Analyzing training plans with a Marathon Shape workbook

Based upon Runalyze‘s Marathon Shape metric, I created a workbook for myself to analyze the projected Marathon Shape for a runner with a given Estimated VO2max, based on the projected mileage from a given training schedule.

Realizing I’ve briefly and vaguely brought this up before, I should first go into some detail on Marathon Shape and why I care about it:

Runalyze (having shared this information in their Forums) calculates Marathon Shape based upon:

  • All mileage from workouts in the previous 182 days (26 weeks).
  • The distance of any run from the previous 70 days (10 weeks) that is longer than 13km, scored on a base 1.0 scale.
  • The runner’s Estimated VO2max score (EVO2).
  • 2/3 of the Shape is weighed by the total 26wk mileage, and 1/3 is based on the percentage of the total score of any 10wk long runs (13km+) against a 10.0 point standard.
  • Shape is scored on a 100% scale, with 100% indicating you can hit your EVO2-projected goal time, and you can exceed 100%. While you can obviously run a marathon at less than 100% Shape, your performance will not match your EVO2’s potential projection, or most projections for that matter.
  • While some parameters can be adjusted (time periods, the 13km long run standard), Runalyze generally recommends they be left as-is unless you have a clear compelling reason to change them. So I leave them as-is.

Basically, the Laufcampus research behind Runalyze found that marathon performance was a function of runners’ total training mileage within the last 6 months, as well as the distance and frequency of their long runs in the final 10 weeks.

Like most training app platforms, Runalyze projects expected race times based on your EVO2 and fitness. At lower distances, you don’t need to run much to be able to reach your optimal race times. As the race distance increases, the needed volume increases. Obviously, the marathon is a much different beast as you’re hitting the back wall of your glycogen reserves (emphasis on Wall) and a certain amount of training volume is needed for your body to develop the fitness to handle this, in-race fueling or not.

In my experience, the Shape metric (in hindsight from its calcuation of past data) has been remarkably accurate in projecting my ability to handle a given marathon, as well as that of others. I quickly grew to trust it and implicitly trust the number today, for myself and for anybody else.

Your required weekly mileage over 26 weeks to get 100% Shape is a function of your EVO2, and it loosely corresponds to your EVO2… e.g. an EVO2 of 37 generally means you need to run an average of 37 miles per week. As EVO2 increases or decreases, the threshold of needed mileage and long runs increases or decreases as well. What would be 100% at an EVO2 of 37 probably becomes 90% at an EVO2 of 40. An elite runner with an EVO2 of 70 might not even be 48% ready at the same mileage.

Your long runs within the last 10 weeks are scored exponentially on a base 1.0 scale against the formula’s projected required weekly long run, e.g. at 37 EVO2max, this is around 17 miles. That required long run distance goes up or down as the EVO2 goes up or down, the idea being (like the aforementioned mileage total) that better runners looking to run faster marathons need to handle more distance on their long runs, and the longer they can capably go in training the better.

(I’m sure the Hansons of the world would disagree, but Laufcampus has a variety of redundant research and data behind this and all their calculations.)

The objective is to possess 10.0 or more total Long Run points on race day, and the value of every long run degrades with each passing day. Shape scales the long runs exponentially not just based on distance (at 37 EVO2, five weeks out: a 9mi run scores almost nothing, a 14mi run might be worth 0.5pts, the average 17mi run might be worth 1.0, a 20mi run might be worth 2.5pts)… but on recency: A(n ill advised) 20 mile run the day before the race by a 37.0 EVO2 runner is worth 3.8 points, but the exact same run done 10 weeks ago is only worth about 0.1 points!

Basically, you want valuable long runs closer to race day, while still bearing in mind the need to taper (please don’t run 20 miles the day before your marathon!) and thus not run long during those taper days, as well as step-back weeks during the final 10 weeks, where those shorter long runs won’t count for very much.

Side note: The easy 201 level answer to Shape for a 37 EVO2 would be to run 17 miles every weekend, including the week before the race (not the best idea). The more advanced and probably best answer is to plan to do a few longer long runs (more like 18-20 miles) most weeks or every couple weeks prior to taper, knowing you’ll have some shorter ones counted during down weeks and during taper.

It sounds complicated, but all of this can be calculated automatically from your training data with Excel functions in a well crafted workbook, let alone instantly through code on a site like Runalyze.

While Runalyze tracks your current Marathon Shape in their dashboard, they don’t have an available function for you to project shape based on an anticipated training plan. This is understandably difficult to create within the standard mainframe code, a key reason other app brands like Garmin and Coros haven’t done it yet.

But with an aforementioned well crafted Excel workbook, there is a reasonable way….

I started seriously running late in life. I’m certainly no elite or local podium runner, but I’m also not a rank beginner just hoping to cross the finish line in one piece with some dignity. I have enough experience and knowledge to know what a given plan will give me or someone else, and thus can calculate what kind of Shape one can expect out of it.

As I’ve shown on this weblog, I like reviewing and analyzing different marathon training plans, seeing how they work and figuring how they’ll work for different people and fitness levels.

At some recent point, I decided to start calculating the mileage (or likely mileage, with time-based training plans, e.g. “run 45 minutes” instead of “run 5 miles”) of various plans against the Marathon Shape metric to analyze what kind of Shape I could expect them to give me or someone at my fitness level. I crafted a simple single-sheet Excel workbook that would let me do this somewhat quickly, in as simple and direct a way as possible.

(The workbook sample shown in the pic takes a typical 30 mile training week with an EVO2 of 37, across 25 weeks before no running on race week. Not the best marathon training approach, obviously!)

The Marathon Shape workbook I created has a no-frills setup, laying out each day of a 26 week period prior to a marathon in its own row, by week number and day number. For each given day in “Miles”, I enter the day’s mileage as a single number. The cumulative Shape is calculated after each day, based upon the VO2Max figure entered in the Header column.

The Long Run quotient is shown for all cells, but is only calculated for daily mileage in the final 70 days, weighed based on the value in the corresponding Days To Race column. The Shape equation factors this in addition to the cumulative mileage.

Once you get to the bottom of the sheet, race day is indicated by “26.2”, along with the Long (Run) total. The prior day’s calculated Shape is your total projected Marathon Shape for that race.

Obviously, one limitation of this basic workbook is that I can only list one mileage total for each day. With Marathon Shape, if you run 14 miles in a day but between two separate 7 mile runs, Runalyze gives no Long Run Shape credit because neither single run met the 13km (8.08mi) threshold, even though you ran 14 miles. As most runners would tell you, the uninterrupted duration of a long run is a key component of its marathon training value, so the metric accounts for that.

To account for this (bear in mind no plan I analyze with this has any less than 1 day off per week), I will “spill over” the shorter run into one of the nearby off days. For example, a plan has 13 miles in a weekday that I of course will split up (I can’t feasibly run that long at once on a weekday). I do know I’m going to run 10 miles in the morning, and run the rest later. So I’ll put a 10 in for that day, and put a 3 (the difference) in one of the off days. The long run will be properly calculated, and all the mileage will get counted correctly.

(What if you have more such overlong days than scheduled days off? … well, that’s not a plan I’m going to analyze. Such a plan is probably too much volume for the level of fitness I’m analyzing! And a fitter runner is probably fast enough to do such a distance at once anyway.)

Also, of course, this analysis is very reductive. We’re looking at mileage and distance covered on a daily basis. This ignores the quality of that mileage, any non-running cross training, any side work like strength training or drills, nutrition and recovery, etc. This is hardly specific to this process: Pretty much everybody all but ignores calculating the training impact of any non-running work, including Runalyze themselves. But any contributing factors in either direction are basically ignored, and all mileage is treated the same.

Also, as most platforms do, Runalyze completely ignores the effect of temperature, wind, and altitude on your performance (though they do factor in the effect of hills and other elevation changes). Training at 2300-3000′ in dry, dusty, often hot Las Vegas leads to my performances varying substantially from my same efforts on flat ground at sea level in Vancouver or Chicago (which at altitude of 600′ is closer than Vegas). All are weighed equally, so when you run slower or the same speed at higher heart rate at altitude or in high heat, the calculations downgrade and penalize you. Even though programmers like Davide Liu have shown how to accurately calculate and account for these effects, virtually no platform will adjust for these factors at all.

Side note: I do make one adjustment with cross training. Any Zone 1+ endurance work done on a bike or on my feet, I adjust in Runalyze as a run workout, labeling it as “Cross Aerobic” for reference, with a “composite mileage” total. Having separately calculated my expected pace for my personal EVO2 at given heart rates between 50-75% of max, I take a cross training workout’s duration and heart rate and calculate an equivalent mileage in Runalyze.

While I obviously didn’t run these workouts, my cardiovascular system and much of my relevant running muscles did have to do work at that heart rate for that duration. The calculated equivalent paces typically are way slower than I run, often in the 13:00-15:00/mile range (my heart rate is usually lower than during a run). So the equivalent mileage isn’t as high as it’d be had I run for that duration.

Doing this does put a mileage number on my cross training so Runalyze’s Shape metrics won’t ignore those workouts, plus the equivalent pace is at my close-to-current EVO2 so it doesn’t significantly change that overall calculation. I still find when I go on faster/slower than expected runs that my EVO2 still shifts significantly despite all this cross training data behind it.

You can debate my practice of doing this, but I’ve found doing it improves the accuracy of the EVO2/projection calculations versus before when I’d let Runalyze leave all that cross training out and therefore underestimate my training volume. I’ve also found that my calculated composite mileage/hour is very consistent between all cross training over long periods of time. Usually, an hour long cross training workout is roughly worth about 4-5 miles depending on average heart rate. End side note.

Despite these caveats, I forged ahead over the last few weeks with calculating the Marathon Shape of various plans, while adding in equivalent mileage from any scheduled cross training, assuming an EVO2 of 37.0 (not quite mine, but close!), and pre-plan mileage loosely matching the plan at 30 miles a week (a reasonable expectation for runners in this fitness range).

I’ve since analyzed a couple dozen training plans, most of which I’ve previously written about here on this weblog, along with many others I’ve never covered or referenced here.

Rather than make an overlong post double or triple long, I’ll leave you with a general cliffhanger: I’ll cover the Shape Analysis of individual plans over time going forward, likely in individual posts, before writing a summary post ranking the relevant ones. It wouldn’t be very helpful to just give a single ranked list without some depth of individual explanation, and that’s not going to be reasonable in a single post. I also would leave off several plans as they ultimately aren’t reasonable for someone at this fitness level, or it’s clear in one or more ways that they’re not going to suitably prepare a runner for the marathon as written. (There’s unfortunately a couple plans I’ve written about that I’ve since discovered cannot get someone at this level close to ready for the race, though they can do so for runners at a different ability level.)

Tagged , , , , ,

Leave a Reply

%d bloggers like this: