12

Race time predictors, revisited (Read 200 times)

     

    It's probably doable. The margin of error may need to be increased to account for the larger variability of the terrain, though. Seems like a probability and statistics problem, not a running problem.

     

    Not without data on each course to match to the time. The problem is that the races for each distance aren't the same, as well as different distances. They're mostly hilly trails (but not all), some are better surface than others..... so on and so on.

    2013 Goal: Make 3:00:16 go away - FAIL.

    2014 Goal: Make 3:00:16 go away.

       

      It's probably doable. The margin of error may need to be increased to account for the larger variability of the terrain, though. Seems like a probability and statistics problem, not a running problem.

      Agree with Viich.

       

      This is really more of an engineering / physics issue for the hill aspects (see the tm incline tables) and a coefficient of slowness (for lack of better term) to account for technical issues of the trails and snow conditions - or a piecewise fit for each section. The coefficient will be runner specific since some people are better at uphills and others at downhills. Probability comes into play with the weather, which also affects snow and mud conditions.

       

      A couple of my races are 2000ft uphill in about 2.4 mi, not rough footing, only uphill. One 26.2 mi loop has about 3500ft of uphill with the bulk of it in an 1800ft hill and contains both road and trail. (when looking at mile splits of the faster runners on this course, there's a 50% variation depending on which mile) A 10k loop has about 800ft of uphill with the middle half having a bunch of roots and rocks (as technical as I run). And this year, some of that course will probably be under snow. The winter races are usually on frozen swamps, but snow conditions - packed or like loose sand - make a huge difference in time. On longer races, I'll piece together the various hills and footing segments and add them all up to get an idea of how long to expect - mostly in order to plan fluids, electrolytes, fuel, pack / waistbelt, extra gear, whatever.

       

      Trying to come up with a predictive formula is just too much hassle.

       

      MTA: For perspective, my avatar is from the crest (looking downhill toward finish) of a HM that has about 3000ft up in the first 4mi. There's actually some runners in that picture - specks off in the distance.

      "So many people get stuck in the routine of life that their dreams waste away. This is about living the dream." - Cave Dog

        This is all getting to what is missing from most/all calculators:  the uncertainty of the estimate.  Riegel's formula was based on world record times, meaning 1 time for 1 distance, not even the same runner. It is an estimate of the best possible outcome only. Using data across many runners, it is difficult to separate out the people that went out too fast and had a death march to the finish in the marathon times, for example.  This massively long tail contributes a great deal of positive skew.

         

        Why not have a calculator that gives a range of performances?  Such as,

         

        Optimal Race Estimate:  hh:mm:ss  (using 1.07)

        Realistic Race Estimate:  hh:mm:ss (using 1.1)

        Pessimistic Race Estimate:  hh:mm:ss (using 1.15 or higher)

         

        Optimal Race Estimate is the goal time for a goal race on a fast course in ideal conditions given proper training.  Someone trying to set a contemporary PR at Boston, Berlin, Chicago, etc. would use this.

        Realistic Race Estimate is the time for e.g. a tune-up race, or a goal race on a course with some added difficulty such as hot weather.

        Pessimistic Race Estimate is the time where there are major factors that preclude an ideal effort, such as lack of training.  The average race falls into this category.

        2013 H1:  7 hours/week base.  Q3: Train for goal race.  Q4:  Goal Race.

          And then we get back to the place where people with enough data in their running logs to make these kinds of predictors actually work, already know pretty much what they should be capable of.

          Runners run.

          12