Thursday, April 05, 2007

2006 Reds Fielding Review

Fielding is a more difficult for the average fan (well, me, at least) to evaluate over the course of the season than pitching and hitting, because traditionally there haven't been very good statistics to help us make sense of what we see. With pitching and hitting, it's relatively easy -- we can count hits, homers, runs allowed, etc, and have a good century of more advanced statistical assessment to fall back on as we evaluate players.

But fielding hasn't had that benefit. Sure, we can count errors, and this clearly told us that Eddie Encarnacion wasn't doing a very good job of making accurate throws from third last season. And we can count assists from the outfield whenever someone gets gunned down at the plate. But these numbers don't tell us much about the more subtle, and arguably more important aspects of fielding, like how much range a fielder has, how well a second baseman pivots on the double play, or whether runners are intimidated by an outfielders' arm and therefore avoid even trying to take the extra base.

Fortunately, tools to assess some of these things are finally becoming available, and here I've tried to use them to piece together some details on Reds player performance last season--hopefully I'll be able to explain enough to keep you folks with me as we move along, as some of this stuff is a bit new and somewhat obscure. Let's start with a look at total team defenses from around the league last year:
Here I've plotted two +/- totals of Reds' defensive performance. The horizontal axis plots BIS's +/- totals (those found in the Fielding Bible) for 2006 teams, which I extracted from The Hardball Times 2007 Annual. The units are +/- plays saved vs. average, and each data point is essentially just the sum effect of all players on each team. On the vertical axis are THT's +/- team totals, which are based on batted ball statistics. They ask how many runs the team defenses saved, given the average outcomes of the batted ball types (i.e. ground ball, fly ball, bunt, liner, etc) these defenses encountered.

As you can see, both the BIS and THT data indicate that the Reds' 2006 defense was bad--either 2nd or 4th to last in the league, depending on how you measure it. Yikes. One positive I'll throw out is that this is actually a substantial improvement compared to 2005. THT's +/- runs saved totals for the 2005 Reds came out as a disastrous -57 runs, so their -18 total this year represents almost a 40-run improvement (that's ~4 wins!). BIS's data in the fielding bible rated the '05 Reds as -96 plays below average, so by that metric the 2005 Reds were better by 35 plays...even if this still put them 2nd to last in the league.

An aside: the last time the Reds were above average in fielding according to BIS's numbers? 2003, when they were 12 plays over average thanks to an above-average infield (Casey, Jimenez, Boone, and Larkin...with help from Castro, Olmedo, Lopez, and Branyan).

We can look closer at the BIS +/- plays statistic, which is split into the infield and outfield in the THT Annual, to learn where the Reds are failing:
Totals for infield and outfield defense using the BIS +/- system, as extracted from The Hardball Times Annual 2007. Units are plays made vs. average.

It turns out that the Reds infield and outfield were both really bad last season, though the outfield was responsible for a few more plays that didn't get made. In the case of the Reds' infield, the middle infield (-15 plays) was slightly worse than the corner infield (-8). Hmph.

With this not-so-rosy picture of the Reds' defense in mind, let's go through the positions and try to understand the underlying causes of these trends.

Before we get started, a few quick notes about the graphs that follow: +/- ZR data were extracted from The Hardball Times' stats output, and were converted to a +/- stat following the methodology I developed here, though I did standardize them to roughly a full season's playing time (defined as 200 BIZ for 1B, 400 BIZ for all other positions) for all players. +/- PMR data are reported as actual plays - expected plays, as reported by David Pinto's PMR system. I also standardized them to a ~full season's worth of playing time (defined as 4000 balls in play).

Infield
+/- PMR vs. +/- ZR for all infield positions. The blue vertical and horizontal lines indicate average performance according to PMR and ZR, respectively. Green bars indicate the range of the first upper standard deviation for the 2006 MLB infielders (it varies a bit by position), whereas the red bars indicate the range of the first lower standard deviation. Good fielders will be in the upper-right quadrant, with truly outstanding fielders being above and to the right of the green bars. Bad fielders will be in the lower-left quadrant, with truly terrible fielders below and to the left of the red bars. Scott Hatteberg and Brandon Phillips overlap slightly in this figure (sorry).

Let's take this one position at a time.

First Base

Last year first base was essentially a platoon position, with Rich Aurilia and Scott Hatteberg splitting time. PMR rated them as roughly equivalent performers there defensively, while ZR rates them as drastically different: it says Hatteberg had slightly above average range, while Aurilia was awful.

In the case of first base, I don't really trust ZR as much because there are apparently a TON of plays made by 1B's outside their zones of responsibility--so much so that out of zone plays sometimes are more numerous than plays in the zone. I don't think BIS defines their 1B zone very well. So in this case, I'm comfortable with concluding that Aurilia and Hatteberg were both roughly average defenders last season. Hatteberg may actually have been a bit better than average: he was listed as the 4th (tied with several others) best first baseman in baseball last season according to the BIS +/- system as reported in the Bill James Handbook 2007...the only Reds fielder who made one of those good lists.

There is more to playing first base than fielding grounders and liners, of course--the best first basemen are often lauded for their ability to help their infielders by converting wild throws into outs. Unfortunately, while BIS apparently has a mechanism in place to track such efforts, I don't have access to it. About all I can report is that both Aurilia (0.994) and Hatteberg (0.996) had fielding percentages that ranked near the middle of the pack for qualified first basemen last season (mlb range = 0.988 to 0.998, median = 0.995). So at least they weren't dropping balls that they should catch.

Second Base

Shortly after Krivsky acquired him on April 7th, Brandon Phillips took over the second base job and never gave it back. The primary reason that he played there initially was his defense, which certainly looks impressive when you watch it. The kid's got flair.

But both ZR and PMR agree that Phillips is actually only marginally better than average defensively. Of course, an average defender, like an average hitter, is quite a bit above replacement level, so he still has a lot of value--particularly if he continues to improve his offensive game like his spring training slugging percentage indicates he might. But it's not fair to say that he's among the league's best defenders, at least based on his 2006 performance.

Third Base

(Note for the technically oriented: I used a straight [+/- PLAYS] + [+/- OOZ] calculation for ZR at third base. I'm no longer convinced that 3B should be treated differently than all the other positions, despite the findings I reported here--see comments in that link for discussion).

In 2005's second half, Edwin Encarnacion was actually rated by the Fielding Bible as a plus-defender. But last year, he clearly struggled, as evidenced by his 25 errors in just under 1000 innings (0.915 fielding percentage -- last among qualified 3B's in 2006). This hurt his overall range substantially, with both ZR and PMR listing him as a full standard deviation below average. Eddie absolutely has to improve his defense this year. He seems to have reasonably good ability to get to balls, but he has to be able to consistently field and throw the ball across the diamond to allow those skills to be a positive for the team. If he could drop his errors by 10 this year--a reasonable number--he'd move up to being almost (though not quite) average.

Nevertheless, for all the bitching and moaning about EdE's defensive struggles, the guy who received the second-most innings at that position--and sometimes caused Eddie to be benched in his favor--was actually worse. While it's true that Rich Aurilia didn't make nearly as many errors (0.953 fielding percentage), PMR indicates that he would have made upwards of 20 fewer plays overall than Encarnacion, once we normalize for a full-season's playing time. 20 plays! ZR rates them closer, but still indicates that Aurilia was not as good as Eddie. After writing my Reds 2006 Hitting Review, I was ready to nominate Aurilia as the Reds' most valuable position player for 2006. But when you factor in his defense, I don't know that I'd say that any more.

Shortstop

(Note: Data for all three Reds' shortstops listed above are based on the total 2006 data for each player, not just those innings they played while employed by the Reds).

Felipe Lopez was often criticized as a miserable defensive shortstop, and these data indicate that those criticisms were entirely justified. I'm a big proponent of strong defense up the middle (I'll readily compromise on the corners in favor of offense), so I was fine with the idea of dealing Lopez prior to The Trade. Unfortunately, the guy they got to replace him--while better--was also below average. Royce Clayton wasn't awful last season defensively, but he was still below average. And given the minimal amount of offensive production he provided, there wasn't much point in having him in a starting role. In fact, it would have made far more sense to declare Juan Castro the starting shortstop. Castro was rated well above-average at short last season by PMR, and actually out-hit Clayton substantially (Clayton: 0.235/0.290/0.329, Castro: 0.284/0.320/0.421) while with the team. I doubt Castro would maintain that sort of offensive production in the long term, but if you're going to have a no-hit shortstop, you might as well have a guy who can catch the ball.

Speaking of no-hit shortstops, how would our new SS, Alex Gonzalez, fit in? Since this is a 2006 Reds review, I didn't include him in the graph, but Gonzalez's playing time-adjusted ratings last season were +3 outs above average according to PMR and a whopping +27 plays above average according to ZR. This would place him near the top of the graph, just right of the vertical blue line. I'm inclined to split the difference between these two stats and place him somewhere around 10-15 plays above average. That would put him slightly below Castro in terms of range, but Gonzalez will probably hit better than Castro, at least in terms of power. Hopefully.

Bunts and Double-Plays

The BIS statistics reported in the Hardball Times 2007 Annual include separate assessments of the ability of teams to make plays on bunts and double plays. Here's a graph of all the teams' performance in those statistics (vertical and horizontal blue lines indicate league averages):It's interesting to see that no teams fell deep into the -GDP, -Bunt category...though the Reds were on its edge. Their performance on bunts was actually right around league average, but they didn't do well on turning double plays.

While I don't have access to individual the BIS player stats, David Pinto recently released PMR-based assessments of shortstop double-play ability that shed some light on those numbers. All three Reds' shortstops last season were rated as below average: Lopez and Castro were both rated 4 double plays below average, while Royce Clayton was rated a miserable 14 below average. Alex Gonzalez, by contrast, was rated 3 plays above average, which ranked him in the upper third of MLB shortstops. Again, defense should be improved with Gonzalez at short.

Outfield

Here's a graph comparing ZR and PMR for qualified Reds' outfielders (all lines and bars are defined in the same way they are in the infield graph--scroll up for details):
First, let's talk about the good: Ryan Freel. Freel came out looking brilliant in this analysis, showing some of the best playing time-adjusted range in baseball. I always knew he was fast out there, but his performance places him a standard deviation above the mean in both metrics. Amazing.

Unfortunately, the other two mainstays in the outfield--Adam Dunn and Ken Griffey Jr.--were bad. Zone Rating liked Dunn better than PMR, but both agree that he's below average...it's just that PMR thinks he was a disaster. My impressions were that Dunn had an awful first half in the field, and then improved from that point on. But I don't have data on that.

Even worse, however, was Ken Griffey. Junior, according to both PMR and ZR, was the worst center fielder in baseball. PMR puts a full season of Junior in center field as resulting in 32 plays below average. My ZR translations put him at -41 plays per season. We knew he was bad, but that rate has him missing a ball that an average fielder would have caught in one out of every four games. Yikes.

A few outfielders didn't make the graphs because they played in too few innings to make the PMR reporting cut, but below I'll list the ZR +/- figures for a few of these individuals. Keep in mind that these are extrapolations for a full season's worth of play based on a small number of innings in the field, so I do not consider them very reliable:
  • Left Field
    • Ryan Freel, -5 plays/season
    • Chris Denorfia, +11 plays/season
  • Center Field
    • Chris Denorfia, +32 plays/season
    • Quinton McCracken, -14 plays/season
    • DeWayne Wise, +11 plays/season
  • Right Field
    • Todd Hollandsworth, -9 plays/season
    • Chris Denorfia, +41 plays/season
    • Norris Hopper, +55 plays/season
Outfielder Throwing Arms

John Walsh has put together a very nice system for evaluating throwing performance. It has two primary measures: Hold Rate, which measures the proportion of runners who try to take an extra base on an outfielder's arm, and kill rate, which measures the rate at which a fielder actually guns down base runners. Here, I'll borrow a page from Walsh's most recent article on this system and plot Reds' outfielders for for these numbers (100 is the mean for each score, higher is better). Unfortunately, I could only find data on three Reds outfielders:
Both Griffey and Kearns were successful at impacting the running game without flashing a lot of kills, given their opportunities last year. Dunn, on the other hand, had only average success at controlling base runners, while also having a below-average kill rate (wasn't he a QB??).

Griffey, in particular, surprises me here. His hold rate was about the same in 2005, but his kill rate was a substantially higher (though still below average) rate of 71. I wonder if his throwing arm is starting to go along with his other skills--and that base runners hadn't caught on to this yet, resulting in a still-positive hold rate.

It's also possible, of course, that Griffey makes up for any lack of throwing arm by being really good at getting the ball back into the infield. My own observations (caveat: I'm no scout!!) suggest to me that he plays angles really well, that he's really quick to get rid of the ball, and that he has an accurate arm, even if it's not that strong. Doing those things can go a long way toward preventing base runners from advancing.

Catchers

There are two primary elements to catching: working with the pitcher to help them excel, and shutting down the opponents' running game. The latter is easier to examine than the former:
Two measures of catchers' abilities to affect the running game: % of would-be base stealers caught, and stolen bases attempted per game. The latter might be considered an indication of runner intimidation. As usual, solid lines indicate MLB averages among qualified catchers (I used 578 innings, as there's a big drop-off between that player, Gerald Laird, and the next catcher, Gregg Zaun). Dotted lines represent +/- one standard deviation from the mean.

Reds catchers were extremely good at stopping the running game in 2006. All caught an above-average number of would-be base stealers, with Dave Ross and Jose Valentin coming in more than a standard deviation above average. In fact, only these qualified catchers had equal or better caught stealing percentages than our two guys last season: Ivan Rodriguez (0.46), Gerald Laird (0.41), Yadier Molina (0.41), & Jose Molina (0.41). Perhaps reflecting their arm strength (and probably also reflecting the reputation of GABP as being a hitters' park), there were also very few stolen base attempts against Reds catchers last season.

Handling pitchers is tougher to assess. One important aspect of a catchers' job is to serve as a backstop--one often hears announcers talk about how important it is that pitchers can trust their catchers to knock down anything they throw up there, even if it's bounced in the dirt. Another aspect of catching has to do with "calling a good game," which essentially means helping pitchers get the most out of their skills on any given day. Here's an attempt to assess those things:
On the vertical axis is Wild Pitches + Passed Balls per game. As long as you're not catching a knuckleballer, this is generally an indication of how good you are at stopping errant pitches and preventing runners from advancing. The solid blue line is the MLB mean of qualified catchers, with dotted lines being +/- one standard deviation. The horizontal axis lists catcher ERA, which is simply the ERA of pitchers while the catcher was in the game. The red vertical line is the Reds' team ERA in 2007.

First, let's deal with the y-axis. Both Jason LaRue and Javy Valentin looked like absolute brick walls up there, stopping almost every pitch that comes their way. In contrast, Dave Ross was "only" average.

Catcher ERA is interesting. It indicates, on the surface, that pitchers threw far better with Ross behind the plate than Valentin. But one needs to be very careful about interpreting this stat. First, Ross was the personal catcher of Bronson Arroyo even before he took the starting catching job from LaRue, and Arroyo enjoyed a career year last year, finish up 4th in the league with a 3.29 ERA AND leading the league in innings pitched. Ross was behind the plate for most of those starts. Valentin, on the other hand, due to his Spanish skills, was the personal catcher of Elizardo Ramirez. Ramirez had a nice couple of months for the Reds, but imploded badly later in the year. So those two pitchers alone could explain a big part of the variation we're seeing here.

What I should have done, but haven't taken the time to do, is to compare the ERA of each pitcher with each catcher, and then noted the average effect each catcher had on the staff. Maybe I'll sit down and try to figure out an efficient way to do that sometime. But for now, I'm going to punt: for what it's worth, Valentin's CERA the previous three years was 4.92 (team ERA 5.15), 5.51 (team ERA 5.19), and 4.77 (for tampa bay, team ERA 4.93). So I don't think his 2006 numbers are revealing any kind of real deficiency in his performance behind the plate.

Looking ahead to 2007

Defense is still an underappreciated part of the game, but I think the statistical fan community is finally coming to grips with how important a role it plays in determining team wins. And this season, the Reds defense looks like it will be much improved.

With the departure of Aurilia, Lopez, & Clayton, and the addition of Gonzalez, along with the improvement I expect to see from Encarnacion, the Reds infield defense could actually be average or even slightly above-average this season. That would go a long way toward helping to improve the Reds pitching staff's performance--and a lot of our guys might need that help. This will be a solid core to build on, and I'm looking forward to watching them play.

Similarly, the outfield may be much improved, simply by shifting Griffey from center field to right field. Based on 2006 data, a full season of Ryan Freel in center field instead of Griffey would net the Reds a total of somewhere between 46 (PMR) and 72 (ZR) additional outs. If you assume slightly less than a run saved per half a play, that works out to somewhere between 20 and 30 runs saved, which is the equivalent of ~2-3 wins. We might lose some off this total by sacrificing Freel's defense in right field to Griffey. But Junior should be much closer to league average out there than he was in center, so the loss shouldn't be too dramatic.

Dunn will, in all likelihood, continue to be bad, although his comments this spring about trying to become a good outfielder are encouraging...simply because it's never appeared to be a major emphasis in his game. If he could make himself even an average fielder, given his offense, that'd be huge.

Tables
As with the hitters and pitchers, I'm closing by posting a set of relevant fielding stats for the 2006 Reds, courtesy of The Hardball Times and Baseball Musings. Many, many thanks to those sites for allowing free access to such great data.

Last Pos Inn PO A TE FE FPct +/-ZR +/-ZR (~162G) PMR+- PMR+- (~162G)
Hatteberg 1B 1088 997 69 1 2 0.996 2 4 3 4
Aurilia 1B 329 301 26 1 1 0.994 -10 -39 2 6
Dunn 1B 17 15 1 0 1 0.941 -2 -93 --- ---
Encarnacion 1B 9 9 0 0 0 1.000 0 -43 --- ---
Phillips 2B 1216 331 334 8 8 0.977 2 3 7 7
Freel 2B 91 23 25 1 0 0.980 2 34 --- ---
Aurilia 2B 41 13 14 0 0 1.000 0 -3 --- ---
Womack 2B 40 11 17 0 0 1.000 1 34 --- ---
Olmedo 2B 40 5 13 0 2 0.900 1 34 --- ---
Harris 2B 12 4 3 0 0 1.000 -3 -266 --- ---
Castro 2B 5 0 2 0 0 1.000 0 34 --- ---
Encarnacion 3B 931 74 196 16 9 0.915 -16 -28 -13 -18
Aurilia 3B 356 24 78 3 2 0.953 -7 -31 -11 -40
Freel 3B 88 7 25 0 0 1.000 3 58 --- ---
Castro 3B 64 4 16 0 0 1.000 3 88 --- ---
Lopez SS 1336 187 389 14 14 0.951 -20 -21 -35 -33
Clayton SS 1054 171 340 6 12 0.965 -8 -10 -7 -9
Castro SS 548 89 190 3 5 0.971 6 15 12 28
Olmedo SS 21 4 6 0 1 0.909 -2 -111 --- ---
Phillips SS 15 1 3 0 1 0.800 -3 -178 --- ---
Aurilia SS 198 37 68 0 1 0.991 -3 -16 --- ---
Dunn LF 1321 279 7 4 8 0.960 -11 -9 -21 -21
Freel LF 41 10 1 0 0 1.000 0 -5 --- ---
Denorfia LF 27 4 0 0 0 1.000 0 11 --- ---
Wise LF 22 4 0 0 0 1.000 1 65 --- ---
McCracken LF 14 3 0 0 0 1.000 1 145 --- ---
Ross LF 9 1 0 0 0 1.000 -2 -155 --- ---
Hollandsworth LF 8 1 1 0 0 1.000 -1 -122 --- ---
Hopper LF 3 1 0 0 0 1.000 0 145 --- ---
Griffey Jr. CF 870 229 6 3 2 0.979 -27 -41 -22 -32
Freel CF 400 127 4 0 0 1.000 9 31 4 14
Denorfia CF 85 20 0 0 0 1.000 2 32 --- ---
McCracken CF 40 15 0 0 1 0.938 -1 -14 --- ---
Wise CF 36 14 1 0 1 0.938 0 11 --- ---
Hopper CF 14 1 1 0 0 1.000 0 11 --- ---
Kearns RF 1228 346 7 5 2 0.980 16 13 3 3
Freel RF 360 101 7 1 4 0.956 8 23 8 29
Hollandsworth RF 118 22 0 0 0 1.000 -1 -9 --- ---
Denorfia RF 118 34 2 0 0 1.000 5 41 --- ---
Hopper RF 70 25 1 0 0 1.000 4 55 --- ---
McCracken RF 10 3 0 0 0 1.000 0 -28 --- ---

7 comments:

  1. Heh, now THAT's a long post. Congrats to anyone who makes it to the bottom. -j

    ReplyDelete
  2. Whew. It was worth reading the whole post, but you are right -- it took a long time!

    Good stuff.

    ReplyDelete
  3. I'm hoping that the graphics are intuitive enough that folks can look at them and get something out of this, even if they don't want to read through the whole post. I won't do stuff this long very often (takes too long to put together!!), but this is a once/year review, so I wanted it to be reasonably comprehensive.

    ReplyDelete
  4. Thanks. :)

    I do consider this something of a stats blog (see ZR post a few days back, for example), though my primary interest is in using stats to analyze Reds players. :) -j

    ReplyDelete
  5. Great stuff once again. I expected Edwin to be a little higher than he was, but everyone else I placed about the same as what your numbers came up with.

    ReplyDelete
  6. I have to read this post on the installment plan!

    I found it interesting that a recent report on the Nationals mentioned that Felipe Lopez was resisting a move back to shortstop, saying it would hurt the team - while I hadn't even heard that he'd moved to second. But apparently he was quite aware of his defensive shortcomings at SS.

    The use of the color bars on your graphs is a great addition.

    Some of these things that you're doing - e.g. the standard deviation bars, or plotting WPA against VORP in your 2006 Reds Hitting Review - are so interesting and informative that after seeing you do them, I'm amazed I haven't seen anyone else do them as standard practice. And that should be taken as a great compliment; it's an excellent original perspective you bring to these analyses.

    ReplyDelete
  7. I have to read this post on the installment plan!

    Yeah, I probably should have posted this in parts (e.g. infield, then outfield), but I just wanted to get it out there. Live and learn I guess.

    Some of these things that you're doing - e.g. the standard deviation bars, or plotting WPA against VORP in your 2006 Reds Hitting Review - are so interesting and informative that after seeing you do them, I'm amazed I haven't seen anyone else do them as standard practice. And that should be taken as a great compliment; it's an excellent original perspective you bring to these analyses.

    Thanks! I'm really liking this graphical approach I've been taking lately. I'm finding it to be far more insightful, not just because you can make such interesting comparisons between pairs of stats, but also because you get a much better feel for the distributions.
    -j

    ReplyDelete