What to make of the Yankees’ RISP woes

Batting average with runners in scoring position is one of the first metrics trumpeted when discussing a team’s offense or the nebulous notion of “clutch.” The common belief among many is that teams who can maintain a high average with runners in scoring position will be able to maintain a better offense and win more games.

Despite the Yankees’ 11-7 start, buoyed by what is currently the second-best offense in baseball, there have been rumblings among New York Yankee fans that such a good start is unsustainable because of the team’s .221 average with runners in scoring position, 30 points below the league average .251 mark.

I thought I’d test this prevailing theory that average with runners in scoring position is correlated to wins or offensive performance, and started by plotting the correlation between the 30 MLB team’s 2017 winning percentage and their RISP:

This appears to be a pretty strong correlation, and it is! The correlation coefficient for these arrays is .694, which would indicate a relatively strong, positive correlation. That is, as a team’s RISP increases, so should their winning percentage.

This isn’t the whole story, however. No team has played more than 22 games, and because opportunities with runners in scoring position don’t occur all the time, most data sets are vulnerable to swings because of small sample sizes. To remedy this, I plotted the same correlation for the entire 2016 MLB season:

That looks slightly different. The apparent positive correlation disappears, with a correlation coefficient of only .121. The RISP scale (x-axis) also shrinks, showing that the data points tend to centralize given a larger sample size. Over an entire season, a team’s ability or lack of ability to hit with runners in scoring position doesn’t have that much of an effect on their ability to win games.

However, one full season still isn’t a great indicator of sample size, and so I took the data from both 2016 and 2015, averaged it, and plotted that:

The correlation coefficient is almost halved, dropping to .063. There is now almost no correlation between a team’s RISP and their winning percentage. As sample size increases, the inferences made about RISP and its affect on winning percentage begin to fall apart.

So we can accept that RISP doesn’t have much of an affect on a team’s ability to win games, but that might be because we’re completely ignoring the role of pitching. When the Yankees send out Masahiro Tanaka to pitch, generally the team only needs three or four runs to win. Knowing this, it’s appropriate to see if RISP has any affect on the overall offensive production of an MLB team.

Following the same pattern as before, I plotted a team’s RISP against their 2017 wRC+. For those who don’t know, wRC+ is probably the best single measure of a player or team’s offensive impact. The metric is normalized after accounting for park factors, meaning a wRC+ of 100 is league average production. Each mark above or below 100 is 1% better than league average. For example, Mike Trout led baseball in 2016 with a 171 wRC+ (71% better than league average), while Chase Headley’s 184 wRC+ currently is best on the Yankees.

The plot between RISP and wRC+ in 2017 came out like this:

Shockingly similar to the initial winning percentage plot, we see that there is a positive correlation between a team’s RISP and wRC+. The correlation coefficient is 0.427, so not quite as strong as the very first plot, but still noticeable.

How does this relationship hold up with submitted to the same sample sizes as RISP and winning percentage? Below is the plot of RISP and wRC+ for all of 2016:

Again, the similarity to RISP and winning percentage should be obvious. The correlation coefficient falls to .255.

For the same two-year average as before:

There we go: an almost identical pattern to the first round of RISP and winning percentage. For this relationship, the correlation coefficient drops to .095, indicating again, virtually no correlation between RISP and a team’s overall offensive production.

This presents, in my opinion, a perfect long run vs. short run dilemma. When analyzing small sets of games, a team’s RISP might be indicative of how many wins they have. This could be because the team getting the one “big hit” can help win a close game they might have otherwise lost by a run or two. However, when looking at an entire season or more, it becomes obvious that RISP has nothing to do with a team’s chances of winning, or its overall offensive ability.

Data courtesy of FanGraphs and ESPN