Is WMATA really that bad? This data scientist says Metro is unfairly vilified

Maybe it’s good that Keith Kelly didn’t have much of a bias against the Metro when he started digging into WMATA’s new real-time train position API. “I don’t really ride it much,” he shrugs. So how does he get around, you ask? “Well, I like to walk…”
The real reason, though, is that Kelly is new to D.C. He moved here just six weeks ago, from Austin, Texas, after graduating with a master’s in computational physics. His girlfriend lives in the District.
For such a new addition to the city, though, Kelly very quickly threw himself into the middle of so many D.C. residents’ favorite conversation and complaint topic — our trials and tribulations with WMATA.
In a blog post published at the end of August, Kelly used WMATA’s train position API and his data science skills to try to assess just how bad metro wait times are. The inspiration came from similar work done on NYC subway wait times. And what he found, well, it might not be what you’d expect.
So how’d this happen? Where did his Metro Math blog post come from?

Kelly started his eponymous blog back in May when the domain became available, thinking it’d be a good place to build a portfolio. “I wanted a way to showcase some of my skills,” he said. Given his recent graduation and move Kelly has been doing a lot of job hunting in the data science realm, so he view the blog as “a good way to practice what I hope to be employed doing.”
Kelly’s done a number of other posts to the blog, including ones analyzing Capital Bikeshare data, but none struck quite the same nerve as the Metro Math post.
“Transportation is an important issue for cities,” he nods, understandingly.
So what happened with the Metro Math post? Well, in short, Kelly found that “wait times are really not that bad.”
“The metro is really heavily criticized here, and what I found is that it maybe isn’t that deserving of all that criticism,” Kelly said. Yes, it’s true that wait times are longer at night or on the weekends or during the middle of the day when less trains run. His analysis also doesn’t take into account situations where trains are full and that makes de facto wait times longer. But in general, Kelly’s analysis found, the system works pretty well. Take that, metro haters.
Kelly certainly isn’t wedded to any outcome, though — he just enjoys playing with the data. Got an idea for a dimension he could add to the metro investigation, or another data-driven question entirely? He’d love to hear about it.