Monday, May 12, 2008

The Flaws in SLG and its Relatives

Slugging Percentage is a flawed stat because it divides TB by AB's. What's wrong with this? Most MLB players have more AB's than TB's but what if they didn't? What if player X got called up to the majors and in his first 11 PA's had 10 walks, 1 homer, 10 steals and 10 caught stealings, this would mean player X's SLG would be 24/1=24 yes he would average 24 total bases per at-bat. How is this possible? It's not, that's why SLG is flawed. TB(total bases) includes walks which AB's don't, only PA's include walks. Since SLG is flawed ISO is also flawed because ISO is just SLG-AVG. Also, what's the point of OPS? It just adds OBP and SLG, that's like adding 2B and RBI's and calling it a stat. It has no meaning whatsoever.

My proposed ways to fix these problems are:
a) To change SLG to (TB-CS)/PA this way you can't have someone averaging more than 4 total bases per PA/AB
b) To get rid of OPS or to equalize OBP and SLG and then average them(or something like that)
c) To change ISO's formula from SLG-AVG to [(TB-BB-HBP-SB)/AB]-AVG=ISO+(my invention)

Baseball's old stats have so many flaws it's ridiculous. I'm just pointing out one but there are many more like OBP(arguable), RBI's, Defensive Stats, Saves, Holds and ERA. Baseball Prospectus is putting together many stats that solve these problems like EqA(equivalent average, accounts for all offensive facets) for OBP, EqR(equivalent runs) for RBI's, FRAR(fielding runs above replacement) for D stats, NRA(normalized run average) for ERA and Leverage(measures game importance based on the level it was at when the game started) for bullpen appearances. I think everyone should put all these phony stats behind us and start using these new clear stats.


Moish said...

Basically, you're right. Here's what I think the right stats are and why:

The expendable currency of baseball is not at-bats but outs. In other words, a team does not get a fixed number of at-bats but rather a fixed number of outs. So production needs to be measured in terms of outs made by a hitter (or gotten by a pitcher).

Once that is understood, there are two basic characteristics that a good "universal" stat must have:

1. Its ingredients should be the kind of stuff you keep around the house (such as in your daily newspaper).

2. It should correlate with the number of runs a hitter is worth per game. Roughly speaking, that means it should approximate the number of runs a team would score if it consisted of nine copies of this guy.

Here is about the closest your going to come to a stat satisfying those criteria:

NOPS = (SLG/1-BA) + (OBP/1-OBP)

Note that SLG + OBP is your basic OPS. The difference here is the denominators which normalize by outs made. (NOPS stands for normalized OPS.)

Here are a few key facts about NOPS:

1. The major league average over the past 30 years has been consistently extremely close to 1.0. So the scale is perfect: an average ballplayer is at 1, and a good/bad ballplayer is above/below 1.

2. NOPS correlates really well with runs scored. Over a season, the number of runs a team scores per game is very very close to 6.7 x NOPS - 2.4.

An even simpler measure that has a great correlation with team runs is OPS* = OPS/(1-OBP) where runs = 6 x OPS* - 2.

Obviously, then, we can use the same formula to evaluate individual players. Clearly this fails at the extreme since, for example, a hitter batting .000 will be worth -2 runs, which makes no sense. But it's really not bad in the typical range.

On all this, see

Ari Berkowitz said...

Thank you for this, I new that something like this would make sense. Still, if a hitter has a 1.5 NOPS what does it mean? The stats with the same format as yours based on out are RC/27 and MLV which I think are both terrific stats. Why do you have to multiply it by 6.7 and subtract 2.4 I don't understand? Is this your own work?

moish said...

Yes, those are my stats. The advantage is that you don't need any hard-to-get data to compute them. The 6x-2 is to convert it to runs scored per game. For example an OPS* of 1.1 is equal to 6 x 1.1 - 2 = 4.6 runs per game.

Anonymous said...

Your example is wrong. TB does not include walks, nor steals. It's just bases gained on hits (single=1, double=2, triple=3, HR=4).