For the past two months I’ve been running with a new toy, a GPS trainer that tracks (almost) exactly where I’ve run and how fast I ran. Lately I’ve been improving a whole lot, which has been extremely gratifying. But on this morning’s run I took a step backward, running slower than I have for weeks. What’s up with that?
Then I noticed on Facebook that several of my running acquaintances were reporting the same sluggish pace today. It was quite warm and muggy this morning. Could it be that all my “improvements” over that past month have really been due to cooler weather? Since my GPS device records every run, I thought I’d compare my Wednesday morning run pace (over about six miles) to the weather. This graph shows the results:
The blue line shows my pace in minutes per mile — so lower numbers are better. Green columns show the high temperature for that day, while gold columns show the previous day’s low, which I think comes closest to approximating the temperature at run-time, 6:00 a.m. As you can see, starting October 6, I improved dramatically, shaving almost a half-minute per mile off my pace. For a runner, this is huge! Then I continued that improvement on the 13th and the 20th, before regressing back to roughly my October 6 pace today. The pace changes don’t seem to bear much relationship to the high temps — my slowest pace was on the coolest day. But low temps tell a different story — my first dramatic improvement came on a day where the low was 19 degrees cooler than the previous week. My regression today came after a 15-degree rise from the low temp last week.
I also computed correlation coefficients for my run pace and temperature. The correlation between pace and high temp was 0.35, while the correlation between pace and low temp was a whopping 0.58!
Is this statistically significant? Nope — my N of just 8 means that p = 0.13. But it’s a fairly dramatic trend. However, since Wednesdays are the only days I run 6 miles, it would be hard to get a larger sample. Maybe I’ll reassess when things start warming up again in the spring.
Update: There have been a few requests on Twitter / Facebook to include more weather variables. I’ve heard dew point is a better proxy for humidity than “relative humidity,” so here’s a graph showing dew point instead of high temperature:
Just taking a look at the graph, dew point does not appear to explain the variance in my pace as well as low temperature. Indeed, the correlation coefficient for pace vs. dew point was 0.39 — slightly better than the high-temperature number, but not nearly as good as the low-temp correlation. And of course, like the other correlations, it’s not significant.