FB probably works well for POWER HITTERS. But, players who typically don’t break 20-HRs per season have *ROUTINE* swings in HR totals that are utterly unpredictable. The upper number is simply too small and variable to rely on the conclusions.
Someone else noted that it is “possible” for park effects to have major impact on HR totals, also.
This does not mean your analysis of Vidro is wrong (or right). But, I definitely think a look at isolated power would be more likely to avoid the inherent weaknesses I see in the HR/FB methodology. (Barry Larkin ranged from 7 to 33 HRs during his career, so I’m betting his HR/FB ratio was all over the place).
While I think it is worth exploring, I still have doubts based on your graph, which shows that Vidro’s BABIP has *NOT* suffered in conjunction with his HR/FB rate. If I’m starting with the theory that HR/FB is a ‘tell’ for BABIP, then I wouldn’t jump for joy if the first case I’m looking at to prove this shows zero correlation.
Don’t get me wrong. I think it is a valid arena to study. I just don’t think the data you’ve gathered so far is particularly indicative of causation.
=================
On the subject of how often you see bounces? My perception is “regularly”. I’ve had incredible success in fantasy leagues by betting on bounce backs after career bad years.
Mike Lowell would be a near perfect example of a guy who crashed for an entire season, (don’t know what his HR/FB was in 2005 – but with only 8-HRs, I’m thinking it couldn’t have been good), and two years later is 5th in MVP voting.
Setting aside whether the actual argument is valid or not, which I’ll let Dave address if he wants…
If you want to disagree with something posted here, folks, that’s the standard for how you do it. Well thought out, with specific reasons for the disagreement and examples to support those reasons.
Good post, firemane.
55 – Nice. In case you’re wondering, Mike Lowell’s 5 year ‘trend’ from 2003-07 goes like this:
17.1
10.6
3.8
9.5
8.6
This ratio seems remarkably variable for most players.
The one thing I’d quibble with is the idea that Dave’s post started with the idea that a bad HR/FB presages or is correlated with bad BABIP. Both metrics are variable, and the latter has much more to do with LD% than if FBs are creeping over the wall or not.
I think we’re in agreement on the broad point though – who knows though, maybe further investigation will produce some parameters in which HR/FB really is a great diagnostic tool. We pretty clearly haven’t found such parameters yet, however.