I understand the Ryan Howard "he got older and slower" explanation, but could someone looking at data and seeing a guy go from a very high number of runs saved in a season on defense, to being negative the rest of his career, not think there is some flaw in the first number?

It may not be a flaw. Good defense is not necessarily linear due to interactions between defenders.

It could also have just been a fluke good season for him defensively. Just like a player with one offensive season that is much better than the entire rest of their careers.

Jason Michaels saved 11, Ryan Howard saved 11

Jason Michaels, while useful and versatile in the OF, also was nobody's idea of a run saver. The three-headed, two-position monster of Utley, Bell and Polanco saved at least 45 runs in 2005. (Utley's # is not cited but he must have had more than Rollins' 18). Hmm, I wonder if Polanco really saved 9 runs before June 8 (the infamous Ugueth Urbina trade) or the author grabbed all of Polanco's 2005 totals?

Fine; so we assume that Brady Anderson's 5 year home-run stretch of 12, 16, 50, 18, 18, while playing for the same team in the same park, is just 1 fluke, good year? There, with an outlier number, we worked hard for an explanation other than one fluke year, right? Here, with an outlier number in what I assume is a less than concrete result (if a guy hits 50 HRs, we know he hit 50 HRs; we do not have a guy in the stands judging whether he hit the ball hard enought to be a HR, where, with fielding, there is a subjective element on runs saved), why don't we more rigorously question the number?

There is a difference between rigorously questioning the number and assuming its false.

Okay, I hereby rigorously question the +11 number. I cannot watch game film of the 162 Phillies game that year, though, so it is impossible for me to check that side. However, looking at the subject player, and his history, and knowing that the numbers are derived at least in part with a subjective element by the recording party, I doubt the number.

Jason Heyward swings really hard, and the ball just sounds differently coming off his bat (lots of stories on each of these; just look them up). He increased his GB/FB ratio last year, so my by watching the Braves games, I have written him down for 37 HRs. How do we feel about this figure? That is how i feel about the +11 fielding for Howard.

David Bell? And we thought that just because he was old and slow and couldn't hit he was overpaid.

The 2006 Fielding Bible sez Howard is "particularly strong moving to his right. His agility and footwork are surprising for a man his size..." He was rated as +15 on plays to his right, +6 straight on, and -2 to his left.

Dewan's "personal ratings" had Howard as the 10th best defensive 1B in MLB.

Maybe Chase gobbled up a lot more balls to his left that year, giving Howard less chances to #### up the play and accumulate negative numbers?

just gut reaction but i don't 'buy' the phillies position. completely ok with seattle and tampa.

surprised not to see any cardinal teams listed

Worth noting that that is only about half a seasons worth of date on Howard. That is beyond unreliable. Fangraphs UZR has him +2.4 runs in that time. I feel pretty comfortable saying the +11 is in fact crap.

UZR has the 05 Phillies at 64.5 runs above average fwiw.

Indeed, he played 706.1 innings at 1B in 2005, almost exactly half of what he would play in 2006.

I understand the Ryan Howard "he got older and slower" explanation, but could someone looking at data and seeing a guy go from a very high number of runs saved in a season on defense, to being negative the rest of his career, not think there is some flaw in the first number?

There have been a few spectacular defensive flameouts. Griffey is the first example that jumps to mind, and Andruw went from brilliant to sack of tomato juice pretty quickly as well. So it is at least possible that the Howard number just reflects his particular career arc. Howard spent a lot of his younger years blocked by Thome, so we don't have MLB number from his younger, and presumably more nimble, days.

Howard spent a lot of his younger years blocked by Thome, so we don't have MLB number from his younger, and presumably more nimble, days.

Waah! Look it up -- you can attribute about 1/2 year waiting for Thome. No more. Thome got conviently hurt for a good part of the year when Howard was ready. Howard was merely interesting, until he exploded at AA.
There was a brief LF experiment with the "younger, and presumably more nimble" with Howard but he returned to first base within days, IIRC. :) The eye test said that he improved a little AFTER a couple of years in the majors. It could have been fluky, but it wasn't noticeable.

i'm not sure how much it matters since the best phillies defenders in 05 were mostly infielders, but that was the 2nd year that the phillies played in citizens bank park, and between years 1 and 2, they moved the LF wall back about 10 feet. i'm not sure how that would effect the numbers, but it could be an important consideration.

Indeed, he played 706.1 innings at 1B in 2005, almost exactly half of what he would play in 2006.

So, if I understand the facts correctly, he saved 11 runs in about half a season of play, and then never had a positive runs saved season the rest of his career? And that doesn't suggest a flaw in the data?

UZR has the 05 Phillies at 64.5 runs above average fwiw

I'm not clear on why we are supposed to have confidence in these ratings. UZR uses exactly the same data, and yet often comes up with significantly different assessments. Here are the combined 2011-12 ratings for every team by the two systems (I took 2 years to reduce the random noise). UZR says Marlins were a bit below average the last 2 years, while DRS says they were clearly the worst defensive team in baseball (-27 vs. -120). UZR says the Jays were equal to the Marlins (somewhat below average), while DRS says they were a very strong defensive team (+64). UZR says Cubs and Pirates were average, DRS says they both sucked.

Am I cherry-picking? Yes. There is certainly some correlation here (R^2 of .50). But the two systems differ by 31 runs on average, and that's before we get to questions about the reliability of the underlying data (of which there are many).

Team UZR DRS
Angels 87 87
Red Sox 78 60
Dbacks 75 77
Rays 62 107
Reds 56 76
Rangers 40 -4
Braves 30 80
Twins 30 -32
Giants 21 -4
Brewers 18 30
Mriners 14 18
Yankees 12 -31
Padres 10 22
Dodgers 6 17
Cubs 6 -50
A's 2 -13
Pirates -2 -54
W Sox -3 -8
Royals -4 37
Nats -6 -9
Phils -18 -67
Marlins -27 -120
Jays -30 64
Tigers -32 -18
Astros -48 -59
Cards -50 -10
Rockies -59 -54
Orioles -80 -34
Mets -83 -102
Indians -104 -57

So, if I understand the facts correctly, he saved 11 runs in about half a season of play, and then never had a positive runs saved season the rest of his career? And that doesn't suggest a flaw in the data?

Remember that guy who hit .400 for a month and a half, but never hit .300 ever in his career?

And that doesn't suggest a flaw in batting averages?

It's been said that defensive data requires twice as many games to be as useful as batting stats, so that's essentially the same thing.

Marlins -27 -120
Jays -30 64

That's some swing.

Howard, yearly changes in Rbat vs. Rfield, starting with 2006 vs 2005

+41 -20
-27 +5
-13 0
+18 0
-15 -9
0 0
-24 +7

I think you'd have a hard time arguing the offensive series is more stable than the defensive series. The average absolute change of the offensive series is nearly 20 runs per year while it's about 6 for the defensive series. There's a lot of year-to-year variation in performance in all aspects of baseball, defense doesn't seem particularly extreme in that regard.

Howard, Rbat, b-r vs fangraphs, 2005-12 (remember, this is vs. average batter)

14 16
55 61
28 34
15 22
33 36
18 23
18 20
-6 -3

Howard, Rfield, b-r vs fangraphs, 2005-12

11 2
-9 -4
-4 -2
-4 -1
-4 0
-13 -13
-13 -5
-6 -6

Fangraphs consistently rates Howard a better hitter and an equal or better fielder every year except 2005. The variance in the hitting numbers is about the same but the variance in the defensive numbers is lower in fangraphs.

#16, the LF wall was reconfigured after 2005, not before.

A bit more fun. Andruw was mentioned so why not. Yearly change in Rbat vs Rfield starting with 98 vs 97

+16 +7 (understated as move from RF to CF, +1.4 dWAR)
-3 +1
+21 -11
-36 +2
+27 -8
-10 -5
-5 -6
+21 +8
-8 -4
-29 +7
-9 -25 (238 PA in year 2)
+19 +4 (331 PA, shift to DH/LF)
+8 +6
-3 -4
-9 +1

Nobody really bats an eye when a player adds 21 runs of offense then drops 36 then adds 27 then drops 10 then (a year later) adds 21. That's an average absolute change on offense of about 15 runs and on defense of about 6.5. You get more PA than defensive chances but cut the offensive change in half and it's still a little larger than the defensive shifts.

How about a Howard type player. Here's Thome, starting with 96 vs 95 and ending at 2004 vs. 2003 (pre-injury and move to DH).

+20 0
-13 -6
-8 -7
+1 +11
-7 +8
+24 -15
+17 +1
-26 +2
-17 +4

Again no contest on which series is more variable. Again the average absolute offensive change is about 15 runs vs about 6 for defense.

Take a look at Cano sometime, he's a real roller-coaster.

None of this means that the defensive stats aren't more variable than they should be -- it's obviously possible that they are. But it does mean the "OMG, look at the way those numbers bounce around!!" type of evidence is absolute nonsense. They don't bounce around wildly ... or at least no more wildly than most other baseball numbers. If the fluctuations in offensive numbers don't bother you there's no reason the fluctuations in defensive numbers should either.

If the fluctuations in offensive numbers don't bother you there's no reason the fluctuations in defensive numbers should either.

They are bothersome, the numbers are just undeniable.

Remember that guy who hit .400 for a month and a half, but never hit .300 ever in his career?

And that doesn't suggest a flaw in batting averages?

The difference is that we know that defensive numbers
a) contain significant amount of measurement error
b) require a long time for the difficulty of chances faced to stabilize (over a short time a fielder may face a large number of balls that are easier/harder than balls on average hit into the same zones)

Add to that that we have another defensive measure which flat out contradicts the +11 measure, and Occam's razor is that the +11 number is worthless.

I keep flashing onto Kevin Maas and his 21 home runs in 79 games his first season (150 OPS+, the only above average number for his career).

On television broadcasts, when the defensive alignment is introduced, they'll usually give the team fielding percentage. Wouldn't a better first approximation be team BABIP? Maybe make adjustments for double plays, pick-offs and caught stealing.

On television broadcasts, when the defensive alignment is introduced, they'll usually give the team fielding percentage. Wouldn't a better first approximation be team BABIP?

Well that's basically defensive efficiency (1-BABIP). And yes, obviously still flawed, but much more useful than FP.