Friday, July 08, 2005

 

The PoP Question - Again

So.
It's been almost 24 hours since I asked for input on the positive outing percentage cut-off point. About forty "unique visitors" have at least had the opportunity to consider the question during that time. That guy over there on the right is the only one to have put his two cents in, which were... indecisive, shall we say?
(By the way, have you ever met anyone who is psychologically unable to say their own name? Well if you've met Jere, you have. Dude can't even say "cherry" because it sounds too similar. I'm serious. I was talking to him last night, and he inserted my cat's name for his own while reading me a quote.)
So, back to the PoP. It's most likely that no one cares about this stat, and therefore no one cares what would be considered "positive" as they are unlikely to look at it anyway. Best case scenario, there are people who DO care, but consider me to be all-wise and better qualified to make the call. I don't think that's the case, but since no one's chiming in I guess I'd better just figure it out myself.
When I made up this stat, I had no standard to go by, obviously. I just had to relate the numbers to how I felt about the pitching in a particular game to understand the kind of quality that number indicated. My original reaction was that all the really good games were under .500. As I began charting, however, that area between .5 and .6 began to look more and more solid. In considering this question, I have tried to think about it in terms of the simple ratio involved. A pitcher giving up three bases in the span of five batters (.600) is sometimes scored upon, but not most of the time. But the tables turn with the next simple ratio: two bases awarded for every three batters (.667) is quite a bit more marginal as a performance. I think scoring happens a lot more with this ratio. So I'm going to say that anything .600 and below gives the team a very solid chance to win. As I mentioned in the comments yesterday, this would change Miller's PoP from 55 to 81%, and Clement's from 61 to 89%. I would have to say that those numbers more accurately describe my gut feeling on the matter. (By the way, this would move Foulkie to 83% last year and 68% this year, and 68% is still pretty unacceptable. A "closer" is not someone who pitches marginally to horribly, risking the team's opportunity to win, nearly ONE THIRD of the time.)
Also, I realize that there's some stat out there relating to quality starts. Of course that stat is not likely to satisfy me as it's criteria is unlikely to be all that similar to BpB. Besides, this is to measure quality outings, not just starts, which in my opinion makes it immensely more useful. Particularly for Red Sox Nation, as we seem to have some... um, ISSUES with our pen at the moment, eh?
This is officially your last chance to express an opinion on this before I do a lot more calculations. All the raw data is available in the spreadsheets linked on the left, so consider it now or forever hold your peace.


# posted by Rebecca @ 9:04 AM
Comments:
I think your assessment seems right on. Granted, I have only really skimmed what you've been working on so far, but it seems very logical.

I think a metric for quality outings as opposed to starts is invaluable, and as such your PoP stat is a good way to evaluate relievers.

Very nice work.

-MRhé
# posted by Anonymous Anonymous : 10:28 AM
 
Yeah, I haven't been getting many comments lately either. How come Finy writes a paragraph about not having written in a while, and gets ten comments on it? And then Surviving Grady gets 80 a day, of course. Spread the wealth everybody.
# posted by Blogger Jere : 11:21 AM
 
He said your cats name??
# posted by Blogger Sam : 3:14 PM
 
Rather than a strict cutoff, perhaps you could have a graded transition, where you output a final "percentage" that is the sum of each outing's rating divided by total outings. That's probably unclear, so here's a quick example--pitcher X has three outings with BpB's of 0.4, 0.55, and 0.8. 0.4 is great, totally successful, so that gets a 1, and 0.8 sucks so that's a 0. 0.55 is intermediate, so that gets a 0.5. Add it up, that's 1.5, divided by three outings gives 0.5 or 50%.

The trick here is how to calibrate the intermediates--is it between 0.4 and 0.6, or some other range, and what shape does it have? You could make it a linear ramp, or sigmoidal, or (I like this the best but it's the most work) you could figure out what the win probability is as a function of starting (or relieving or closing as the case may be) pitcher's BpB and set up the conversion accordingly.
# posted by Anonymous Anonymous : 2:41 AM
 
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?