15 February 2011

Has the Orioles defense improved in 2011?

Motherwell's Mark Reynolds
Last week, I read through Andrew Gibson's work on using infield and outfield Defensive Efficiency on ground balls and fly balls, respectively.  I discuss that post over at the Baltimore Orioles Round Table (BORT), but on this blog I wanted to jumped in on something that I felt was slightly undone.  As you may have noticed, almost every statistic these days is converted over to runs.  It is the high falutin' goal of any respectable statistician to devise ways to convert all measures of player performance into how many runs are scored or given.  It permits a great deal of comparison and it is why when guys like us so repetitively use WAR and other run based numbers, others think we are making too much out of single statistics.  We really are not, we are just converting things over to a relatively easy model for comparison.  Converting numbers to runs does not remove problems with the original numbers.  We are all aware that a high WAR due to the defensive component is probably not an accurate portrayal of the player's talent.

My objective in this post is to take Andrew's numbers and convert them over to runs saved/given as well as compare that to other systems like DRS, TZL, and UZR.  All of that after the jump.

Below is a table showing Andrew's original numbers and my runs conversions.  This was done by recalculating all of the Defensive Efficiencies from Baseball Reference for the Orioles and for the entirety of Major League Baseball.  This enabled the generation of a league average Defensive Efficiency.  Based on the number of plays the Orioles made, the difference in Defensive Efficiency between the average and Orioles performance could then be converted into plays made above or below average.  In turn, the plays were then converted into runs above or below average by multiplying the plays by an average infield or outfield coefficient taken from this particular study.  After all of that . . . we are left with this chart:
By converting it over to runs saved and given, two things become apparent.  One, the Orioles routinely show us what a sub-standard defense looks like.  Two, these numbers are concerning for a playoff contending team, but probably not the Orioles.  The infield in the past has cost the Orioles on upwards of 3.3 games in the standings (2008), but these losses are often compensated by the outfield.  Last year, a slightly below average outfield was paired with a poor infield.  It cost the team 1.6 wins.  In the grand scheme for a 75 or so win team ... this is not much of a deciding factor.  If this team's talent base in hitting and pitching was projecting for the mid 80s or higher, then it might be something to look at with Brian Roberts and Mark Reynolds being the likely culprits in the infield and Adam Jones in center and Luke Scott in left probably not fitting those positions ideally.  But, we are getting ahead of ourselves.

How does the GB and FB Def Eff differ from other defense metrics like DRS, TZL, and UZR?

Here is a quick primer, John Dewan created his Plus/Minus system.  This system looks at double plays, bunts, stolen bases, outfield throwing, stealing home runs, and fielding.  All of this is converted into Defensive Runs Saved or DRS.  Total Zones with Location (TZL) is a system that was created by Sean Smith of the famous CHONE projection system.  This is a good primer on TZL.  This actually marks the first year that all of his ideas are proprietary as he has been mysteriously scooped up by a MLB team.  Finally, MGL's Ultimate Zone Rating (UZR) which is similar to TZL but utilizes different source data in a slightly different way.  This is a great place to go to learn up on UZR.

So how does the infield compare between these different measures?

For the other three established advanced defensive metrics, I summed the runs values for each player on the Orioles for a given infield position and then added those sums.  I did not need to convert the numbers to runs saved or given because each of those three metrics do it automatically.  This information is readily available over at FanGraphs.
As you can see there is quite a good deal of variety in the numbers reported.  Amazingly, the figures estimated for the Defensive Efficiency runs saved or given are often within the range of the other three and at most about 18 runs off the mean.  This may seem like a pretty poor system and it is as we have been spoiled by the ease of determining value for offensive data sets.  However, all we need to do is look at the grunting of fielding percentage and errors to recognize the femur bone that are these new defensive metrics.  Yes, none of these are likely to be silver bullets, but we have to show some patience.

In a more Orioles-centric slant, the other defensive metrics tend to agree that the Orioles have not been great in the infield.  Only in 2007 was the team considered average by the consensus.  If we assume typical aging for Derrek Lee and Brian Roberts (about a loss of 3 runs per year for players in their 30s) as well as Reynolds and Hardy maintaining their performance; plus average defense from replacement . . . I getting an average of -1 run next year.  That improves the infield by 1.2 wins.

This leads us to discussing how much has the outfield improved?

The following table was constructed the same way as the infield table was created above.
What I find interesting about these numbers is that UZR and TZL are piggy backing tightly over the past three years while DRS as well as the fly ball efficiency metrics are separately doing their own thing.  The numbers are still pretty similar, with about 18 runs off the mean as being the most extraordinary difference.  I think what does connote is that defense was not a very strong part of the Orioles outfield last year.  In fact, the only consistently good outfielder last year between the systems is Felix Pie and he has been benched in favor of Luke Scott.  Assuming that Markakis and Jones maintain their ability, that the backups provide average offense, and Scott only declines by 6 runs . . . I'm actually calculating the defense as -5.4 runs, which would be a 0.3 run improvement.  This is based on some pretty solid defensive scores Scott earned three years ago.  I think there might be an argument to be had that the defense this year will be similar to the defense last year.  Pie was able to chase down a lot of hit balls well, but he did show a nature for not having an organic understanding of defense.  My own personal feelings are that Scott's absence from the outfield for two years will cause a greater deterioration than 6 runs and we are likely to see something more in the neighborhood of a loss of 12 runs.  That would leave the team with a -11.4 outfield defense or a decrease of 5.7 runs from last year.

So how much has defense improved?

Based on the DRS, TZL, and UZR averages for offense (-1 run) and defense (-5.4 runs), that is an improvement of 13.5 runs or 1.4 wins.

Going back to Andrew's conclusion about how defense is an unlikely source to find a great improvement in wins for this club, I largely agree on the condition that a team that wins about 75-85 wins next year will not largely benefit from that amount of improvement.  If a team like the Cardinals or Yankees were able to improve their defense by that much, it would be far more meaningful.

No comments: