Pitchers and NERP

Diamonddope.com

I've long been a big fan of Dean Carrano's Offense vs Defense in Stratomatic.

I seem to remember a pitcher conversion for the nerp rating found on diamonddope. I don't see one of the pdf now. Does anyone know if one exists? I remember stumbling on the thought that errors really could matter for sp even with the assumption that they only get ~1/5th of your teams defense rolls at the pitcher position.

What are folk's general thoughts on br vs nerp? I've really been doing more nerp lately (alot of work with the cost curve for hitters) but know there can be large disparities between BR and NERP. <-----extra need pitcher defense conversion ; p

Thanks

Obviously nerp attempts to value gbA which is great, but outside of this advantage, how does NERP differ or improve on BR?

Is a NERP a good unit of measurement or too abstract?

Formula for NERP is:
NERP (New Estimated Runs Produced) = .318 * TB + .333 * (BB + HBP) - (gbA * .1875) + .25 * H - .085 * AB

A single consists of 1TB, 1H and 1AB, so the value of a single in this formula is .318 + .25 - .085= 0.48.
Similarly, 2B = 0.80, 3B= 1.12 and HR=1.44.
These values should sound similar because they are virtually the same with BR.
For recall, the formula for BR as defined on diamond dope is 1B=0.47 2B=0.78 3B=1.09 and HR=1.40

What distinguishes most NERP and BR is that NERP integrates the value of gbA (a positive asset) and that the value of outs differ: outs are worth -0.085 in NERP and -0.25 in BR. This last difference though is trivial in itself because they simply reflect differences in the comparator: You should use the value -0.085 when you wish to estimate How many runs a player will produce compared to a player who will produce 0 runs (i.e. A very bad player, roughly worth 0.5M) whereas you should use the -0.25 value when you wish to estimate how many runs a player will produce compared to a somewhat sub-average player (more likely a 2M worth player with baseline defense). You could even use a stronger value if you had another comparator in mind.

So both formulas are similar, and consequently, they both have the same limitations. The most important is they underestimate the impact of on-base. A player who walks not only increases the net run production of 0.33 runs, he also allows his team to have an extra at-bat. Both formulas assume a zero value to this extra at-bat, which is kinda similar to assume that a bench player worth 0.5M will get this extra at-bat. There is even an argument to make that walks are worth more than 0.33 regardless of this last point---they should be 0.34 or 0.35, because walks are randomly distributed in Strat, compared to real life where non-intentional walks are more frequent, and less valuable, with first base open.

The two formulas also don't value clutch hits (roughly +/- 0.1 each) and of course they don't include speed (and holds) and defense unless you follow Dean's suggestions.

To answer your other questions, a very bad fielding pitcher will allow a bit less than 2 singles/E1 per 108 chances and a few si+E1 and E2, so the net value is nearby 1 NERP or 1BR perhaps a bit worse

Very informative, fascinating stuff. Thanks.

On a somewhat unrelated topic, I wonder if we could / should add the eight pitcher batter cards to the diamond dope database to get a visual and statistical comparison of the "contribution" pitchers can make in non-DH leagues. Obviously the one batting card is an automatic out, but the other seven cards may at least be better than someone like Ernie Andres at the plate.

I think fielding and hitting for pitchers are too often overlooked and ignored when choosing pitchers. I've had teams where 20% of my team's errors were by the pitching staff - not a trivial factor at all.

Agreed Mr.Typhoon. Way back I had an all e0 pitching team that just felt like such a killer edge. I'm sure I came up with it via some kind of Nerp adjuster for pitchers pitching cards.....or at least the rough implementation of an imagined one. It matters. Errors are a b and I remember, Russ Ford with his 43 errors or so, won't pitch 162 games, but he'll pitch alot.

Incredibly interesting stuff Marc! Thank You! My main two Qs were:

1). If some has a pitcher NERP defense adjuster or knows how to calculate it? Dean's article says 1Bmen and P have the same # of defensive X-rolls (18) per 108 PA. ---- is holding the runners the only different thing, and can thus be ignored for pitchers? ALSO AND MAINLY This NERP adjuster is to be added to a pitcher's PITCHING CARD. Is this appropriate and overallllllllll

2). How good of a metric do you strato-tyrants find NERP to be for Pitcher's pitching ability?

1) NERP defense adjuster for pitchers

The worst fielder among pitchers avaiable in ATG is Frank Bertainia. His fielding rating is 5e51. The NERP value of his defense is +1.09. On the opposite hand, the Glavines of ATG world (1e0) have a NERP fielding value of -0.17. Everything else is in between.

To put that in a formula is kinda complicated, because some errors are added to hits, but if you want it, here it is

A) 2*(ROUNDUP(1.5*(range-1);0)/20)+errors/54*2*ROUNDUP(1.5*(range-1);0)/20+errors/54*2*(20-(roundup(1.5*(range-1);0)))/20.

B) Anything in excess of 2 of this sum must be put in another column.

C)Then you take the sum (in A) X 0.56 + the excess (in B) X 0.31 minus 0.17.

BK and WP have a slightly greater range. The worse ratings are found in Dave Stewart and Ontiverious (curiously, both Oakland As members of the 1988) with balks and wp at 17 apiece. The net NERP adjustment is + 2.09 for these two pitchers. You have to consider though that WP happen twice as frequently than BK, so for example David Cone's 1988 card (BK=17 WP=9) has roughly the same NERP adjustment on this regard compared to Nolan Ryan's 1986 card (BK=0; WP=17)--both at 1.4. 1990's Dave Stewart (BK=0, WP=6) has a NERP adjustment of 0.5. In fact, these are roughly estimates...in theory you should adjust these ratings for the on-base presence (a pitcher with a WP of 8 will allow more wild pitches than a pitcher with a WP of 6 that allows half the on-base).

Holding too is affected by the level of on-base, so in theory it too should be influenced by how much on-base a pitcher will allow. I'm less on solid grounds on how to adjust NERP for holds in ATG, but my rule of thumb is to adjust the NERP by adding +1 for the very bad holders (+5 and above) and by substracting -0.15 to the best holders (-6):
+5 and above +1
+4....0.92
+3....0.83
+2....0.66
+1...0.51
0...0.38
-1...0.26
-2...0.15
-3...0.07
-4...0
-5...-0.1
-6...-0.15

Supertyphoon mentions the adjustment for hitters' pitching card. I didn't calculate the value of, say, a 8N hitting card or a 5w hitting card, but I have estimated that roughly speaking, by the time a pitcher will get through his pitching card (216 PA), there is roughly 3 offensive at-bats where a 8N or a 5w will get a hit instead of an out. So it would make sense to adjust the NERP for the hitting cards from 0 (for worst hitters) to +1.5 (for better hitters). This won't make choose Krukow over Klippstein (the latter has a much better card) if I'm looking for a cheap pitcher, but it would make me opt for Walter Johnson's 10.98M over Pete Alexander's 10.99 card, so it is something to consider.

2). How good of a metric do you strato-tyrants find NERP to be for Pitcher's pitching ability?

Not good enough considering that the impact of on-base is neglected somehow, as explained in other post above.

Something that works for me is to estimate how many outs a pitcher will get from 216 plate appearences, compared to an average pitcher. To illustrate, let's say I estimate that Maddux (in a league with highly potent hitters) will get 155 outs for every 216 PA, whereas an average 5M pitcher will get only 140 outs.

Then I take Maddux's final NERP and do the following adjustment:

raw NERP/140 X 155 = NERP-adjusted to onbase.

By doing this, I put Maddux's NERP value on an "out" basis, where it should be, rather than on a PA basis.

NERP should be based on outs because the bottom line of this game is to win after your opponent reached 27 outs. Your opponent may get 40 PA or 50 PA to reach those 27 outs, depending of the quality of your pitchers, but the numbers of outs are invariable if you win a 9-inning game.

Please note that the pitchers, like 1B men (and corner OF's), have two X-chances per pitcher card. For hitters it is necessary to convert to X-chances per card (by multiplying their number of chances by 9), but for pitcher if you do that you will be wildly ovecompensating for pitcher defense.