An Expansion on xISO, Plus 10 Noteworthy Names

Last week, I introduced xISO, a metric that calculates a player’s expected isolated power based on his batted ball profile (per FanGraphs’ recently added batted ball data courtesy of Baseball Info Solutions). Having looked at a handful of underachieving National League outfielders for its induction, I’ll expand the analysis of xISO here today.

I’ll reiterate some key points. I used all 12 years’ worth of batted ball data for all player-seasons in which a hitter qualified for the batting title. The OLS regression specified pull rate (Pull%), hard-hit rate (Hard%) and fly ball rate (FB%) as explanatory variables and produced the following equation, which I deliberately omitted last week:

xISO = –.1396 + .1814*Pull% + .5136*Hard% + .2344*FB%
Adjusted R-squared: .631

In a perfect world, the explanatory variables would not overlap; or, if they did — and they do — then I would be able to tease out the amount of overlap. However, FanGraphs’ batted ball leaderboard doesn’t dissect the data to that extent*, thus contributing to a degree of multicollinearity (correlation among explanatory variables). The multicollinearity is small, however, and I have tried to quell the principal concerns of overfitting the model or producing biased estimates by keeping the model simple.

*A diligent and curious mind, however, could manipulate the leaderboards using the “splits” dropdown and append multiple data files to manufacture the ideal data set, if he or she so desired.

Moreover, the sabermetric community greatly concerns itself with how quickly statistics stabilize. Unfortunately, I don’t know yet when, by number of plate appearances, xISO becomes reliable. Nor do I know when pull or fly ball rates stabilize, either. Sean Dolinar informed me that hard-hit rate becomes reasonably reliable when a hitter accumulates as few as 70 balls in play, but that denominator does not necessarily hold for the other two rates.

Therefore, it is not yet safe to make inferences based on the following graph alone, which I have generated for the sake of demonstration. Below, I depicted ISO and xISO moving simultaneously for one player-season. I somewhat arbitrarily selected Adam Jones‘ 2014 season, recalling he started off last year rather slowly.

It’s a pleasant coincidence that, by game 159, Jones’ ISO and xISO are almost identical; the end-of-season xISO estimates for everyone else won’t always be so accurate. Regardless, it appears that xISO stabilizes a lot faster than ISO does. (Again, it cannot be stressed enough that, in the instance of Jones, n equals one). Thus, there may be some merit to xISO stabilizing more quickly than ISO. The former, by attributing a certain amount of weight to each input, dampens the effects of sample-size noise in each one. It is also computed based on expected, rather than actual, outcomes; the latter is at the mercy of probabilities.

Please find below a table of the ISOs, xISOs and batted ball information of all 2015 qualified hitters. Underachievers and their varying degrees of underachievement are highlighted in shades of red; overachievers, in blue.

Statistics exclude May 12 games.

Four overachievers, of whom some may be more obvious than others:

Eric Hosmer, KCR 1B

2015	FB%	Pull%	Hard%	ISO	xISO	diff
Eric Hosmer	27.0%	34.0%	27.2%	.222	.125	+.097

Is this the Hosmer we’ve been waiting for? (Better question: How many times have we asked ourselves that?) The 25-homer pace is pretty, but Hosmer’s .125 xISO almost perfectly aligns with his 2012 and 2014 ISOs (both .127), confirming he’s probably the same, frequently disappointing hitter he has always been.

Stephen Vogt, OAK C

2015	FB%	Pull%	Hard%	ISO	xISO	diff
Stephen Vogt	39.8%	49.4%	36.1%	.316	.229	+.087

Sure, his hot start is unsustainable, but that’s not what’s peculiar: Vogt’s .229 xISO slots in just below Nelson Cruz, Miguel Cabrera and Justin Upton and just above Jose Bautista. So, again: his hot start is unsustainable. But just because it is unsustainable doesn’t make it illegitimate.

Jake Marisnick, HOU OF

2015	FB%	Pull%	Hard%	ISO	xISO	diff
Jake Marisnick	32.4%	35.5%	25.0%	.198	.129	+.069

This is a timely development, as David Wiers, in his May tiered rankings of American League outfielders, expressed pessimism regarding Marisnick while I, for reasons not worth trying to articulate, ranked him 1) optimistically, and 2) as a National League outfielder. Anyway, xISO sides with David, and I’m inclined to side with both of them now. Still, the steals are nice. As a slight-power, moderate-speed threat, Marisnick seems like the player, both offensively and defensively, we once wished Adam Eaton would be.

Joey Votto, CIN 1B

2015	FB%	Pull%	Hard%	ISO	xISO	diff
Joey Votto	26.1%	45.5%	33.0%	.235	.174	+.061

Votto’s 2013 and 2014 xISOs clock in at .185 and .168, respectively, so I have few reasons right now to deny an impending decline from the aging and declining slugger. It’s reasonable to expect the return of 2013 Joey Votto — it’s closer to what his current xISO anticipates than his dreadful 2014 season — but he currently looks like 2011 Joey Votto, and given that RBI are tied to isolated power, those are pretty different Joey Vottos.

Three underachievers, of whom some may be more obvious than others:

Victor Martinez, DET DH

2015	FB%	Pull%	Hard%	ISO	xISO	diff
Victor Martinez	37.4%	38.5%	28.6%	.061	.165	-.104

V-Mart’s struggles are well-chronicled, and it’s not the first time he has been slow out the gate following an injury. If one harkens back to 2013, one may recall an incredibly slow start before ravaging the league for the latter half of the year. The .165 xISO is a far cry from last year’s .230 ISO, but it points to a more promising rest of season.

Danny Santana, MIN SS

2015	FB%	Pull%	Hard%	ISO	xISO	diff
Danny Santana	26.0%	38.1%	34.5%	.071	.168	-.097

The plate discipline is horrid, but xISO still thinks Santana is due for more power given how hard he has hit the ball this season. His batting average and lack of power would be tolerable were he running, but he’s not, so they’re not. An interesting buy-low in deeper leagues, perhaps.

Troy Tulowitzki, COL SS

2015	FB%	Pull%	Hard%	ISO	xISO	diff
Troy Tulowitzki	36.7%	54.4%	45.6%	.188	.279	-.091

Tulo is straight-up hitting the snot out of the ball when he makes contact, sporting an xISO that mirrors last year’s .263 ISO amid a half-season in which he simply punished opposing pitchers. A straightforward metric such as HR/FB would scream regression right now, so this one isn’t so bold. Also, there was a pretty strong caveat in when he makes contact: he’s striking out at a career-worst rate and walking at an astonishingly bad 1.9-percent clip. Sorry, OBP!

Four dudes who are performing exactly how one might expect them to perform, so that’s interesting, man:

Carlos Gonzalez, COL OF

2015	FB%	Pull%	Hard%	ISO	xISO	diff
Carlos Gonzalez	27.1%	31.4%	27.1%	.120	.120	+.000

It’s just ugly at this point. It’s easy to convince oneself to wait for CarGo, a perennial all-star, to come around and bust his funk. xISO, quite definitively, denounces such patience as unwise. I encourage you to get as big a return as you can for him, and to do it sooner rather than later.

Chris Davis, BAL 1B

2015	FB%	Pull%	Hard%	ISO	xISO	diff
Chris Davis	40.0%	60.0%	40.0%	.271	.268	+.003

Crush is back. (Mostly.) The xISO splits the difference between his ridiculous 2013 and miserable 2014 — right where most of us probably thought he’d be.

Mookie Betts, BOS OF

2015	FB%	Pull%	Hard%	ISO	xISO	diff
Mookie Betts	45.5%	40.7%	31.0%	.194	.200	-.006

The kid’s a monster. He has already outdone himself, and it all seems legit.

Alex Rodriguez, NYY DH

2015	FB%	Pull%	Hard%	ISO	xISO	diff
Alex Rodriguez	42.3%	43.6%	47.4%	.308	.282	+.026

I’ll let you make sense of this on your own. We all cope in different ways.

Dare I ask: could A-Rod actually produce enough value this year to justify his salary? (Not his entire salary, obviously — just this year’s.)