THE BOOK--Playing The Percentages In Baseball

Here’s my model:

http://www.baseballprospectus.com/article.php?articleid=11839

Or at least, the start of it. Additional starter innings per game displace below-average relief performances before they displace above-average relief performances.

I still would like to see the final impact numbers using your model. 

Given a league average of 4.70 runs per game, a replacement starter giving you 6.00 runs per game, a mopup relief guy giving you 5.00 runs per game, what happens when:
1. You lose your 238 innings of 3.80 runs per game starter
2. The rest of your bullpen has to work harder to compensate

I’m suggesting that for #2, the extra runs given up by overworking your relievers will be offset by not letting your replacement-level starter pitch that third time through the order; that it works out to the same thing.  And if not the same thing, not enough to worry about.

That’s what’s on the table here.  I’d like to see any kind of working model at this point.

How are you splitting up your innings between starters are relievers? Is a replacement-level starter 6 RPG in a league average number of IP per game, or is he pitching fewer innings?

Colin/3:

I think I addressed that already:

His 35 starts are cobbled together at the replacement level of 6.00 runs per game, with only 4.8 innings per start.
...
You still need 70 innings to find somewhere in relief. 

Or are you asking something else?

We have little way of knowing at this time at least, but my WAG would be that a .20 bump for the entire bullpen is generous (I think it might be less, if anything at all).

IIRC, in The Book we showed some evidence that perhaps relievers don’t pitch enough.

And certainly a more optimal use of the bullpen could mitigate the effects of having to pitch more innings.  For example, when you have a low leverage situation, you pitch some replacement level reliever for several innings.  When he gets burned out after a few weeks, you go get another one from your farm clubs. Managers spend way too many resources trying to come back from 5-run deficits and trying to protect 5-run leads…

MGL, sure it could be no change at all, or increase of 0.50 runs.

What I think is important is to create a framework for discussion, so that’s what I’m offering as an idea here.  Lots of these discussion talk in such theoretical viewpoints that we have no idea what the impact of what they are saying.

Once you quantify and create a framework, we can come to a consensus in no time at all.  I think that’s what the WAR framework gave us.

Sure, I like your framework.  One of the keys to the final “answer” of course, is how extra use affects the pen.  I think that is a very open-ended question.  One of the areas of baseball we know little about…

Agreed, a potentially fascinating discussion.

I think bullpens could pitch more without loss of effectiveness if the manager could plan for a regular usage pattern.  But that doesn’t work because baseball is too streaky. 

You’ll have a few stretches with a lot of extra inning games or starters getting knocked out early.  You’ll have situations where your best relievers are just not available because they’ve pitched a lot recently.  And then you’ll have stretches where you hardly even need a bullpen, and bring your ace reliever in to finish a 6-1 game just to get him some work.

In the illustration I’ve offered here, I increased the workload for all relievers by about 15%, and increased their runs allowed by about 5%.

Whether that is a reasonable or unreasonable position, I’m not sure. 

Rally makes a good point that the more reliant you are on relievers, the more juggling you have to do, then the less a fixed pattern you can follow.  You are putting pressure on the relievers.

On the other hand, you can really emphasize the relievers more by giving them even more work, fixing the pattern by inning, in order to control workload, and thereby possibly improving efficiency.

I don’t know.

As I said before, let your worst reliever pitch 3 or 4 innings in low leverage situations.  And let your good starter keep pitching even if he has allowed 7 runs (which we see more and more of these days, as Tango has mentioned many times) as long as his pitch count is not too high.  That would solve a lot of problems.

Tango - I think you have laid out the elements for the discussion.  That said, my instinct is that the reliever ERA would be much higher.  Also, I dont think that it is fair to assume that managers will use their bullpens properly for the example (which I dont think you did, but some have suggested).  Managers think they are using their bullpens properly right now but I think most of us agree that is not the case.

As for the main gripe I have with the reliver ERA is that (in the case of Sabathia and the Yankees) those innings are going to be covered by Mitre, Gaudin, etc.  That 4.35 ERA assumes that Mariano gets a propotional number of those innings when in fact he will get virtually none of them.

The nature of the Yankee bullpen (Mariano is really the only good reliver) makes the extra innings that Sabathia pitching EVEN MORE VALUABLE.

Socrates: I have no qualms there.  Just quantify it, since the model is there.

My day job and my brain power (or lack of brain power) prevent me from really being able to do what you guys do. 

Although maybe I could do this one… I guess the data needed is to look at Yankee reliever ERA in innings pitched in from the 4.8 inning mark to the 6.8 mark.  I think that ERA should be the benchmark for what will happen with the replacement starter.

Does that data exist somewhere or would I need to recreate it from the game log at baseballreference.com.  Happy to do my end of the work, but I am really a novice at this and realize that I am probably not using the best resources. 

Direct me brother.  I am here to learn.

I like your idea.  I’ll try it when I get home.