When is an error more likely to occur in a game?

Last Friday I wrote an article for The Hardball Times breaking down events by their average Leverage Index. One event that especially intrigued me was the Reached on Error.

Ultimately I discovered that an ROE appears to have an average LI of around 1.05, meaning that it tends to occur more often in more important game situations. (Note that this value of 1.05 is actually different from the one I originally reported in the article. I go on to explain why in the comments.)

So I decided this morning to break down the frequency of errors by various game states in order to see what situations are more prone to a fielder's flub.

I used Retrosheet data from 2003-2012, regular season games only, and defined "error rate" as the number of times an error occurred per ball in play only. This means all pitcher pick-off errors and catcher's caught stealing errors are excluded. Home runs were not considered balls in play. In addition to ROE, this method also includes errors that have occurred during base hits. The league average error rate over this period was 1.96 percent.

In order to keep track of a potential scoring bias, I've also included a version of BABIP we'll call Non-Outs per Ball in Play, which is basically (H+ROE)/BIP.

Inning

Let's start with innings. Do more errors on balls in play occur in the later innings when games are more often on the line? Or do fielders slack off in the early goings of a contest?

If we break up error rates for each game from 2002-2012 by inning we see that there is a slight tendency for fielders to make more errors in the later innings rather than early on. Error rates then peak if the game goes into extra innings, where we see 2.1 percent of balls in play botched by the defense:

Inning	Errors	BIP	ERR%	NO/BIP
Early (1-3)	8,509	443,633	1.92%	30.3%
Middle (4-6)	8,696	453,183	1.92%	30.5%
Late (7-9)	8,034	394,152	2.04%	30.2%
Extras	508	24,011	2.12%	30.3%

We see the lowest error rates in the second inning (1.86), where presumably the fielders have not fatigued, and the better hitters at the top of the order have likely been passed up. Outside of extra innings the eighth inning actually holds the highest error rates with 2.09 percent, with the ninth inning close behind at 2.03 percent.

The trend is ever so slight, but it is noticeable with such a large sample. I imagine fatigue plays some part in the decline, but I'm not sure how much of an impact it has. I am also not so sure there is any worry of scorer bias here-- it would seem cruel to hand out errors just before you hit the showers.

Score

Our next concern then would be run differential. Are fielders more likely to punt a grounder when they are several runs up in a blowout? Or do their nerves get to them in tight contests, tensing up their throwing arms and wreaking havoc on their release points?

Interestingly, we see that when the fielding team is ahead their error rate is at its lowest, and when the fielding team is behind their error rate is at its worst. During tie games, the error rates are almost perfectly average.

Score	Errors	BIP	ERR%	NO/BIP
Ahead	9,278	491,865	1.89%	30.2%
Tie	6,713	341,083	1.97%	30.4%
Behind	9,756	482,031	2.02%	30.4%

How much of this is caused by the fact that the losing team is more often the worse team, and the worse team more often has the worse defense? Good question. That is certainly something we can test for, but it will have to wait for another day.

Error rates are at their lowest when the fielding team is up by three runs (1.77), and their highest when the defense is trailing by two (2.08).

Base state

But we find the biggest factor in determining why batters reach base on error more often in higher leverage situations is base state.

Base state	Errors	BIP	ERR%	NO/BIP
Bases Empty	11104	728837	1.52%	30.5%
Runners On	14643	586142	2.50%	30.1%

With runners on the fielders commit almost 60 percent more errors than in situations where the bases are empty. Of all the eight base states, error rates are at their lowest with the bases empty and at their highest with the bases loaded at over 3.5 percent. Intuitively, this makes perfect sense, but I'll admit I did not anticipate the effect of runners on fielding errors to be this dramatic.

With men on, fielders are out of position holding the runners. For the same reasons, we see more hits on balls in play with runners on as well. More grounders get through the infield and more grounders slip just out of reach, deflecting the glove. But the scorer has no sympathy for this.

This may also be an issue of distraction as well. Not only in trying to keep the runner close, waiting for a pickoff play up until the final millisecond of delivery. But also when the ball is put into play. Runners on situations provide many more moving parts, many more obstacles to the routine grounder.

Outs

There is a jump in error rate from no outs to two outs, and my first guess was that this is only due to the fact that we are likely to see more baserunners as the inning matures.

So in order to attempt to control for this I looked at error rates by out state during bases empty situations only.

Errors with bases empty

Outs	Errors	BIP	ERR%	NO/BIP
0	5108	325826	1.57%	30.9%
1	3334	227580	1.46%	30.2%
2	2662	175431	1.52%	30.2%

This is likely for the same reason we see higher home run rates with none out and none on than we do with two outs and none on. My best guess is that pitchers are more likely to pitch in the zone with a clean slate. Both strikeout rates and walk rates are also lower with no outs and none on. Pitchers may be looking for quick outs in such cases, making the batter work his way aboard, rather than risking the free pass.

As a result we might see better quality of contact with no outs. And this means more long balls as well as more errors. But we should also be aware that we are much more likely to see better pitchers pitching in situations with no one on and two outs.

Base/out state

While error rates jump with two outs during bases empty situations, for nearly every other base state we see a drop in error rate as we hit two outs. This is most likely because the double play is no longer a concern with two outs.

Bases	ERR% 0 out	ERR% 1 out	ERR% 2 out
___	1.57%	1.46%	1.52%
1__	2.71%	2.47%	1.74%
_2_	2.51%	2.20%	2.14%
12_	4.02%	3.47%	2.30%
__3	2.20%	2.03%	1.43%
1_3	3.15%	3.51%	1.73%
_23	3.16%	2.95%	2.33%
123	4.04%	4.63%	2.40%

Bases loaded, one-out scenarios seem to be the messiest for major league defenses, while runner on third, two outs just barely edges out bases empty for lowest error rate.

Batted ball type

A few other notes before we wrap it up here. Errors mostly occur on plays in which ground balls are hit as you'd probably expect. Fewer errors occur during pop ups than fly balls, but just barely. Here is the full chart (keep in mind we're still including throwing errors here as well):

Batted ball	Errors	BIP	ERR%	NO/BIP
Flyball	1856	337533	0.55%	17.0%
Grounder	20624	621984	3.32%	26.1%
Line drive	2796	249302	1.12%	70.9%
Pop up	471	106160	0.44%	2.5%

Position

Third basemen have committed the most errors (of any kind, not just limited to BIP errors) from 2002-2012, just barely beating out shortstops for that dubious honor. Center fielders have recorded the fewest errors over that same time period.

Pos.	Errors	weighted AGE
CF	1590	27.9
LF	1776	29.2
RF	1933	28.9
C	3208	29.5
1B	3296	29.4
P	4122	28.5
2B	4396	28.7
OF	4418	28.8
SS	6629	28.2
3B	6640	28.7

Center fielders, incidentally, appear to be the youngest position over that time period, which is interesting because younger players on average do tend to commit more errors than their more seasoned veteran teammates. I'll spare you the chart for this, however, because 1.) it is probably wrought with survivor bias, and 2.) I think we've had enough charts for one day.

Any other aspect of the game state you'd like to see broken down with regard to error rate? Anything I may have missed? Anything I may have misinterpreted? Great potato salad recipe? Let us know in the comments.

. . .

James Gentile writes about baseball at Beyond the Box Score and The Hardball Times. You can follow him on twitter @JDGentile.