NERP and Catcher Defense

That's a pretty cool rule of thumb Rich

I like Rich's rule of thumb as well. And the plethora of factors involved underlines just how complex any attempt to assess catcher defense truly is.
I think one of the challenges of evaluating catcher defense is that there are so many facets to weigh against one another. For infielders, you only have range and errors. If a SS has a 1e24 rating, it's possible to compare this--in terms of NERP analysis--to a 2e10 or a 3e17 at SS. For this comparison I use Dean Carrano's charts.
OF throwing arms add an additional factor that has so far, as I understand, been hard to assess in formal terms. Still, we know that a 2e10(-2) RF is going to be to some dedgree better than a 2e10(+1) RF, even if we don't know exactly how much better. Dean Carrano, as far as I know, hasn't yet tried to assess OF arms.
But, to use egvrich's example, what about a 2(-1)e1 T3 PB 3 catcher?
What is catcher range worth? What is the throwing arm worth? What about fielding errors (e1)? What about throwing errors? What about passed balls? And then, we have to mix in the hold and WP proclivities of the pitching staff, which might have an impact on the catcher's defensive value. A staff full of -2 and -3 holds might tend to minimize the damage of a +1 arm or exaggerate that damage if the pitcher holds are bad.
For example, when I've used Yogi's c-2(-2)e7,T-15(pb-2) card, I try to team him up with pitchers with minus holds. A -2 catcher's arm and a pitcher's -2 hold = -4. This would tend to minimize steal attempts, making Yogi's high throwing errors less of an issue.
Given all the complexities, I kinda doubt that we'll ever achieve a definitive means of assessing catcher defense in terms of NERP. This makes Rich's rule of thumb all the more handy, especially if adjusted on lab's gut level, in terms of the conditions of the league, park, and pitching staff.

I think it is all guts and thumbs if you wanna do well

Well any estimates?

A 1e4 at SS has a 7.2 NERP---1e10 has 5.9---2e22 is a neg .9
1e9 in LF/RF has a neg.9 NERP----3e9 has neg4.7----4e12 is a neg 8.2

So the range is basically +6 to -8 NERP. I max-score fairly usable DH-types at neg 7.5 NERP

In looking at the "average" team over 26 ATG8 80mil DH leagues, offenses posted: .281/.349/.444 for 876 runs with 110 steals/45 CS (71.2%) and 182 HR
----------------------------------------------------------------------
How close to that 1e10 SS (~5 NERP) do catchers like Bench get? 1B, 3B, and OFers top-out at about 2.5 NERP if they're near-perfect.

I've got the ATG9 non-Catcher Pool compared to the Catcher pool charted (total NERP and Price on the axis). When you plug in Catchers at that -8 NERP like they're a terrible DH, they show about maybe about 10 NERP below where they "should" be. It does imply that even the worst defensive catchers are still above that -8 NERP in value as on the whole, the group is way below the non-catcher baseline when penciled in at -8 NERP.

Jack Clements at 4(+2)e7, T-15(pb-12) sits just on the low end of players at his price at 19.83 NERP---looks like about 21 is market average there. (this is Clements at -8 NERP) Only players like him approach the line. 9.61 Piazza (29.1 NERP) is just below it at 4(+1)e6, T-10(pb-2). The next lowest total NERP in Piazza's strata is ~32 NERP, or 3 above Piazza's. So that puts Piazza's glove at at least -5 NERP (3 above the -8 floor)----comparable to that 3(0)e9 LFer?

Elmer Flick $8.68 with 29.34 NERP is the next guy to have comparable NERP to $9.61 Piazza (29.12) at -8 dNERP.

~B. Hamilton $10.11 (30.93), Speaker $10.22 (32.59) are close

Harper, Bryce $10.49 35.96
Larkin, Barry $10.49 38.78
Vaughan, Arky $10.49 35.45
Giambi, Jason $10.48 35.10
Cobb, Ty $10.44 35.03
Foxx, Jimmie $10.42 36.46
Mays, Willie $10.40 37.03
Averill, Earl $10.34 35.21
Snider, Duke $10.34 36.79
Wilson, Hack $10.29 34.97
Rodriguez, Alex $10.26 36.20
Speaker, Tris $10.22 32.59
Jeter, Derek $10.21 34.96
DiMaggio, Joe $10.19 36.12
Mantle, Mickey $10.19 33.61
Mays, Willie $10.19 37.12
Mays, Willie $10.12 36.34
Trout, Mike $10.12 35.51
Hamilton, Billy $10.11 30.93
Cuyler, Kiki $10.08 36.06
Pujols, Albert $10.05 37.39
Sisler, George $10.05 35.29
Heilmann, H. $10.04 36.00
Lajoie, Nap $10.04 36.01
Helton, Todd $10.02 35.60
Jackson, Reggie $10.01 35.78
Boudreau, Lou $10.00 33.67
Rodriguez, A. $9.98 35.66
Snider, Duke $9.97 35.72
Seymour, Cy $9.95 33.42
Goldschmidt $9.94 35.39
Cobb, Ty $9.91 35.06
Bagwell, Jeff $9.86 34.83
Delgado, Carlos $9.86 34.05
McCovey, Willie $9.86 34.99
Simmons, Al $9.84 35.97
Thome, Jim $9.81 34.24
Walker, Larry $9.79 35.38
Kemp, Matt $9.76 36.51
Schmidt, Mike $9.76 35.82
Alomar, R. $9.72 33.96
Boone, Bret $9.7 35.89
Biggio, Craig $9.68 35.54
Pujols, Albert $9.68 35.44
Musial, Stan $9.66 33.05
Kiner, Ralph $9.65 33.59
Delahanty, Ed $9.61 33.99
Piazza, Mike $9.61 29.12
Gehringer, C. $9.59 33.43
Williams, Ted $9.59 32.62
Stanton, G. $9.55 34.53
Carew, Rod $9.54 32.46
Yastrzemski $9.54 33.17
Clark, Jack $9.53 31.92
Heilmann, H. $9.53 32.91
Wells, Willie $9.53 35.47
Alomar, $9.52 33.88
Dihigo, Martin $9.51 36.01
Trout, Mike $9.51 35.20
Waner, Paul $9.51 31.80
Williams, Ted $9.5 32.65
...
Mauer, Joe $9.22 24.36


*table feature doesn't work on boards? Sorry.

Joe Mauer's exceptional glove at $9.22 mil shows 24.36 NERP when his dNERP is an obviously untrue -8 dNERP. This shows about 6.5 - 11 NERP "behind" what $9.22 should get you (~32 NERP), which puts Mauer's stellar c-1(-3)e1,T-4(pb-4) at about +2.25 NERP. Mauer is a *15game risk so all his NERP scores are subject to .975 multiplier (inj risk). So if he's 10 NERP behind market value, he needs to go from -8 dNERP to +2, and that's 10.25 more before inj .975 multiplier.

That +2.25 dNERP makes c-1(-3)e1,T-4(pb-4) be worth what a 1e27 at SS, 1e6 at 3B, 1(0)e0 in CF is worth! Super anecdotal and at the "high-end" $9mil scale---seems low except when you compare it to the CF !!! Still less valuable than a CF who gets +1+ more NERP from a great arm.

Nels,

I went back and looked at the assumptions in my original catcher charts, so maybe some of this helps:

J. Mauers -3 arm will on average produce 80 SB attempts at a 65% success rate. So 52 SB's and 28 outs. Additionally, his T rating of 4 will produce an extra 2.4 bases 80*.15*(4/20) on throwing errors. That converts to about 7.5 runs (using ERP factors of .19 and -.0976).

M. Piazza +1 arm will on average produce double that 160 SB attempts but at the same 65% rate, so 104 and 56. Also his higher T rating will produce an extra 12 bases on throwing errors. That converts to about 16.6 runs using ERP. (around a +2 arm, you tend to start seeing the SB % go up as well as the SB attempts, but primarily at arms better than this it is simply the SB attempts that are impacted by the arm not the safe %).

However, to adequately compare these two catchers, you have to make one final adjustment, since baseball is measured by outs, not PA's, SB attempts, etc. In the above example, Mike Piazza actually produces 28 more outs than Mauer does, which means 28 outs through normal PA's no longer needs to be generated. If you go through a normal ATG lineup trying to generate another 28 outs, you can expect roughly another 5 runs to be generated, so the gap between Mauer and Piazza is not 9 runs, but closer to 4 runs specific to throwing arms. This is yet another piece of the secondary impacts on catchers arms and why I don't believe the spread is nearly as impactful as a number of others might believe.

(although Piazza gets used a lot, so there might be a few more believers out there).

Comparing Mauer to equivalents at other positions, I think the AT9 pricing model effectively puts him at the equivalent of a 1e32 or a 2e13 at SS, which basically highlights that the top catchers can not produce the defensive value that the top SS can, since those are average defensive ratings at SS for your typical ATG team.

I will make one other point specific to T ratings. A bad T rating, only becomes a big issue when it is combined with a bad throwing arm. Using the examples above, Joe Mauer would only make approximately 12 throwing errors, if his T rating were a 20, because his -3 arm keeps runners from ever attempting to steal in the first place. Piazza is only expected to make 12 throwing errors with a T-10 and a +1 arm, so even at this level of poor defense, it still isn't too impactful. Now at the upper end of +1 arms and a T-20 you can expect 24 throwing errors, so somewhere in there you cross a really bad line and once you go with a +2 arm, is when you get the scary throwing error numbers.

The take away is that if you are going with a decent throwing catcher a bad T rating should not be the deal breaker.

Fascinating stuff childsmwc I always appreciate your insights into pricing. I admit T-rating is an area I've been dismissive of thinking the effects aren't large enough to be worth measuring but you've inspired me to give it another look. Your comments about adjusting defensive runs for OBP are also interesting and the effect of going from 9 runs to 4 runs sounds much larger than the adjustment I've been making for the same thing so I will have to give that some more thought as well.

Question: I assume catchers were priced with the assumption that the catcher-blocking-the-plate rule is operative? Any thoughts on this thread which casts some doubt on whether that setting is actually in use? https://forum-365.strat-o-matic.com/com ... ababb6fcec

Great insights.

I think the (lack of) value of the catcher D is something that has to be specifically exploited.

So, if it is a smallball team that went with a Fred Carroll or something, and they play against the type of smallball team that Rich or I might put together...they will be in trouble. All those runners create an asymmetric advantage that is not part of the average modeling and will result in additional bases than expected. But it is a fringe result and we are talking teams in the 300 SB range for the season.

I see a lot of "large" smallball teams that play to the big OBA/Hits and don't have much speed. This can be very successful with the right players in an average league of course. But an extreme "tiny" smallball team that is fast and steals up and down the order and plays great D/keeps games close, will almost always have an advantage against that other small ball team model in a small ball park. Or at least that has almost always been my anecdotal experience and is not based on something more than that.

And by average I don't mean in terms of opponent skill. I mean in terms of typical league environment.

J. Mauers -3 arm will on average produce 80 SB attempts at a 65% success rate. So 52 SB's and 28 outs. Additionally, his T rating of 4 will produce an extra 2.4 bases 80*.15*(4/20) on throwing errors. That converts to about 7.5 runs (using ERP factors of .19 and -.0976).

Comparing Mauer to equivalents at other positions, I think the AT9 pricing model effectively puts him at the equivalent of a 1e32 or a 2e13 at SS, which basically highlights that the top catchers can not produce the defensive value that the top SS can, since those are average defensive ratings at SS for your typical ATG team.

Thank you so much!

is (using ERP factors of .19 and -.0976) a reference to NERP ala DiamondDope and http://www.mfooz.com/bblog/wp-content/uploads/offense-vs-defense.pdf?

I see how you get the 2.4 bases etc. Does that plug into NERP? .19 could be .1875 for gbAs and -0.976 could be -0.085 for ABs ???? If so, is there a baseline you're...adding that 2.4 bases to get to 7.5 runs? Does that translate to NERP?

The range you gave from a -3 arm to +1 was 80 to 160 SB attempts (+80) over four "levels. That's just 80, 100, 120, 140, 160 sb attempts if consistent increase? Also the extra errors you mentioned for Mauer T-4, Mauer T-20 and Piazza, the T-rating changes seem consistent as you change the arm?

I have to factor in arm rating to my T-rating NERP cost. I threw together some #s on Sunday to ~get 4(+1)e6, T-10(pb-2) to -6 dNERP and a studly catcher to somewhere around +2 NERP...

Range: - too big a factor in my estimates (?)
1: +1.2 dNERP
2: 0 dNERP
3: -1.4 dNERP
4: -3 dNERP
5: 5 dNERP

Arm - basically -0.7 for each arm. Was 0.5 and maxing at like 2.5 for -5, but I wanted a -0 arm to not be exactly a “0” NERP. Plus gotta reward those rare -4s or Pudge!
(-5): +3 dNERP
(-4): +2.3 dNERP
(-3): +1.6 dNERP
(-2): +0.9 dNERP
(-1): +0.2 dNERP
(-0): -0.5 dNERP
(+1): -1.2 dNERP
(+2): -2.0 dNERP
(+3): -2.8 dNERP - not really relevant/considered
(+4): -3.7 dNERP
(+5): -4.7 dNERP

Error
-.22 dNERP per error, starting at +0.22 for e0 (so 0 for e1, -0.22 for e2 etc) Kinda lazy, but this seems about the rate for an error on Berce’s offense/defense PDF

T-Rating -- So these changes should cross-reference with overall arm?
Should this scale consistently if you keep everything else equal? 2(+1)e4, T-1(pb-0) vs 2(+1)e4, T-20(pb-0) I’m asking if that progression is consistent 1 additional T at a time?
*On the other end of this, the #s you mentioned with Mauer T-4, Mauer T-20, Piazza, the T-rating changes seem consistent as you change the arm.
Here were my Sunday ~#s
T-0: +0.4 dNERP
T-1: +0.3 dNERP
T-2: +0.2 dNERP
T-3: +0.1 dNERP
T-4: +0 dNERP
T-5: -0.15 dNERP
T-6: -0.3 dNERP
T-7: -0.45 dNERP
T-8: -0.6 dNERP
T-9: -0.8 dNERP
T-10: -1.0 dNERP
T-11: -1.25 dNERP
T-12: -1.55 dNERP
T-13: -1.85 dNERP
T-14: -2.15 dNERP
T-15: -2.45 dNERP
T-16: -2.75 dNERP
T-17: -3.05 dNERP
T-18: -3.35 dNERP
T-19: -3.65 dNERP
T-20: -3.95 dNERP

PB-ratings
pb-0: +0.3 dNERP
-0.15 dNERP for every rating. So pb-10 is -1.2 etc

*A 1e8 at SS gets +6.1 in offense vs defense article.

These are the totals it was producing to fit within this kinda range (Berce shows great as negative, I'm just calling it psoitive)
c-1(-4)e0,T-1(pb-0) +4.32 dNERP
c-1(-3)e3,T-4(pb-1) +2.51 dNERP
1(-2)e5, T-6(pb-2) +0.92 dNERP
1(-1)e3, T-17(pb-1) -1.94 dNERP
c-2(-1)e2,T-6(pb-2) +0.17 dNERP
c-2(-4)e3,T-6(pb-1) +1.71 dNERP
c-2(-1)e1,T-11(pb-4) -0.07 dNERP
c-3(0)e1,T-3(pb-1) -1.65 dNERP
c-3(-3)e2,T-6(pb-6) -0.92 dNERP
c-4(+1)e5,T-7(pb-6) -6.3 dNERP

things that felt wrong that your insights "fix" --- will improve as the T-rating tax is tied to arms like you mentioned
c-1(-3)e8,T-18(pb-2) -2.09 dNERP
c-1(-2)e10,T-15(pb-7) -3.08 dNERP
c-1(-2)e13,T-15(pb-14) -4.79 dNERP

Nels,

On the topic of NERP used in that article, I actually prefer Paul Johnson's original Estimated Runs Produced factors when evaluating the ATG strat environment. I have found when you apply ERP values to actual ATG team stats they produce the smallest variances between the estimated runs to actual runs scored over multiple park and salary cap environments.

A few words of "caution" using that article you cited, and these are points I debated with Dean years ago. Not all NERP is created equal. In that article the NERP value you are deriving is based on a fixed number of plate appearances and the conclusion being reached is if two players have the same NERP over 108 PA's they are the same value. That is an erroneous conclusion.

Baseball is not measured by plate appearances but by outs. Therefore, once you have computed a NERP value you must make a replacement level adjustment to your hypothetical lineup. To the extent that the player being evaluated consumes less outs than the "baseline" he has effectively created additional opportunities for the rest of the lineup (which can then be converted to an RC value as well), the opposite is also true. This adjustment applies to the defensive side of things as well.

As a hypothetical, the difference between a SS 1e4 and a 1e11 might be an additional 7 errors during the season (the equivalent of 7 extra hits in an ERP formula). However, the issue doesn't end there, because now your opponent's ATG lineup gets to keep batting until they have generated those 7 outs. Assuming a .350 OBP environment that means another 11 PA's for your opponent to replace those 7 outs, which will also generate approx 1.3 more runs in the process.
So the impact of 7 errors is not simply 7* NERP factor for singles, you must also layer on the impacts of the additional runs to replace the outs.

T- Ratings occur in game on stolen base attempts when the successful steal role is a 1-3 on the twenty-sided dice. So the formula for the T-Rating impact is a function of SB attempts over a season (which is a function of the arm behind the plate). That is why my defensive arm evaluations are based on the combination of arm and T-rating. Assuming constant SB attempts for all catchers with the same arm, then the impacts of the T-rating would be linear as the T-rating in this equation increases (SB attempts*15%*(T rating/20)).