clock menu more-arrow no yes

Filed under:

Advanced Defensive Efficiency Refined

A couple weeks ago, I introduced the statistic advanced defensive efficiency. This statistic measures how well a team's defense performs based on the different types of batted balls they face. One of the problems that I had is that I didn't have exact batted ball numbers. Thanks to the Baseball Prospectus custom stat reports, now I do. I also tweaked my formula a little. Before I considered all home runs to be fly balls, now 88 percent of home runs are marked as fly balls and 12 percent are marked as line drives (don't worry, this isn't an arbitrary number). Here's how the new numbers work out:

Old Rank TEAM DER eDER aDER
1 SDN 71.80% 71.04% 0.76%
2 DET 71.20% 70.64% 0.56%
8 TOR 70.50% 70.05% 0.45%
9 SLN 70.40% 69.96% 0.44%
5 HOU 70.60% 70.76% -0.16%
3 NYN 70.80% 71.10% -0.30%
21 CIN 69.10% 69.56% -0.46%
7 SFN 70.50% 70.97% -0.47%
6 CHN 70.50% 71.03% -0.53%
18 ATL 69.20% 69.84% -0.64%
17 ARI 69.30% 70.05% -0.75%
4 NYA 70.60% 71.38% -0.78%
12 WAS 69.80% 70.63% -0.83%
13 SEA 69.80% 70.81% -1.01%
22 COL 69.00% 70.10% -1.10%
15 OAK 69.40% 70.68% -1.28%
23 PHI 68.80% 70.09% -1.29%
28 KCA 68.20% 69.52% -1.32%
26 BAL 68.50% 69.84% -1.34%
24 TEX 68.80% 70.21% -1.41%
14 MIL 69.40% 70.84% -1.44%
16 FLO 69.30% 70.84% -1.54%
20 LAN 69.10% 70.77% -1.67%
10 ANA 70.30% 72.08% -1.78%
11 CHA 70.00% 71.91% -1.91%
25 CLE 68.60% 70.62% -2.02%
27 BOS 68.30% 70.50% -2.20%
19 MIN 69.20% 71.48% -2.28%
29 TBA 68.00% 70.50% -2.50%
30 PIT 67.60% 70.35% -2.75%

Old rank- The team's rank in defensive efficiency.
eDER- Expected Defensive Efficiency.
aDER - Advanced Defensive Efficiency.
If any of these terms are confusing, see my previous article on this subject.

These results, which turned out very differently from my original results, shows the big problem with batted ball data: it's very subjective. Baseball Prospectus' interpretation of a line drive is apparently very different from the Hardball Times. Because of this, the final advanced numbers are largely negative. To better capture this data, I would need to recalculate the out percentages using Baseball Prospectus' data, which is impossible.

Nevertheless, I think the rankings are largely correct, just the final number to represent the rank is off. From here, I want to introduce park factors into the equation, something that won't be done until I finish my stat database.