Interesting article on the (in)accuracy of race time predictors (Read 524 times)

zonykel


    Interesting article. i suspect the wall has something to do with the longer than expected times For the marathon. If we could gather total volume of training, it might give us some insight as to the cause for the deviations.


    Got Hills?

      Interesting numbers, but he sums it up well in the last paragraph.

      A good predictor won't tell you what you will do, but what you probably can do...

      "Not to touch the Earth, not to see the Sun, nothing left to do but run, run, run..."


      Muddling through

        Interesting but if you use that 1.15 exponent on shorter distances, how much does it throw off those times?  One point he notes is that Riegel did his calculations in 1977, a time when marathoners had a much different demographic than they do now. He starts to discuss the relation when speed is considered, but then just compares men and women with equal HM times. I'd like to see the relative accuracy based on marathon times. Is the formula more accurate for faster runners who presumably are also better trained? While he did try to consider experience, the time span over which those 5 HMs and 5 marathons may also be important. The calculators were a lot more accurate with my marathon time prediction after I'd been running 10+ years than after only 3 years.

        2014 Goals: Run first trail ultra, first 100K, and see what I can do in a 24-Hour race

          Sure, if you're woefully under-trained then 1.15 is going to make a better prediction that 1.06. That should be fairly obvious.

          Runners run.


          I'm back!

            Sure, if you're woefully under-trained then 1.15 is going to make a better prediction that 1.06. That should be fairly obvious.

            +1. 1.15 is ridiculous for anyone who has trained appropriately. And the guy seems to want people to use his new, improved calculator. Honestly, I will be pretty annoyed if this is now widely used to justify crappy performances as equivalent.

            GC100k


              He cites a 1.06 exponent.  Most predictors in the past used 1.07.  I tried several times on the RW forum a decade ago to have people calculate what their exponent is based on their PRs, but I never got much interest.  Just recalculated my exponent and it's 1.33, meaning I'm relatively slower at longer distances, even though I trained much much harder and longer for my best marathon than I did for my best 5k.  Maybe it's because I'm so heavy; I don't know.

               

              But like he said in the article, I know very few people who have run a marathon in the time predicted by their 5k.  I'm fat and slow, so don't ridicule my times, but I've run a 5k in 23 minutes without ever training for it (speedwork, etc.) and I trained long and hard to run a 4:30 marathon that I'm mighty proud of.  The 1.07 exponent would predict a 3:46 marathon.  That ain't happening.

               

              We could crowd-source an exponent calculation by having a bunch of people calculate theirs, but like I said, I never got much interest in the past.  By the way, I'm talking about getting a best-fit exponent using data from 5k to marathon, not just a half and a full.

              Longboat


              Letting off steam

                For me, a 1.07 exponent works fairly well for my PR times from 5k to marathon, over three years long ago.  I ran lower mileage then -- I never trained as well for a marathon as I did a half or 10k.

                 

                Looking at my most recent series of races, 1.06 is closer for 5k to half.  Had I followed this series with a marathon in 50* rather than 85*Roll eyes, I may have been a little faster than the 1.06 would have called it.

                Neil

                ---------------------------------------------------------------------------------------------------------

                Nearly back to 100% 6 months after Achilles surgery. Now at 35 50 mpw.

                Base building time!

                  +1. 1.15 is ridiculous for anyone who has trained appropriately. And the guy seems to want people to use his new, improved calculator. Honestly, I will be pretty annoyed if this is now widely used to justify crappy performances as equivalent.

                   

                  Well, maybe you're right, bhearn, but I've found in my personal experience as more of a "speed guy" (don't laugh too much, some of you who really are!) that my exponent definitely gets way whacked at marathon distance, and I think I have definitely trained for the distance appropriately many times.  This week case in point.  On Monday I ran a 2:56:39 giving it what I think was my very best effort for where my fitness is at right now.  Working the problem backwards with 1.06 I should be a 18:25 5k guy, and yet I can almost run that fast without trying.  I ran 17:44 yesterday morning on a course that was *probably* slightly long, in fact, with only 5 days recovery from a major marathon effort, and all alone without anyone to push me the final 2 miles.  So, I am starting to wonder, after 20 marathons now, if one of two things might be true:  1.  I am a pansy marathon runner..?? (maybe!?) or 2.  I'm just not cut out for marathon running the way some guys are and I'm just more able at shorter stuff.  I don't know, but based on my experience I'd estimate my personal exponent between 5k and marathon to be more like 1.1.  Although, I agree that 1.13 or 1.15 as proposed by this chucklehead is way out of bounds.

                  - Joe

                  We are fragile creatures on collision with our judgment day.


                  I'm back!

                    But Joe, you weren't fully trained for Boston, right? I don't think this is a case that's valid for comparison.

                     

                    And that gets at my problem with this guy's methodology. It's great to say, OK, I'm only using times from people with at least five halves and five fulls. But you can't tell just from looking at the times the specifics of those performances. Were they equivalently trained in all cases? Almost certainly not. And that fact pushes the exponent higher, because you suffer a lot more running a marathon not appropriately trained.

                     

                    I think the existing calculators work pretty well for most people, IF they are appropriately trained, AND conditions and course are favorable for the marathon. Those three things don't come together all that frequently.

                     

                    That said -- yes, I'll agree, some people are just better at short stuff, some at longer stuff. One obvious component is slow/fast twitch muscle balance, but there are undoubtedly others. However, within the framework of the exponential formula at issue, my feeling is that 1.15 is far too high for anybody, if the formula is used correctly.

                    noahsd


                      Riegel value of 1.15 is a ridiculously slow predictor.  Running a half and adding 10 minutes for me is much more accurate.  1.15 = running a 1hr 30 min marathon prediction time of  3:19:40??  I have never come close to that (+4 mins) in a half but have in a marathon (-1 min).  

                       

                       

                        But Joe, you weren't fully trained for Boston, right? I don't think this is a case that's valid for comparison.

                         

                        Yes, that's right, my training for Boston was indeed very much sub-optimal & I think I see your point that my lack of quality marathon-specific training was to blame in this case.  I guess I was trying to normalize things by saying, same guy, same week, similar weather conditions, using 1.06 = either I was really a woosie in Boston, or else I was just amazingly incredibly faster than I should have been yesterday morning.

                        - Joe

                        We are fragile creatures on collision with our judgment day.

                        doctorjen


                          Just plugged in my numbers from a half 5 weeks ago and full yesterday and get 1.065 (1:48:13, 3:50:37). I think in my case that says I have reasonable endurance, but not much speed. Now to figure out how to get faster ...

                          I know lots and lots of folks who train for the marathon with low mileage plans, though, for whom 1.15 might be accurate. One friend in particular has a 5K pr within 15 sec of mine but my marathon pr is an hour faster than hers.


                          I'm back!

                             I guess I was trying to normalize things by saying, same guy, same week, similar weather conditions, using 1.06 = either I was really a woosie in Boston, or else I was just amazingly incredibly faster than I should have been yesterday morning.

                             

                            Not at all. Same guy, same week, similar weather conditions, but adequately trained for 5k, not adequately (or less adequately) trained for marathon. Even if you were to say you were equally inadequately trained for each distance, I still wouldn't expect the formula to work; it's predicated on you being adequately trained for both distances. Failure modes for 5k and for marathon are very different.

                              Failure modes for 5k and for marathon are very different.

                               

                              Ha!  This much is definitely true!

                              - Joe

                              We are fragile creatures on collision with our judgment day.