General NERP assistance

Diamond Dope does a wonderful job with their tool to make it easy for people to see the various calculations like NERP and BR, without manually doing a lot of work.

My questions revolve around how I should think about these results, the context of the 50/50 probability that the card I am looking at will be in play on any given roll:

First, the NERP calculation is based on the assumption that 100% of the rolls are coming from the card that is being evaluated. This is logical and rational. But I am not sure how to conceptually apply these values to the situation that half the rolls will be coming from the universe of other cards in play.

For example Greg Maddux 1995 has a NERP against lefties of -5.36 at Turner Field. How would I think about applying that against a starting lineup of hitter each of whom have an average NERP of 20. Is it simply a matter of taking the average of -5.36 and 20, which is 7.32? That result seems much higher than I would expect the game play to work out, so my conclusion is that there must be another to way to think about it. Any help would be appreciated.

In fact, you need to ADD the values.

-5.36 + 20 = 14.64

Interpretation: Maddux will allow 14.64 runs per 216 PA

The reason you add the two values is because NERP is a relative number, and the NERP formula is set to estimate the value relative to a "zero-value player" (basically, a player whose value is set very low, roughly speaking a .150 hitter with little pop). It wouldn't change much if you take the mean (7.32 is half 14.64), but the interpretation (and the adaptation below) would not be possible.

This said, per 216 PA is not the real basis you would like to have. I'm sure you would prefer to know how many runs he will allow per innings, his ERA so to speak. Unfortunately, it gets complicated. It depends on the on-base environment---so of course, it will highly differ whether you play in a 200M league or a 60M league.

But here is a rule of thumb. In a typical 80M league, I assume a .350 on-base environment and I assume that an average pitcher allows 20 overall chances of on-base will (stadium hits and homeruns included). In that environment, a typical pitcher will register 140 outs per 216 PA. Let's say that Maddux has 10 overall chances of on-base, so that means he will register on average 150 outs per 216 PA. The rule of thumb would work this way:

-5.36 + 20 = 14.64... 14.64/216 PA X (140 outs/150 outs) X 50 = 3.16

Maddux estimated predicted ERA will be 3.16

Take Donald, priced at 2.45M. Let's say that he allows 25 chances of on-base overall (so 5 more than league's average). His NERP in DD vs lh is at 7.64.

7.64 + 20 = 27.64...27.64/216 PA X (140 outs/135 outs) X 50 = 6.63

Ideally, you would also incorporate gbA in the mix (they generate more outs than other outs), errors from the pitcher (I think Diamond Dope does not consider them) and even hold numbers (paradoxically, a -6 holder will have barely any steal attempts, and therefore no caught stealings...a pitcher with a +1/+2 hold will have more caught stealings)

Of course you have to consider the stadium and the ballpark singles and homeruns from the pitcher.

Thank you. Very Helpful.