The Steamtown Marathon has been on my shortlist for a while. This race has a great reputation. It’s often included on lists of races with a high percentage of Boston Marathon qualifiers. In 2019 about 17% of finishers qualified for Boston. Many reviewers mention that the Race Director sends memorable emails. If their Facebook page is any indication this is definitely true.
This race has an interesting elevation profile. It’s a point to point race (you end somewhere very different from where you start), and there is a net decrease in elevation of almost 1,000 feet. A mostly downhill race sounds great in theory; running downhill is easier, right? Running downhill for a few miles feels better than running uphill, but running hard downhill for a lot of miles is going to tire your quads.
As you can see, findyourmarathon.com has fantastic elevation charts and this one of Steamtown shows where you might slow down.
The race website ominously warns “The significant down hills in the first eight miles tend to beat up your quads making it a challenge to tackle the three notable up hills in the last three miles of the course. The up hills aren’t terribly long (one to four blocks each) or terribly steep, but their locations in the last few miles can take their toll on even the most experienced runners.” At the end of a marathon any incline makes me feel like I’m trying to climb Everest. I wondered how many people slow down in this race. So I turned to the data!
The Steamtown Marathon makes their race results available in a couple of formats, and includes a split time for runners at approximately 18 miles. I was able to calculate the pace of runners up to that point, and their pace for the last 6.2 miles of the race, compare them, and rank runners according to the change in their pace. This approach makes good use of the available data, but it does have a couple drawbacks. I can’t really understand how much runners changed their pace during the entire race, I can only compare those two paces. Looking at the elevation profile though, the 18 mile mark is a good place to look for a change in pace. In my experience, after about 18 miles I start to wonder why I thought it was a good idea to go for a 26.2 mile run.
Many a wise runner has advised marathoner to not start a race too fast, and to try to run an even pace. It is a lofty goal to run negative splits (to increase pace during the race).
In the 2019 Steamtown Marathon, no one ran negative splits. Jordan Greenberg (bib #407) ranked first in change of pace, running almost even splits. Looking at his previous marathon times this race was also a big improvement in overall time. Congratulations Jordan! Second in least change in pace was Cara Ann Frankosky (bib #353). Cara Ann has previously run as an official pacer during other marathons, and is obviously talented at keeping an even pace.
I’ll be keeping the Steamtown Marathon on my shortlist. Understanding the trends in pacing will help me plan my pacing accordingly.
The method I used to analyze the race data at scale was to download the results from 2019 in the available .csv file. I uploaded the file into Jupyter Notebook, and tackled it with Python (huge mentoring credit to the Best Husband Ever). I calculated the pace per mile for each racer to the 18 mile split, from the 18 mile split to the finish, and the change in those two paces. Then I sorted the field according to change in pace. I used data frames and was thrilled to find how easy it is to handle times. Normally runner math can perplex me. To calculate pace for an 18 mile run in 3:10:00, you would have to convert the hours to minutes, carry the 10, divide by 60 and so on. In data frames you can just divide by 18.
I posted the Jupyter Notebook in GitHub and in Binder. You can run the Jupyter Notebook and download the results. If you have never used GitHub or Binder follow these directions to take a look at the results.