Friday, August 24, 2007

'03-'06 Reds Player Total Run Value

One of the most consistent flaws with how we, as fans, view players is that we classically have had far better and more accessible statistical tools with which to evaluate offense than we've had for defense. As a result, defense is often an afterthought when we evaluate a player--perhaps little more than a tiebreaker when evaluating a player's contributions.

Nevertheless, we are increasingly getting better and better tools to evaluate player fielding performance. I think even the most advanced metrics currently available, such as MGL's UZR or BIS's +/- system, are still fairly "primitive" in what they do compared to what we eventually can expect to have at our fingertips, but they're good enough to give us a good idea of how players perform on defense.

What I wanted to do was sit down and try to get a comprehensive view of how position players have contributed to the Reds' success over the past several years--both on offense as well as defense. This is my first stab at this, and therefore I'm going to be extremely simplistic in my approach:
  • Production estimates for all players are based on comparisons to hypothetical replacement players at the same position(s).
  • Offense was measured as Value Over Replacement Player (VORP), a BPro stat that estimates runs produced above what would be expected from what a waiver-wire player could produce. Replacement players at defensive-oriented positions like SS or CF have lower production than replacements at offensive-oriented positions like 1B or LF, so real players must produce more at those offensive positions to get the same offensive credit.
  • Fielding was measured using Ultimate Zone Rating (UZR), one of the more advanced fielding metrics available, which reports runs saved above or below league average. I am making the assumption that replacement players will provide league-average defense. There is precedent for this assumption (though perhaps it works better at some positions than others--I didn't worry about that). If players played more than one position, I simply summed their UZR values at all positions.
  • Total player production, which I'll call total run value, is estimated by simply summing offensive and defensive contributions (VORP + UZR). While one can argue that the value of an offensive or defensive run varies depending on how many runs a team is otherwise scoring or allowing, this is not an unreasonable way to estimate total player contributions.
  • I ignored catchers, because UZR doesn't address them and I didn't feel like trying to come up with my own estimate of their defense (perhaps opposing team SB-CS runs?).
First, here are the data graphically:
The horizontal axis shows player VORP, while the vertical axis shows UZR. The diagonal line is replacement level--where VORP and UZR sum to zero. One would hope that everyone on the team is either at or (hopefully mostly) to the right side of the line. Data points are player seasons (one per player per season), and all represented at least 20 games that season.

The immediate thing that has to concern you as a Reds fan is that there are a number of points well below (to the left of) the line. That means, based on this definition of replacement level, that the Reds have had several players from '03 to '06 who were performing substantially (by 10+ runs) below replacement level. It's pretty darn hard to win if you are playing guys who are performing at a level below what could reasonably be expected of a waiver wire claim (think Jason Ellison or Pedro Lopez).

Let's look at the individual values. This table lists the data for '03 through '06 Reds, sorted by total run value. I set a cutoff of 20 games played to keep the table size under control (players below this threshold were mostly grouped near the middle of the table). I'll also note that I've occasionally encountered oddities with the UZR data, especially in terms of games played. Hopefully those are clerical mistakes in the output spreadsheet that don't affect the defensive estimates, but it's worth keeping in mind:

Year Name G UZR VORP TotalRunValue
2004 Casey 127 5 56.4 61
2004 Dunn 141 -5 53.8 49
2005 Lopez 133 -4 45.8 42
2005 Dunn 152 -7 45 38
2004 Jimenez 137 6 27.3 33
2006 Aurilia 105 4 27.4 31
2003 Boone, A 106
2 25.2 27
2005 Casey 117 3 22.5 26
2006 Freel 113 8 16.4 24
2003 Casey 128 11 13 24
2006 Hatteberg 102 7 16.9 24
2005 Randa 123 2 20.4 22
2006 Kearns 158 4 15.7 20
2005 Aurilia 105 -1 20.4 19
2004 Freel 143 -3 20.1 17
2005 Kearns 110 8 9 17
2003 Jimenez 132 0 15.9 16
2006 Phillips 116 -7 22.6 16
2004 Larkin 84 -7 22 15
2005 Freel 108 -2 15.4 13
2003 Larkin 53 2 9.8 12
2005 Griffey 131 -41 52.4 11
2004 Pena 90 -8 18.7 11
2003 Castro 85 10 0 10
2003 Kearns 82 -4 12.7 9
2003 Branyan 45 3 3 6
2006 Denorfia 23 4 1.1 5
2003 Freel 32 -1 5.8 5
2004 Kearns 58 2 2.4 4
2005 Encarnacion 54 0 3.7 4
2003 Griffey 43 -13 16.5 4
2006 Dunn 164 -22 23.5 2
2004 Lopez 74 -3 4.3 1
2004 Castro 70 3 -3.4 0
2004 Bragg 21 3 -3.5 -1
2003 Dunn 117 -11 10 -1
2003 Stenson 23 -2 0.5 -2
2005 Jimenez 22 1 -2.5 -2
2006 Lopez 147 -17 14.9 -2
2006 Encarnacion 103 -20 17 -3
2004 Griffey 80 -25 21.9 -3
2003 Hummel 24 -3 -0.9 -4
2004 Larson 25 -2 -3.1 -5
2003 Lopez 53 -1 -4.4 -5
2003 Taylor 41 -3 -4.2 -7
2004 Hummel 32 -3 -6 -9
2004 Cruz 24 -8 -2 -10
2003 Olmedo 56 -3 -9 -12
2003 Mateo 50 -4 -8.5 -13
2003 Larson 30 -1 -13.5 -15
2003 Pena 41 -10 -5.1 -15
2005 Pena 90 -26 9.5 -17
2006 Griffey 108 -34 16 -18

The Good
  • Sean Casey checks in with three of the top 10 performances in this dataset, thanks to excellent--or at least fair--offensive production, combined with solid defensive performance according to UZR. I should add the qualifier that UZR looks primarily at range, which doesn't always tell the whole story for first basemen--though Casey was generally considered a good at saving errant throws from infielders, so I think his rankings here are not unreasonable.
  • Adam Dunn turns up twice in the top 4, thanks to outstanding offensive production in '04 and '05 (perhaps his peak seasons), and defense that wasn't nearly as bad as his critics would have you think in those years. Of course, '06 was a different story (see below).
  • Felipe Lopez's All-Star (and probably career) year in '05 comes in at #3. His defense wasn't good that season, but it was certainly good enough for a 40+ VORP shortstop to be a real asset to the team. It's in '06 that he fell off the wagon.
  • Jimenez, Aurilia, Freel, Boone, Kearns, Hatteberg, and Randa all show up as having made good contributions on offense and defense, all producing at least 20 total runs above replacement level in at least one season. ... it's notable that Aurilia, at least, was rated by other defensive metrics as well below average defensively last season, so a little skeptical of these numbers.
The Bad
  • The single worst season in this analysis was Ken Griffey's in '06. His lack of OBP last year caused his offensive value to plummet, while he continued to struggle horrifically in center field. Overall, the data indicate that he may have cost the Reds almost 20 runs compared to a replacement-level player in center field last season when you evaluate both offense and fielding together. Yikes--that's two wins!
  • If you look things on a per-game basis, however, Wily Mo Pena probably wins the title of worst performer. In two seasons, '03 and '05, his defense was atrocious, especially (for whatever reason) in right field. And his offense was certainly nothing to write home about either.
  • Other remarkably bad performers include Brandon Larson (offense), Ruben Mateo (everything), and (to my surprise) Ray Olmedo in '03.
A few other observations:
  • How much did Adam Dunn's awful slump over the last two months in '06 cost the Reds? His offensive production was rated at only 23.5 runs over replacement level in the same year that his defense collapsed to -22 runs vs. average! Ouch. This season, his offense seems to be close to its '04-'05 form, though his defense has been only slightly better than '06 (-8 runs through the All Star break). When he has hit like he is capable, Dunn's been extremely valuable...but when he has struggled ('03 and '06), his value has literally disappeared.
  • Griffey's best contribution in the '03-'06 time period was 2005, but even then, his wonderful +52 VORP was almost canceled by his terrible -40 run effort in CF. Overall, this analysis estimates that he has performed slightly below replacement level for the Reds over the scope of the study (4-3+11-18 = -6). I have a hard time believing that...but there it is.
  • Juan Castro got a lot of playing time in '03, and hit right at replacement level for a shortstop--which is the weakest offensive position. And yet his very good defensive skills resulted in a net contribution of about 10 runs over replacement. Not great, but it's the 25th (of 78) best performance by a Red in the time period of this study.
  • I remember there being a lot of concern about Larkin's defense at the end of his career. These data indicate that he really wasn't all that bad. Even in his last season ('04), his offense made up for his reduced range.
Anyway, this is just a first stab at this kind of analysis. But what the data tell me is that it's very important that we start more explicitly considering defense in player valuations. I'm not entirely sold on this methodology--Griffey being rated as below replacement level is hard to fathom--but at the same time, I have a feeling that these estimates aren't that far off the mark. The Reds' defense over the past many seasons has been terrible, and I think this analysis helps show just how much poor defensive play takes away from a team's ability to actually win games, and a player's ability to help his team win.

3 comments:

  1. Hi,

    Another wonderful post. I just have one quick question. Why aren't Runs Created park-adjusted? 100 RC in a park like Coors or GAPB is not the same as 100 RC in Shea or Dodger Stadium (Chavez Ravine). Shouldn't they be park adjusted to get a truer indication of the offensive value a player brings to the table?

    Thanks,
    Dave

    ReplyDelete
  2. Hi -- great point.

    Here I'm using VORP for offense.. and I'm not sure if VORP is park corrected already or not. BPro's glossary is decidedly unhelpful on that point (have I ever mentioned that I really dislike proprietary stats?). I'll see what I can unearth.

    UZR, at least, does include park corrections. That's one of the things, along with adjustments for handedness of hitter and pitcher, that set it apart from all other fielding systems. Of course, how exactly that's done is more or less a secret, so I'm uneasy about that one too. :)
    -j

    ReplyDelete
  3. Excellent article -- fielding really does make a difference, huh?

    VORP has its problems, but it DOES include park adjustments.

    ReplyDelete