Productive Outs Are Not
Productive

by Cyril Morong

Click here to see my sabermetric blog called Cybermetrics

From the ESPN website “A Productive Out, as defined and developed by ESPN The Magazine and the Elias Sports Bureau: when a fly ball, grounder or bunt advances a runner with nobody out; when a pitcher bunts to advance a runner with one out (maximizing the effectiveness of the pitcher's at-bat), or when a grounder or fly ball scores a run with one out.”

Do they matter?

To look at this, I ran a regression in which team runs per game was the dependent variable and team OPS (on-base percentage + slugging percentage) was the independent variable. I used data from 2004. If you know where data on productive outs in years before 2004 is available, please let me know.

The regression equation was

R/G = 14.17*OPS – 5.99

The r-squared was .942, meaning that 94.2% of the variation in runs per game across teams is explained by OPS. The standard error was .125, or about 20.3.

Then I added in productive outs per game (PO/G).

R/G = 14.15*OPS - .083*PO/G – 5.88

The r-squared was .942. Yes, the coefficient on PO/G is
negative. Meaning that with OPS held constant, __as PO/G rises, runs per game
actually goes down__. The coefficient
on PO/G was not significant.

I also broke down OPS into on-base percentage (OBP) and slugging percentage (SLG). The regression equation was

R/G = 16.85*OBP + 12.83*SLG – 6.31

The r-squared was .944

I again added in PO/G. The regression equation was

R/G = 16.85*OBP + 12.83*SLG – .15*PO/G - 6.20

The r-squared was .945. Again, the coefficient on PO/G is negative, meaning that as PO/G rises, runs per game actuall falls. It was not significant.

** CONCLUSION:**
PRODUCTIVE OUTS DO NOT SIGNIFICANTLY INCREASE SCORING. THEY MAY EVEN
HURT SCORING.

In 2004, the relationship between team winning percentage and team OPS differential (your OPS minus your opponent’s OPS) was

Pct = 1.439*OPSDIFF + .496

The r-squared was .814. The standard error was .0366. Then I added in Productive Out differential and got

Pct = 1.435*OPSDIFF + .0082*PRODDIFF + .496

The r-squared was .8144. The standard error was .0372. Adding in productive outs did not change the nature of the relationship or the predictive power. The t-value for PRODDIFF was .247, so it was not significant. In fact, its impact on winning is very slight. The best team in PRODDIFF was the Angels, at .2778. The worst was the Diamnondbacks at -.494. So the Angels were about .77 better. Multiplying that by .0082 and we get .006 gain in winning pct. For a 162 game season that is about 1 win. So to even win one more game, you have to go from the worst to the best team in productive out differential. If we compare the Angels to an average team, with a PRODDIFF of zero, they would win .37 more games. So if an average team in productive outs became the best team, they could add .37 more wins.

** CONCLUSION:**
PRODUCTIVE OUTS HAVE ALMOST NO IMPACT ON WINNING, HOLDING OPS CONSTANT.

Back to Cyril Morong's Sabermetric Research