The percentage of runners left on base, also known as LOB% or strand rate, is a pitching/defensive stat that evaluates how many runners did not score per runner that could have scored. It is not actually measured by the “Left on Base” numbers in a box score, however. It’s calculated as:
(H + BB + HBP – R) / (H + BB + HBP – (1.4 * HR))
The formula above measures the runners that got on base but did not score out of runners that could have possibly scored. The term – (1.4 * HR) means that runners who scored via home runs are not considered runners. 1.4 runs is the run expectancy of a home run. Also, note that intentional walks are not considered.
If you think about the average hitter reaching base about 30 percent of the time, it makes sense that, if a runner is on base to begin with, he will score about 30 percent of the time. Invert that number and you would expect an average strand rate near 70 percent. Indeed, that is about what it is. Here are the average MLB strand rates for the past five years:
- 2010: 72.2 percent
- 2011: 72.5 percent
- 2012: 72.5 percent
- 2013: 73.5 percent
- 2014: 73.0 percent
You can see how the major-league average hovers around 72 percent but has crept up in the past several years. This is one more sign of the decline in offense around MLB; a runner on base is slightly less likely to score now than he was five years ago. Contrast that with the heart of the high-offense era: the MLB-average strand rate in 2000 was 70.2 percent.
During a single season, the variation in LOB% can be high. Starting pitchers routinely post strand rates in the mid-to-high 80’s and sometimes below 60 percent. Here is an idea of the variance you can expect in a single season:
Starting pitcher, ERA-qualifying season since 1955:
- Highest LOB% – 88.8 percent, John Candelaria, 1977 Pirates
- Lowest LOB% – 58.5 percent, Derek Lowe, 2004 Red Sox
First, note the survivor bias of selecting for ERA-qualifying seasons and 50 IP seasons. Pitchers with poor LOB% numbers are likely to get benched or demoted before they reach these thresholds, so they will not appear in the list above. Still, you can see that there is a 30 percent disparity between the best and worst ERA-qualifying seasons for starters. Why the 2004 Red Sox let Lowe rack up the innings he did, I will never know, but it obviously worked out for them that year.
Relief pitchers have a wider range of variance; the best ones consistently strand over 80 percent of their baserunners.
Relief pitcher, season (at least 50 IP) since 1955:
- Highest LOB% – 99.9 percent, Huston Street, 2013 Padres
- Lowest LOB% – 47.5 percent, Jose Mesa, 2007 Tigers and Phillies
On the surface, Street had a good year for the Padres in 2013 with a 2.70 ERA. Under the covers though, his unsightly 4.92 FIP was the worst of his career by far. LOB% tells part of the story; Street’s K rate plunged that year and he gave up 12 HR in 56.2 innings, which lead to the increased FIP, but he stranded most of the baserunners he allowed.
I cannot explain why the 2007 Phillies and Tigers let Mesa pitch so many innings for them. Perhaps it is because the year before, his strand rate was a respectable 80.9 percent and they thought he would be that good again. But at 41 years old and in his 20th season, thats wasn’t the case. 2007 was his last year in the game.
To give you a more current idea, here are the major-league leaders in strand rate for the past five years (qualifying starters only):
- 2014: 83.1 percent, Doug Fister
- 2013: 83.9 percent, Yu Darvish
- 2012: 82.7 percent, Jeremy Hellickson
- 2011: 82.6 percent, Jered Weaver
- 2010: 82.7 percent, Cole Hamels
So, single-season strand rates can vary quite a bit. Over the course of a career, though, the variance becomes much smaller:
Starting pitcher, at least 1500 IP since 1955:
Relief pitcher, at least 500 IP since 1955:
This is where regression to the mean comes in. Compare the single-season numbers above to the career numbers. The wide variance of the former and the narrow variance of the latter indicate that the spread of true talent is small in this stat and that regression to the mean is common.
If your favorite starting pitcher strands 80 percent of runners in a season, you should not expect him to do that next year or the year afterwards. Even if he is a once-in-a-generation talent like Ford or Koufax (or Clayton Kershaw, whose career strand rate is higher than ord and Koufax, but who is 130 innings shy of the threshold for this study), 78 percent seems like the ceiling in the long run. Look at Candelaria, who stranded 88.8 percent of runners in 1977. His career strand rate is a much more average-looking 74.1 percent. In fact, he never topped 77.4 percent again in his career.
It is more appropriate to recognize that, in the long run, the pitcher will regress to the mean. To which mean? That is a great question. You can’t just assume pitchers will regress to the league average; clearly, as is the case with Ford and Koufax (and Navarro as well), some pitchers can sustain a higher (or lower) strand rate over several years.
So, as is the case with many statistics, if you have several years of data about a pitcher’s LOB%, you will want to regress the pitcher towards his career mean. It does not make sense to predict that a high-strikeout pitcher like Kershaw will regress towards a league-average strand rate next year. We have plenty of evidence to indicate he is better than that.
But, if you don’t know much about a pitcher, say he has been in the league for only a year or two, that is when you can safely predict a regression to the league mean. Remember that the word “regression” sounds pejorative but is not; it can occur in the positive direction as well. If a first-year starter strands only 68 percent of runners, you would expect him to strand closer to 73 percent the following year.
More from Statliners
Note I said above that strand rate is a pitching/defense stat. This was intentional. LOB% is reported for pitchers but it really involves the entire defense because the term H for hits is used, and team defense plays a role in how many hits a pitcher “surrenders”. This is the same concept behind BABIP. So, if a pitcher plays in front of a good or bad defense, or is traded from one type of team to another, then his LOB% will be affected.
One last word. LOB% is just an indicator built on components. It doesn’t mean much to say “Pitcher X should increase his LOB%”. Yes, but, what are you really saying? You’re saying that he should allow fewer runners to score or that he should allow fewer runners on base. Well, great, that is what every pitcher wants to do.
It is more useful to say the pitcher should walk fewer batters, or strike more batters out, or pitch up in the zone to induce fly balls (which fall for hits less often), or that his defenders should pay attention instead of thinking about which bar they are going to hit after the game. It is also more useful to notice that a pitcher’s strand rate is out of line with what you would expect it to be and dig deeper for an explanation of why that might be and whether you expect it to change.
All data for this post is from FanGraphs.
Want an explanation of a stat? Click on it the first time it appears in this article or view our Saber Glossary.