## Saber-Friendly Tip #2: Talkin’ About Power

*In case you missed the first article in this series — in which I talk about another way to look at BABIP — I’m trying to take a look at alternative ways to present sabermetric stats, in order to best represent them to an audience. *

When you stop and think about it, despite the numerous baseball statistics out there, there are only a few limited ways of talking about a batter’s power. While there are a multitude of options when talking about plate discipline — On-Base Percentage, walk rate, outside swing rate, etc. — there are only a handful of widely available stats to use for power: the old standby, Slugging Percentage; a player’s raw total of homeruns or extra base hits; or the sabermetric alternative, Isolated Power.

So when you want to talk strictly about how powerful a player has been, which stat do you use? There are pluses and minuses to each of these stats, but do any of them necessarily stand out from the others? I’d argue no.

*Slugging Percentage*

**The Good: **Everyone knows it. It’s simple, easy to understand, and we all grew up using it.

**The Bad:** Like mentioned above, the formula of Slugging Percentage is very simplistic; it’s a player’s total bases divided by at bats. But if we’re talking about power, why are we including singles in the calculation? And if we’re putting value onto extra base hits, is a homerun worth twice what a double is worth?

*Raw Homerun or Extra Base Hit Totals*

**The Good:** Using the raw totals don’t attempt to place any value on each of the different hits, like Slugging Percentage does. Also, baseball fans don’t need to be told that 30+ homeruns is very good.

**The Bad:** Homeruns is just one part of the picture; doubles and triples are also very important. And if you list a player’s total extra base hits, you then run into the problem that you’re considering doubles as important as homeruns. Who’s the more powerful batter: someone who hits 30 doubles and 0 homeruns, or someone that hits 25 homeruns and 5 doubles?

*Isolated Power (ISO)*

**The Good: **It’s also a very simple statistic: Slugging Percentage minus Batting Average. This corrects for one of the flaws of Slugging Percentage, since this subtraction removes singles from the equation and leaves just the extra bases. As a result, Isolated Power gives more value to hitters that accumulate lots of extra bases but don’t hit for a high batting average.

**The Bad: **It’s on a funky scale. It’s a three-decimal stat, so I expect it to be on the same sort of scale as OBP or AVG, but it’s not. Instead, an average ISO score is around .145 and power hitters normally crack .200. It took me a long time to feel comfortable enough with the scale to begin using it in my writing, and new readers could have a problem with it.

***

I’ve gone back and forth on this question. Do I use a familiar stat, like Slugging Percentage, or do I go with Isolate Power — a stat that’s slightly more rigorous, yet is on a confusing scale for new readers? Believe it or not, the two stats correlate at a very high rate (.90 so far this season), so in general, you’re not losing much in terms of accuracy if you choose to use Slugging Percentage instead of Isolated Power. ISO is still an important statistic to use, since it can show you if a player might be over- or under-rated due to Slugging Percentage, but in the majority of cases the difference between the two stats isn’t as large as you may think. Power isn’t a very subtle skill; it tends to shine through no matter what lens you look at it through.

If you’re looking for a more rigorous alternative to Slugging Percentage, though, I’ve recently started looking at a player’s percentage of extra base hits. Consider:

Both these percentages tell you slightly different things — one tells you how often a player gets an extra base hit when they step to the plate, while the other tells you how often a player rips the ball deep when they get a hit — but they can both be useful when trying to get a full picture of a player’s power. Instead of limiting ourselves to just the handful of stats out there right now, why shouldn’t we use percentages like this, much like we’d use walk rate as an alternative to OBP? Both these stats correlate a high amount with Isolated Power (.91 for XBH/AB and .83 for XBH/H), so you know that they’re telling you similar information, but just in a different form.

I probably sound like a broken record, considering I finished the BABIP article on a similar note, but which stat you use depends on your audience and what questions you want to address. So don’t limit yourself to only Isolated Power; Slugging Percentage shouldn’t be throw out in the wash, as it has its many positives going for it as well. And don’t be afraid to branch out into some of the different percentage stats; they take but the work of a minutes to calculate, they are easy for new readers to understand, and they add nuance to your analysis.

Print This Post

I like ISO for power. Also, I pay attention to HR distance, especially avg HR distance. Says nothing about ability to actually connect but I do think using both gives a decent idea about brute power. Although I do use SLG% for evaluating whether 2B/3B/SS have any actual pop.

Gotta be careful with that though because there’s a huge selection problem. You’re only looking at flyballs that clear the yard. I realize there’s no data on flyball distance, but if there were that’d be the thing to look at.

I’m big into fangraphs but I hate Iso as a stat. It’s not accurate.

Many Home runs are just over, while many doubles are off the wall. These tiny differences in skill reflect huge in ISO.

Bad babip on grounders doesn’t mean you are better, but it will increase your ISO.

Also a triple scores 50% better than a double, but the difference in the two is either luck or speed, neither of which belong in a stat called “isolated power”

I would rather break it up into two stats, %hits for extra bases, and % of extra bases for homers.

Logan Morrison is a good name to use here. He was whacking extra base hits right and left, but getting more wall bangers instead of homers.

So his ISO was middling, and didn’t say anything, his %hits for extra bases was high, and his %XBH for hr was low.

A non homer guy will have low %XBH for homers, but they will also have low %hits for extra bases.

Sleeper power guys will have high %hits for XBH and low %XBH for homers.

You have good points, but FWIW, I don’t believe BABIP on grounders affects ISO. Since it’s SLG – AVG, the singles disappear and end up being the same as outs. Or, put another way, it’s (2B + 2*3B + 3*HR) / AB.

The weird part to me is that triples count twice what doubles do.

Except that ground balls can be doubles too.

Grounders can be home runs too. http://mlb.mlb.com/video/play.jsp?content_id=10664571

Your point?

if you have bad babip on grounders, you will have a lower average. Slugging will be reduced, but if you run some math slugging is affected less. (grounders are mostly singles)

Why wouldn’t speed belong in a discussion about power? Power is about how many bases you can get, not how far you can hit it.

Ruth’s official ISO is .34766. It represents nothing because it is a result of a flawed formula. His true ISO is .33879. It tells us that of 3 extra bases possible in an extra base hit, his averge base in excess of of an initial base of all hits is 1.016 bases. Therefore, his average hit is the initial base plus 1.016 extra bases or 2.016 average bases. Didn’t Ruth get 5,793 bases from 2,873 hits for an average of 2.016 bases per hit?

So, his true ISO is .33879, reflecting his average bases per hit of 2.016, his batting average of .342, his true SA of .504 ,etc, ….

Agreed with all these comments so far. It is also a limitation of power statistics that they count a “soft” doubt the same as a ball ripped off the wall, for example. There are players that are fast enough to turn a lot of softly hit gappers into doubles, and doubles into triples, but that should not be taken into account when measuring power, as it is a product of speed.

I think this is right, but then what are we saying? If “power” isn’t your ability to hit for extra bases, is it your ability to advance runners that are already on base? —The idea that the speedster who legs out the double isn’t as good at knocking the run in from 2nd as the lumbering hitter who bangs a double off the wall?

And then the issue of ballpark should be considered, too. As a Yankee fan, many times I’ve seen Teixeira, batting lefty, bang a hard single off the wall in RF.

Rather than XBH/H, why not have a stat for XBH/batted ball? Wouldn’t that give you a good idea for how often a hitter really connects on a ball? Is there something like ISOBIP?

IsoP annoys me, if your goal is to look at raw power. It punishes high-strikeout, low average batters.

Player A has 4 hits: two singles, a double, and a HR.

Player B has 4 hits: two singles, a double, and a HR.

Player A did it in 20 AB’s, so his IsoP is .100.

Player B did it in 10 AB’s, so his IsoP is .200.

They got the same # of hits, and the same number of bases, so if we’re trying to “isolate” power, why are their Iso’s so different???

The solution: TB/H. Gets rid of the problem described above, and is easily-explainable as “bases per hit.”

That’s called “Power Factor.”

http://www.wahooblues.com/2011/03/15/a-better-way-to-measure-power.html/

For Yirhiyahu — First of all Iso is a flawed stat. In your example both hitters actually have an ISO of .333, meaning they averaged one base in excess of the initial bases of their hits, reflecting a true slugging average of .500.

WHEREAS with a flawed formula, you calculate an ISo of .200 and a SA of .400 for player A and an Iso of .400 and a SA of .800 for player B

It took me a long time to feel comfortable enough with the scale to begin using it in my writing, and new readers could have a problem with it.One feature I’ve always thought would be super useful on this site–granted I have no knowledge of web design and therefore no concept of how feasible it is–would be to automatically color-code sabermetric stats based on where they fall relative to league averages. For example, blue my be -2 SD or lower; green would be -1 to -2 SD; etc.

Maybe it would seem too cutesy or be too difficult to execute, but as someone who’s knowledgeable enough to grasp the concepts but ignorant enough to be confused by scale, I think this would be a great communicative tool.

What about investigating a new stat that is a linear-weights version of ISO in a similar way to how wOBA is an improved version of OPS? If the issue with total XBH is that doubles count the same as HRs, and the issue with SLG and ISO is that they don’t count in the right ratios — for ISO, a triple is actually twice as good as a double, which essentially rewards speed, not power — then why not optimize the coefficients? We could use the weights from the wOBA formula and disregard 1B, BB, HBP, or we could subtract the 1B weight from each type of hit. For the denominator, use plate appearances, at-bats, balls in play, or hits — try them all, see which one seems to give the most valuable information.

We can call it wISO.

It is an interesting idea, except that wOBA uses the coefficients to quantify the relative run value associated with each of the outcomes, and not necessarily the raw “power” associated to the outcome. Ex. triples are weighted more than doubles but triples are typically the result of speed rather than power (as pointed out above).

Save a stat that easily encapsulates both hit velocity and distance on some scale, we will probably be left with a bunch of slightly flawed power stats. Maybe wISO is a less flawed power stat, I’m not sure, but I do not think it would measure exactly what you were setting out to measure.

But again, is power “how far one hits the ball,” or is it “how good is the batter at getting extra bases?” If the answer is the latter, wISO would do the best job at evaluating how good a power hitter somebody is.

I second the issue that none of thee above stats are very good measures of “power” really. They’re measures of ability to get extra base hits. HR correlates to power. But doubles and especially triples correlate a lot with speed also. So by lumping HR with other XBP, you’re looking at a stat that varies with power AND speed.

A real power stat would probably need to look at something like HR and some rating of batted ball velocity in general, regardless of where they fall. Otherwise, you’re going to get big swings based on park effects (unless somebody out there thinks that A-Gonzalez got stronger by moving to Fenway?). That way you can also determine when a hard-hit double is close to a HR and when a weak hit ball was a double due to speed (or defensive incompetence).

But, in the context of baseball, when we discuss “power”, don’t we mean “the ability to get extra base hits” ? Cause, if you just want to measure actual power, why not have guys just do bench-presses?

I think, in the context of this discussion, a triple should represent more power than a double. Cause it’s a whole ‘nother base. I don’t care whether the source is speed or muscle.

I get that, to a certain degree. But there really isn’t any more power involved in hitting triples than there is in hitting doubles. The park didn’t get any larger for one than it was for the other, so no additional power was needed.

Triples are more a function of speed (and the luck of where the ball lands) than power.

Maybe the terms should be changed, and as somebody pointed out earlier ‘power’ should reflect HR and velocity and distance.

Doubles and triples should be in a separate metric designed to measure the ability to get XBH.

Why not Power Factor? More year-to-year consistency, and it accounts for the flaws in ISO that give advantages to good contact hitters (i.e., a .350 hitter with .550 SLG has just as much “raw power” as a .150 hitter with a .350 SLG).

http://www.wahooblues.com/2011/03/15/a-better-way-to-measure-power.html/

Lewie –

I like this stat a lot (as someone interested in extra bases, as opposed to power); in your article, you mention Adrian Gonzalez. Has it changed appreciably in Fenway? If so, can you add park factors?

Good stuff…

Adrian’s power numbers (whether TB/H, XBH/H, or HR/AB) have actually been down so far this season, compared to his 2007-2010 norm. He’s been making up for it with a lower K% and higher BABIP.

Who knew that Fenway would make him into a singles hitter.

I use XB/H since it is less dependent on BABIP. It is again on an unusual scale but I like the information I get better especially for minor league or college where BABIP often gets to different ranges than in the majors.

I read the first time about it in the four factor article series here on fangraphs last season. It can be quickly calculated by dividing ISO and AVG.

I have a stat to solve the problem: wISO

(1.24 * 2B) + (1.56 * 3B) + (1.95 HR) / AB

Assigns a value to each type of XBH and doesn’t factor in batting avg like ISO does (Bautista’s high ISO was partially a product of a low batting avg).

I recently came up with it and I think it does a good job of measuring power.

You like?

Depends on where you got the coefficients from.

Isn’t HR/AB fine?

The problem is that “power” in the form of extra bases can look a lot like speed, especially doubles and triples. Jose Reyes and Curtis Granderson both put up big slugging percentage seasons by hitting lots of doubles and triples, and not off the wall. Line drives to the gaps are double and triple’s heaven.

If I tell you Jose Reyes has an ISO of .187, how many home runs do you guess he has?

If I tell you Granderson has an ISO of .251 (2007), how many HR do you guess he has?

Measuring “power” by using ISO just seems like an unnecessary step, when HR/AB will do the job, with metrics everyone already knows, and in ways everyone is already familiar with.

Slugging includes total bases, including those that are achieved with speed. ISO just removes the singles aspect of it. That’s really the best power measurement we have?

We also have to recognize that when we fans think “power”, they think “home runs” … so when talking about power, why measure anything other than home runs only? We’re throwing in doubles and triples and acting like it’s an improvement?

You don’t need ISO to tell you how many HRs a guy hit. There’s a stat for that. It’s called Home Runs.

Who cares if a guys speed is a factor? Aren’t we trying to calculate who is advancing himself around the bases fastest? Isn’t that the “power” that we’re trying to represent?

I’ll stop with the questions.

I don’t know.

What the baseball definition of power?

If you ask baseball fans, who is the most powerful hitter are they going to name the guys with the most home runs or the most total bases?

To me, this is one area, where we probably don’t need a sabermetric stat.

Is there a stat that tracks warning track fly balls? I know many analysts credit doubles as an excellent way of predicting breakout homer campaigns, but wouldn’t someone who clubbed 15 warning track fly ball outs be a more accurate representation?

I prefer wOBAcon:

((1b*.9)+(2b*1.24)+(3b*1.56)+(HR*1.95))/(AB-SO)

As strictly a measure of power the linear weights do a good job of scaling the hits and since we’re only interested in balls in play the denominator also works nicely.

This becomes even more confounding when trying to evaluate minor leaguers playing in different leagues, in different ballparks, against different levels of competition. ISO or SLG is ok but maybe it’s not as good as XBH%, which is something I like a little better for minor leaguers than for major leaguers. Perhaps going w/ XBH/AB is even better but that doesn’t discount for good/bad luck caused by hitting balls right at people or finding holes.

There is power in a Single. Think of the fatties who hit a single off the wall. The power hitters who were able to get a single through the infield that a weaker player couldn’t (speed is power). It’s not the traditional power that people think of, but it counts.

Actually, the formula for slugging average is flawed. It reflects the average size of an at-bat in terms of bases negotiated. Slugging average should reflect the average size of a hit. Like all stats, a perfect SA should be 1.000, not 4.000.

For instance, instead of Ruth having a SA of .690 of a possible 4.000, it is actually .504 of a possible 1.000.

Ruth had 5,793 bases negotiated of 33,596 baases possible from 2,873 in 8,399 at bats and a batting averge of .34206453. Correlating the components of that batting record we can determine Ruth’s correct SA. 5,793 bases negotiate divided by 33,596 bases possible; then divided by the batting average of .34206453 equals .5040898

That is an expression of an average hit bieng slightly greater than a double. A SA of 1.000 reflects an average hit being a home run. The traditional .690 SA expresses a batter having run down the first base line 62 feet before he was tagged out

Not only is SA flawed but ant other stat incorporating SA into its formula is also flawed — especially Isopower and OPS.