Comparing marathon training plans with real (Strava) training data

Hi All,

You might be interested in this analysis I did, where I downloaded the training data for 1,000 strava athletes in the lead up to VLM 2016. I then compared their training patterns to a number of Runner's World marathon plans:

http://blog.scottlogic.com/2017/02/28/london-marathon-training-visualisation.html

Interestingly people are putting in far fewer miles than the plans suggest. Those achieving a sub 4:00 marathon time are running 34% fewer miles than the Runner's World plan!

- Colin E.
Tagged:

Comments

  • YnnecYnnec ✭✭✭
    That's a really interesting analysis, Colin. Thanks for posting.
  • Hi All,

    You might be interested in this analysis I did, where I downloaded the training data for 1,000 strava athletes in the lead up to VLM 2016. I then compared their training patterns to a number of Runner's World marathon plans:

    http://blog.scottlogic.com/2017/02/28/london-marathon-training-visualisation.html

    Interestingly people are putting in far fewer miles than the plans suggest. Those achieving a sub 4:00 marathon time are running 34% fewer miles than the Runner's World plan!

    - Colin E.

    Or, they're only logging 66% of their mileage on Strava.  You're drawing conclusions from incomplete data.
  • Pfft - everyone knows that if it's not on strava - it didn't happen.... ;-)

    (i know all my stuff is on strava - I'd assume it's the same for most strava users ? Even if you had a run that failed - you'd add it on manually ?)

    I like this from the notes : 

    "Interestingly you can see that the 2:30 - 3:00 finishers are the only ones where their average training pace is slower than their marathon pace."

    So basically everyone else is racing everywhere in training and race slower on the day. Which ties in with an awful lot of posts on here ?
  • HA77HA77 ✭✭✭
    I thought about that Dave. You could be right but I think most people on Strava log all their runs, even putting in manual entries if they don't have GPS data for a run or even for a section of a run where their GPS stuffed up. That's what most people I know are like anyway. My thoughts are that the number of runs not logged are probably small and won't make a huge difference.
  • HA77HA77 ✭✭✭
    Cross post cougie. Stole my thunder again.
  • That's spooky HA77 - maybe we're the same person ? 
  • "Interestingly you can see that the 2:30 - 3:00 finishers are the only ones where their average training pace is slower than their marathon pace."

    Not very surprising really - I would guess that most sub 3 finishers are pretty close to their target time.  The results for slower groups will be skewed by people who were targeting a faster run but crashed and burned on the day.  If you could plot training pace vs race pace for people who were aiming for (say) 4:00 to 4:30 and actually finished in that range, you'd see a more conventional result of training being slower than race pace.

    I would agree that most people using strava will log most of their runs on it.  I just took objection to the OP making a sweeping statement without acknowledging that he's not got all the data.

  • HA77HA77 ✭✭✭
    Thanks for that Colin by the way. I found it interesting. Very geeky, but interesting.
  • > @"Cheerful Dave" said:

    > Or, they're only logging 66% of their mileage on Strava.  You're drawing conclusions from incomplete data.

    You're right, I've got no way of proving that for sure. However, almost every Strava user I know is very particular about logging every last mile. We have a little mantra among my running friends, "If it's not on Strava, it didn't happen".

    It was also funny seeing the panic a couple of days ago when the Amazon outage meant Strava was out of action!

    Colin E.
  • > @cougie said:
    > "Interestingly you can see that the 2:30 - 3:00 finishers are the only ones where their average training pace is slower than their marathon pace."
    >
    > So basically everyone else is racing everywhere in training and race slower on the day. Which ties in with an awful lot of posts on here ?

    That's one way of looking at it, and you might be right. Although, another perspective is that the faster (2:30-3:00) finishers are the only ones logging enough miles that they need to slow down a bit for some of their runs to avoid damage.

    When I ran my very first marathon I was struggling to find enough time to put in the miles. As a result I tended to run faster than the plans suggested just to make better use of the time. I only did about 20 miles a week, which meant I could do quite a lot of my running at a tempo pace.

    Colin E.
  • > @"Cheerful Dave" said:
    > If you could plot training pace vs race pace for people who were aiming for (say) 4:00 to 4:30 and actually finished in that range, you'd see a more conventional result of training being slower than race pace.

    I would love to know a bit more about the aims and goals of these athletes - but unfortunately that is not possible!

    > I would agree that most people using strava will log most of their runs on it.  I just took objection to the OP making a sweeping statement without acknowledging that he's not got all the data.

    Yeah, you make a fair point. I should have acknowledged that I might not have all the data.

    Although conversely, you make a statement above "you'd see a more conventional result of training being slower than race pace" - without any data at all ;-)

    Colin E.
  • You're right Colin, I did, and I guess I knew it was a bit hypocritical at the time!  Seems logical though, that if you could analyse a subset of people who finished within (say) 10 minutes of target time that training paces would be slower than the strava data suggests. 

    It would be interesting if you took the halfway timings for VLM finishers and compared your data with those - times at 13.1 are probably a better reflection of target pace than finish times.  Most of us make it to at least half way before blowing up! 

  • That's SO true CD ! 

    That's why when I become President of the World - the marathon will be set at a sensible distance - 20 miles.  Avoids a lot of messiness that way. 
    Vote for me !
  • Vote for Cougie.... and his 5k run to the corner of the street and back
    lets be honest.... its all prep for an Ironman on my 100th birthday
  • You KNOW it makes sense. 
  • NickW2NickW2 ✭✭✭
    Thanks Colin, that's an interesting analysis.

    Out of curiosity I just looked up my own data to see how I fit into it - to give some background I was new to running in October 2015, having got a place in the ballot for London 2016, but had decent base fitness from playing squash approx twice a week - I did a (in hindsight probably ill-advised) half marathon in November 2015 in 2:05 off 3 weeks running.

    In the 16 weeks culminating in, but not including, the marathon, I ran 338.3 miles across 40 runs, with an average pace of 8:49 per mile. My target for London was sub-4, which I just achieved, with a 1:57:20 first half and 2:02:32 second half.

    Interestingly, this mileage puts me pretty damn close to the average 3:45-3:59 runner in your analysis (total 342 miles), despite as you say being a fair bit less than recommended by some plans.

    I ran approx 50% (by mileage) of my training runs at faster than MP, 25% within 10 secs/mile of MP, and 25% slower than MP, with longest runs of 20 (x1), 19 (x1), 16 (x1), 15 (x4). Weekly mileage was 21,26,14,28,2 (injured), 22, 11, 31 ,32, 25, 40, 19, 22, 15, 24, 9 (plus marathon).
  • god I love data geeks! 

    (takes one to know one btw)
    lets be honest.... its all prep for an Ironman on my 100th birthday
  • My 16 weeks last year  - 342 miles over 35 runs. Average pace 9 mins 20. 
    MP turned out to be 7.49 - not as good as I'd hoped.  My legs did fall off a bit. 

    This years going a bit better.
  • > It would be interesting if you took the halfway timings for VLM finishers and compared your data with those - times at 13.1 are probably a better reflection of target pace than finish times. Most of us make it to at least half way before blowing up!

    In my previous analysis I plotted the half marathon split times for various pace bands:

    https://bl.ocks.org/ColinEberhardt/1da0f37fb51146d6a3600dc5c53d7b76

    So I guess what you are asking is, if I took one of these groups, say 3-4 hour finishers, I could split the athletes into positive and negative splits, then look at the weekly mileage for each of those groups?

    I might give that a go tonight ...
  • edited March 2017
    My 16 weeks were 458 miles over 70 runs. Average pace 8 mins 23. 
    MP done at 8 mins 59, as I took it really steady, as I was building up to my A race, Ironman Lanazarote a few weeks later.

    This year going great so far... slower, further and feeling better!
    lets be honest.... its all prep for an Ironman on my 100th birthday
  • 16 weeks before VLM last year were 635 miles, average 6'57" pace.  7 long runs in that time (that's runs of 20 miles or more, including one on NYD which was a few days before the first of the 16 weeks) average 7'24" pace.  Plus a few unlogged miles which would have been slower recovery runs.  MP on the day 6'05".

    "So I guess what you are asking is, if I took one of these groups, say 3-4 hour finishers, I could split the athletes into positive and negative splits, then look at the weekly mileage for each of those groups?"

    I wasn't thinking of that, although it would be interesting (a wild guess: low mileage runners are more likely to have a big positive split than higher mileage ones!).  I was just thinking that the average pace to half way would be an overall better indicator of target pace than the finish time, so you'd get a better idea of whether people are training above or below their target MP.

  • NickW2NickW2 ✭✭✭
    @Magna CarterI wonder how many treated London as a training run like you? Probably not enough to be statistically significant.
Sign In or Register to comment.