Ultra Running and Science: How Do They Blend?

Esa Nurkka, 23.4.2015

University of Zurich's Institute for Primary Care and Gesundheitszentrum St. Gallen have published “Pacing strategy in male elite and age group 100 km ultra-marathoners” report, which analyzes Biel 100 km race pacing strategies over time period 2000-2009. Despite the promising title, the content of the research is a letdown. The report is based on data from the race website, and it seems that, before moving into statistical analysis and conclusions, a bit more background research would have been in order.

In this review I have chosen some questionable issues from the report (direct quotes from the report are in italic red color) and comment on them.

Pacing strategy in male elite and age group 100 km ultra-marathoners (Knechtle-Rosemann-Zingg-Stiefel-Rüst, 2015)
The aim of the study was to investigate running speed over segments in male elite and age group 100 km ultra-marathoners competing in the 100 km Lauf Biel in Switzerland in 2000-2009. The dataset for the study was obtained from the official race website in January 2015, and running speeds for athletes were calculated from split time information from the three Time Stations along the route.

The method sounds appropriate, but unfortunately it seems like something went wrong.

 A quote from the report:
“Split times at three time stations (TS) TS1 “Oberramsern” (38 km), TS2 “Kirchberg” (56.1 km), and TS3 “Bibern” (76.7 km) were taken identically during the 2000–2009 period by using an electronic chip system… Since 1999, split and overall race times have been recorded by Datasport in the same manner.” [source: Pacing strategy…, page 2]

My comment:
Could it be that although the data has been recorded “identically using an electronic chip system” and “in the same manner”, the location of the chip mats has not been identical over the years?

 A quote from the report:
“There seems to be a trend towards even pacing after 2005, with more negative splits at the final segment before 2005." [source: Pacing strategy…, page 6]

My comment:
1. This is not a trend. As you can see in Figure 3, this is a sudden and inexplicable change in the data.
2. This did not happen “after 2005”, the change happened after 2003.
3. This kind of sudden change cannot be explained by “changes in training theory or trends” or "some ‘running guru’ advertising towards ‘putting some time in the bank’" or “dawn and a flat course”. Both the data from 2000-2003 (with an unprecedented amount of negative splits) AND the sudden change in 2004 are utterly implausible. This should have set alarm bells ringing.

It would have been easy to ask one of the Biel 100 km veterans, whether there had been a major change in 2004. The names of all the finishers can be found at 100km.ch web site, so that would have been easy to do.

I have never been to Switzerland, so in order to find out the truth, I had to spend half an hour digging the internet. The waybackmachine found older versions of 100km.ch website, and the outcome was not surprising.

TS3 has been in Bibern (76,7 km) only since the change of the route in 2004, while in 2000-2003 TS3 was in Gossliwil (82 km). The distance from TS3 to the finish line has been 23,3 km since 2004, but before that it was a mere 18 km sprint.

 A quote from the report:
“…using repeated measures one-way analysis of variance with Greenhouse-Geisser correction and Sidak’s multiple comparison tests… the effect of change in altitude on running speed was estimated using means of the Akaike information criterion… using one-way analysis of variance with Dunnett’s multiple comparison post hoc tests… calculated Cohen’s d using means and standard deviations…“ [source: Pacing strategy…, page 3]

My comment:
It is very easy to see that basic source data is seriously flawed, so running extensive statistical tests on this material is a waste of time. Actually, it is rather surprising that the flawed dataset somehow “survived” all these tests. 

The final verdict

The split time data used as the base of this report are seriously flawed. The data from 2000-2003 and 2004-2009 are not commensurate. Trying to identify trends from such data probably leads to flawed conclusions.


Disclaimer: I am not properly qualified to comment on scientific research, as I have no medical education and my expertise in sophisticated statistical methods is next to nothing. I am a pure amateur in these areas. Despite my academic limitations, I have done some (extremely simple and un-academic) research on pacing strategies in Spartathlon 2008 and Oxroad 100 miles 2013.