Peripheral Vision or, Passing my physical

[Wally Mo Pena]

Well, it’s been a wild off-season for the S3F, hasn’t it? Threats to move, changes at the top, feelings hurt, bridges mended, and anticipation for the new season. Why, it’s just like (insert team name), now that I think about it. It all leaves me feeling like the first article of the year is my version of reporting for spring training - just before the march 1st deadline no less - and what better way to kick it off than with pitching drills?

Let’s start with an easy one: Peripherals. We all know the tale. You’re arguing with a friend about a pitcher, you cite his dominant ERA as the final word on his greatness, and your friend - laughing haughtily - tells you that his peripherals are those of a dead moose and he is certainly destined for humiliation and worse. What are you to say? ERA is so simple it has to be right, it’s just runs divided by innings, what could possibly… ohhhhh.

Then it hits you. All those runners left on base, circus catches in the outfield, and miracle double plays may catch up with your phenom one of these days. That’s when you turn to The Peripherals: those rate stats (i.e h/9, bb/9, whip, k/9, and hr/9) that seek to predict how well a pitcher will do without looking at runs.

But lets be fair. The great pitchers ARE good at getting out of jams, and sometimes ball 4 is really strike 3 and inducing the DP ball to end the inning was a work of art. ERA matters in the end because that’s what’s on the board as they say. So the question becomes, what peripherals best correlate with ERA in a large sample? Over the long haul, if a peripheral doesn’t correlate with ERA it should be moved into that area of the periphery known as the trash.

My method for answering this question was simple, relatively. 1) determine the relationship between ERA and each peripheral noted above, by dividing ERA by the other stat (how much did ERA change when K/9 changed for example), for each pitcher in the sample. 2) determine which of these relationships changed the least from pitcher to pitcher. The one that changed the least would be the best predictor of ERA and therefore the best weapon against you’re know-it-all friend. [The math version of this is lowest standard deviation as a percentage of average, if you want to talk about methods later. I used as my sample every NL pitcher in the 2005 season, for a total n = 357.]

What I found was surprising in some ways and not so much in others. Here are the peripherals ranked by how well they predict ERA (deviation %), with the average ERA predictor for the league:

1) WHIP, 3.27 points of ERA/point of WHIP, 40.2% deviation
2) Hits/9, .523 points of ERA/hit per 9, 46.6% deviation
3) BB/9, 1.35 points of ERA/BB per 9, 62.6% deviation
4) HR/9, 4.16 points of ERA/HR per 9, 63.7 % deviation
5) K/9, 92.3 % deviation (k’s don’t work the same, but the deviation is still interesting)

So what does all that tell us? Well, the averages there can give you a general guideline for where someone may move toward if their ERA seems high or low based upon a peripheral. For example, if someone gives up 4 BB/9 and has an ERA of 2.70, you might guess that their ERA will work it’s way back to (4 * 1.35) 5.40 over time. That would be returning to the average relationship between BB/9 and ERA. But then you take into account the deviation percentage (which can be stated: except in extreme cases, people will move to a range of the average +/- that percentage of the average). So, applying that range, you could expect to see an ERA between 2.01 and 8.78 in this case. That shows that the 2.70 ERA actually isn’t that ridiculous for some, but it's getting pretty far to one end of the range so you could expect to see some increase toward the middle.

WHIP has the strongest correlation with ERA, but hits/9 is close (I was surprised how close). K/9 has the worst relationship with a variation of almost +/- the whole average. So it’s possible to be very successful with few k’s and terrible with many.

Whew. I’m ready to hit the showers and drive a golf cart to beach, and i'm sure you are too if you finished this. As always, comment or PM any questions or requests for the raw data.

[redsrbetter]

I really like this. Its a way to measure a pitcher and his luck far better than just trying to say that you saw every game he pitched and realized that he gave up far too many hits. "He was lucky to have a 2.71 ERA because he gave up 10 hits/9 IP." We all know that it was some form of luck that gave that up. But this equation is a nice way to figure out just what that luck was.

Very well done Wally once again!

[Fishercat]

That's pretty damn interesting. However, in your opinion, would this be a better judge for future performance of players who have hit their peak or prospects? For instance, if you saw a 1.1 WHIP on two different players,
would it be more indicative of how the 33 year-old veteran will do or the 22 year old prospect (if they stayed on the same level)?

Of course, if one is determining the worth of a player, he should focus on more than one stat? I mean, WHIP may be the best determinant, but looking at H/9, or HR/9, or TB/9 would also help in determining.

Just out of curiosity, how would some stats that deal with total bases (to show the difference between that 10 hits/9 innings for Player A and Player B (Like if one is prone to doubles and one only gives up singles), how close is that?

Good job though. Very nice work.

[love_that_reefer]

Determining "luck? I know you are trying but I can't buy into any of this stuff. Guys who give up hits but get out of jams aren't lucky, sometimes they get the ground balls they need and simply let the defense work for them. Too many stats. The best measurement is to see it with your own eyes.