MLB

# Using Power Factor to Predict Future Performance in Fantasy Baseball

By dividing total bases by hits, we can get a better look at which hitters are the most reliant on raw power.

Think about how we measure power in baseball.

You could use slugging percentage, which had been the default for years. You’d run into trouble, though, with slap hitters who get a ton of singles.

Take two players who both have 200 at-bats. Player A gets 80 singles and no extra-base hits, and the other hits 10 home runs and 20 doubles. Player B has hit for more power, but both players would have identical .400 slugging percentages.

You’ve surely seen this example before, and know this is the part where I bring up isolated power rate (extra bases per plate appearance), which does a much better job at measuring true power hitting.

In terms of ISO (which can also be measured by subtracting batting average from slugging percentage), the first player would have a rate of .000, while the second player would have a .250 ISO.

ISO does a great job at measuring power hitting and  takes a smaller sample to become stable than slugging percentage.

But, we can actually go one step further, by isolating power itself from the act of hitting.

This is where “Power Factor” or total bases per hit comes in, which does an even better job at measuring raw power, as it does so on a per hit rather than per at-bat level.

This means that a player who makes out frequently, but hits the ball very far, can post a very high Power Factor (though if he is making out often enough, he might not actually be a very good hitter; PF is more descriptive than prescriptive).

In a good post on this topic at Beyond the Box Score from 2011, Lewie Pollis notes that while “it's impossible to fully isolate power from contact skill and luck...their impact is much smaller on Power Factor than on ISO or SLG.”

Consider the second player from the sample above again, the guy who hit 10 homers and 20 doubles in 200 at-bats.

Now let’s say Player C hits 10 home runs, 20 doubles, and 40 singles in 200 at-bats (for the sake of simplicity, let’s just say neither player drew a walk somehow).

Clearly, Player C is the superior hitter, thanks to a .350/.350/.600 slash line, but both players would have identical ISOs.

Both players are different styles of hitter, though, with Player B taking the part of power-reliant slugger, and Player C as the all-around hitter.

A stat like Power Factor illustrates this, as Player B has a PF of 2.7, while Player C would have a 1.7.

Essentially, this stat is trying to tell us when a player gets his next hit, whenever that may be, this is how many bases we can expect him to get -- this is different from ISO.

In a 2009 Walk Like A Sabermetrician post, the distinction between the metrics is spelled out well: PF measures a player’s style of hitting with regards to raw power, while ISO measures “power contribution.”

Though Player B and his PF of 2.7 is more likely to get an extra base hit than Player C when he does get a hit, both players had an identical extra base hit rate -- their power contribution was equal, and their equivalent ISOs show this.

It should also be noted the distinction between ISO and PF is often academic in practice -- a player who posts a high ISO will almost surely have a high rank in terms of PF as well (last season, the two stats correlated with each other with a coefficient of 0.76, with 1 implying a perfect relationship and 0 implying no relationship).

With that said, here are the leaders in PF last season, followed by a look at how we might want to use this information.

PF rank Name PF ISO ISO Rank PF rank - ISO rank
1 Chris Davis 2.145038168 0.3 2 -1
2 Jose Bautista 2.144 0.285 5 -3
3 David Ortiz 2.025641026 0.28 6 -3
4 Edwin Encarnacion 2.010830325 0.28 7 -3
5 Nolan Arenado 2.003484321 0.287 4 1
6 Carlos Gonzalez 1.992619926 0.269 9 -3
7 Lucas Duda 1.991803279 0.242 14 -7
8 Joc Pederson 1.985714286 0.206 31 -23
9 Mike Trout 1.973244147 0.29 3 6
10 Albert Pujols 1.967213115 0.236 16 -6
11 Bryce Harper 1.966666667 0.319 1 10
12 Todd Frazier 1.952941176 0.242 15 -3
13 Alex Rodriguez 1.944 0.235 17 -4
14 Ryan Howard 1.934497817 0.214 24 -10
15 Jay Bruce 1.920353982 0.209 29 -14
16 Josh Donaldson 1.912457912 0.271 8 8
17 Russell Martin 1.908333333 0.218 21 -4
18 J.D. Martinez 1.897163121 0.253 11 7
19 Brian McCann 1.88362069 0.204 36 -17
20 Evan Gattis 1.882113821 0.217 22 -2

The league average Power Factor was 1.59 last season, with a standard deviation of 0.21.

Again, this stat is a descriptive tool, rather than one that speaks to a player’s value. It can tell us, for example, that despite similar ISOs, Paul Goldschmidt is a very different hitter than Lucas Duda. It can also show that players with differing skills, say Duda and Carlos Gonzalez, are more similar than we might expect in terms of raw power. Gonzalez gets more hits, but on these hits, he averages about the same number of total bases as Duda.

### Fantasy Applications?

While Power Factor doesn't tell us who the best power hitters are, there are a few other ways we can use this data.

As Pollis noted in a separate post, Power Factor “gives us insight about whether a slumping hitter is maintaining his power, or if his lack of pop is the cause of his offensive struggles.”

The biggest takeaway here, though, especially for fantasy owners, might be Power Factor’s use in predicting future performance.

In 2015, 101 players had at least 250 plate appearances during both the first half and second half of the season. Here's how the following stats in the first half correlated with themselves in the second.

Stat Correlation Coefficient (R)
Strikeout Rate 0.83
Power Factor 0.74
Walk Rate 0.73
Home Run Rate 0.69
ISO 0.62
OPS 0.39
Slugging Percentage 0.38
OBP 0.38
Batting Average 0.18
BABIP 0.12

Well look at that. Power Factor ranks right up there with the famously reliable walk and strikeout rates in terms of consistency.

On its own, this might not seem surprising or particularly relevant...until we look at how Power Factor in the first half correlated with various other power stats in the second.

Stat 2H ISO 2H HR% 2H SLG%
1H PF 0.685201185 0.699688113 0.517768304
1H ISO 0.61681539 0.631129045 0.479525496
1H HR% 0.622349583 0.649442041 0.47872798
1H SLG% 0.456787298 0.473167015 0.379612636

Now we’re talking. In terms of predicting future ISO, home run rate and slugging percentage, Power Factor actually performed better than each stat on its own.

If last season was any indication, this is actually something we could have put to use. Consider Jose Bautista, Chris Davis, Edwin Encarnacion and David Ortiz, who ranked 4th, 6th, 10th and 11th in first-half Power Factor, respectively. This same group ranked 9th, 15th, 17th and 23rd in ISO during the first half.

In the second half, Davis led the league in ISO, Ortiz was second, Encarnacion was third, and Bautista was fifth.

This group of four happens to include players who already established their power-hitting chops. However, it might still be useful as a tool to see if a player has changed his approach at the plate, from a contact-oriented style to one based on power. In cases such as this, Power Factor might be among the first stats to pick up on this.

Is this the be-all, end-all of predictive offensive stats? Certainly not, but its high self-correlation and relationship with future power metrics seems to make it something worth looking at..