Big Data Reveals New Marathon Training Approaches

New big-math research into GPS data has, for the first time, made significant contributions to our understanding of efficient training. While much remains to be learned, the big-data approach has revealed previously-unrecognized links between training plans and marathon performance. This is especially true for midpack and slower runners.

Historically, runners have been able to read about Eliud Kipchoge’s training, and that of other elites. The training of Olympians is frequently covered and glorified in media reports. But how does this help if you need a sub-3:35 to qualify for Boston?

That information actually exists in large digital databases maintained by companies like Strava, but it hasn’t been easy to access or interpret. That’s where the big-data folks are stepping in. They’ve been analyzing the training logs of thousands of marathon runners to look for successful patterns.

Last week my friend Rick Lovett wrote a PodiumRunner article that focused mostly about the “predictive” power of new big-data equations published in Nature Communications, a highly-regarded journal. I’ve been corresponding with the author of that paper, Paris-based Thorsten Emig, about the training implications of his work.

Another group, led by the University of Dublin’s Barry Smyth, has been following a similar path in recent years. Smyth often works with “recommender systems” like the one Netflix uses to predict what movies you might enjoy — or what works for you. In running, another name for this is “coach.” This group has unlocked some training = success patterns I’ve never seen before. I doubt you have, either.

Emig has worked with aggregated data from Polar, and Smyth with Strava data.

I’m going to leave out the math (there’s a lot of it), and just give you the most useful findings, first from Emig’s paper, then from Smyth’s recent work. Emig reaches four key conclusions that can help guide your marathon buildup. Smyth has one significant take-away.

1) Training more, even at a slow pace, can make you faster.

This is a universally recognized result of training, though rarely supported with hard data of the kind Emig has uncovered. It explains why running more, within reason, normally pays off.

2) Fast training builds your endurance more effectively than slow training.

This is why almost everyone does some amount of speedwork. If you can boost your performance at short distances, you can maintain a relatively faster pace over the longer distances required by endurance races. Achieving the right balance between 1) and 2) is what training is all about. Also, recognize that speedwork carries more risks than slow running.

3) Elite runners generally don’t push as hard in training as midpack and slower runners.

When elite athletes cover 100 miles a week, they have no choice but to run most of those miles somewhat slowly. When you run 20 to 40 miles a week, you can run those miles harder relative to your ability.

4) There’s a limit to how far and hard you can train.

Beyond that limit, your efforts turn south. You become over-trained, fatigued, and slower. Emig found a limit at 27,000 TRIMPS (“training impulses”) during marathon training. This number won’t help you much, because it’s not easy to equate to your own training. But you’d better understand that there’s a limit: More is not always better.

5) Adopt a new pattern for your training weeks

In his recent work, Smyth (along with Jakim Berndsen and Aonghous Lawlor) has looked at how to put together a marathon-training program. Their advice, greatly simplified:

1) Focus on training weeks, not individual workouts, and

2) Alternate your weeks in this month-long manner: hard-hard-easy-moderate. In the weeks prior to race-day, taper in this manner: hard-easy-easy.

Here’s what the Irish researchers mean by hard or easy:

• A hard week is one that includes more fast training than normal. For example, it might include a tempo workout and/or a workout where you run as fast as your 5K race pace. You’ll get a substantial training stimulus from this kind of week, but also build fatigue.

• An easy week is one where you do 50 to 67 percent less training than a hard week, but at the same pace as some of your previous week’s training. You’ll recover, and consolidate your fitness gains.

• A moderate week is anything between hard and easy. You don’t get either fitter or more fatigued.

The important findings here are the repeated hard-hard pattern, and the degree of cutback during an easy week. Most current marathon training programs include cutback weeks, but few if any cut as deep as 50 to 67 percent, while emphasizing the pace of those runs.

Even though most plans don’t call for such a pattern, according to Smyth’s analysis, about 40 percent of marathon runners come close enough to his proposed outline that any change would be unlikely to produce a faster marathon time. On the other hand, 47 percent are under-training and 14 percent are overtraining. This 60 percent of marathon trainers might improve by following his suggestions.

None of the big-data experts believe they have found all the answers. Many questions remain. Smyth et al employ the verb “nudge” to describe an appropriate use of their results. If you can nudge your training in the right direction, it could pay big dividends.