Ip + Era+

[Blackout]

does anyone think this is a decent way to evaluate a pitchers season if they pitch say 200 innings?

maddux:
91: 356
92: 434
93: 438
94: 473
95: 471.7
96: 407

93 and 94 were clearly his best seasons.

does anyone think i'm on to something here??

[Tango Tiger]

In one word: no.

In more words:

Let's say the league ERA is 4.5, which means 0.5 earned runs per inning, or 2 innings per run).

x = lgIP / lgER = 2

So, you are saying:
= IP + (IP/ER / x) * 100
= IP + (IP/ER / 2) * 100
= IP + 50*IP/ER

Other than "bigger is better" (more IP, less ER), what is it you are trying to say here?

If I try to baseline it with 180 IP, 60 ER, that gives me a score of "330". With 9 more innings, I need 7 more earned runs to get the same score. Seems somewhat reasonable.

But, let's say I start wiht 180 IP and 90 ER. That's a score of 280. If I bump the IP to 189, I need 14 more ER to get a 280 score. Adding 9 IP and 14 ER is not what I call anything reasonable.

[brett]

does anyone think this is a decent way to evaluate a pitchers season if they pitch say 200 innings?

maddux:
91: 356
92: 434
93: 438
94: 473
95: 471.7
96: 407

93 and 94 were clearly his best seasons.

does anyone think i'm on to something here??

If you like ERA+, I would take ERA+ (divided by 100) minus some baseline (probably .70 or .80) and multiply it by innings pitched.

So a guy with a 200 ERA+ for 200 innings would be 2.00 - .80 or 1.20 x 200

or 240.


Another guy with a 1640 ERA+ and 300 IP would be 1.640-.80 or .80 x 300 or 240

So a 160 ERA+ for 300 innings would match a 200 ERA+ for 200 innings.

[Tango Tiger]

Except you need to take the recipricol of ERA+ before doing that.

You need to do: (1.25 - 100/ERA+) * IP

That'll give you the runs above replacement-like thing you are looking for.

Yet another instance where ERA+ calculated improperly causing more confusion.

[brett]

Except you need to take the recipricol of ERA+ before doing that.

You need to do: (1.25 - 100/ERA+) * IP

That'll give you the runs above replacement-like thing you are looking for.

Yet another instance where ERA+ calculated improperly causing more confusion.

Yea, I figured.

And factor out defense...

And then I guess you need to use some pyth-like estimator of wins.

[Tango Tiger]

No, the recipricol thing is at least correct. You have fielding and starter/relief biases, but those are biases. Biases can be accounted for.

ERA+ is simply bad math. You can't multiply it the way you were doing (but you already knew that!).

[weskelton]

I was playing around with ERA+ and IP last week following Tango's original response. I viewed ERA+ as the expression of quality and IP as the quantity over which that quality is expressed. This lead me down the multiplicative path similar to what brett and Tango have provided. I realized that 100/ERA+ is actually the % of lg(ER/IP) a pitcher is saving/costing per IP. So I ended up with...

(1 - 100/ERA+) * lgERA/9 * IP

Note that lgERA/9 = lg(ER/IP).

This actually ends up being a really good estimator for Palmer's Pitching Runs. I suspect that it may actually be equivalent when you take out the park adjustments.

PR = lg(ER/IP) * IP - ER
or
PR = (lgERA/9 * IP) - ER

Here's Clemen's career numbers as an example...

IP = 4916 2/3; ER = 1707; ERA = 3.12
lgERA = 4.46; ERA+ = 143

PR = (4.46/9 * 4916.67) - 1707 = 729.5

estPR = (1- 100/143) * 4.46/9 * 4916.67 = 732.6

[Patriot]

(1 - 100/ERA+) * lgERA/9 * IP

= (1 - ERA/LgERA)*LgERA/9*IP

= LgERA/9*IP - ERA/9*IP
= LgERA/9*IP - ER

So yes, it is mathematically equivalent as you surmised.

[Bothrops Atrox]

It seems to me that with the metrics that combine IP and ERA components, whether it be Pitching Runs, or WSAB, or some variation of what we are doing here - the all-time leaders in each category tend to be pretty close down the line for all of them. Most of the more advanced metrics similar to the ones mentioned are not much different than taking RSAA and making an adjustment for team defense and/or league quality.