Ultra Running Races – shallow competition?
Recently I have heard a growing number of comments from experienced competitive ultra runners that ultra running races have a very shallow competitive pool of participants. I too have had this sense of the competition and about a year and a half ago I conducted an analysis to test the following hypothesis- ‘ultra running races are less competitive than marathon races’. This is not intended to be an exhaustive analysis, just an exemplary submission, but it does provide a data-based perspective and, in my mind, an interesting result.
Much discussion ensues when this topic comes up but I have yet to see any data-based argument provided. I hope that the following adds in a positive way to the discourse. These data are just a sample of a more comprehensive analysis that I have been working on. Although not complete, the extensive analysis appears to be in general agreement with what is presented here.
Although more rigorous analysis techniques are utilized in the literature to measure individual sport competitiveness, e.g.:
such analyses are not directly applicable to trail ultramarathons as the courses (terrain, aggregate elevation change, altitude, etc.) and weather play a major role in the finishing times. The simple approach utilized below allows for analysis and comparison of individual events and results in a reasonable metric for assessing competitiveness.
Background and Data Analysis
To test the soundness of claims that ultra running races are not as competitive as marathon races I chose to compare three races, two ultras (Western States Endurance Run (WSER) 2011 and the Ultra Trail du Mont Blanc (UTMB) 2011) and one major marathon (London 2011). The most defensible metric for such comparisons is the distribution of finishers as a function of percent back from the winning time. One can use other metrics but this one is clear and easily interpreted. Presented below are the histogram data for the 2011 Western States Endurance Run, one of the most competitive ultramarathon races in the world. Plotted is the number of participants who finished within set ranges of percentage back from the winning time – 0-10%, 10+%-20%, etc. Note that there were 14 racers who finished within 10% of the winning time. These 14 racers represent about 4.5% of the 310 finishers. Keep that number in mind for a discussion to follow. Also note that the results are truncated because WSER has a cut-off time of 30 hours- all competitors who finish after 30 hours ( and some that miss cut-off times along the course) are not included in the results.
Now let’s take a look at the same results from the 2011 London Marathon, one of the largest, most competitive marathons in the world. The 2011 event had 34,806 finishers and the complete percent back histogram dataset is presented here:
For comparison purposes, it is appropriate to truncate the London Marathon results in a similar way to that of the 2011 WSER. Looking over a number of years of WSER results (2004-2011)*, it is found that typically there are no finishers beyond 110% of the finishing time due to the time cutoff. Another very competitive ultramarthon that will be used for comparison later typically has finishers within 130% of the winning time. The last finisher at the 2011 London Marathon was 400% back from the winning time. To allow comparison, we will use a percent back value 110% for truncation. The 2011 London Marathon histogram is as presented here, now with 14,253 finishers within 110% of the winning time:
The shape of the 2011 WSER and the 2011 London Marathon (truncated) percent back distributions are not dramatically different, represent entire (truncated) populations and therefore, to first order, can be used for reliable comparisons. Accepting this, then comparisons of the percent of the truncated population in each of the percent back bins is comparable- and allows for a base metric that allows analysis of the two substantially different population sizes (n). Presented below is a comparison of the percent of population as a function of percent back from the winning time:
Clearly, these populations are very different, particularly at the competitive end of the tail of the distribution. The data from the 2011 London Marathon demonstrates exactly how unusual it is for a competitor to be within even 30% of the winning time (1 in 74), let alone 10% (1 in 800). Whereas for the 2011 WSER, 1 in 7 competitors are within 30% of the winning time and an amazing 1 in 22 are within 10% of the winning time. Based on this, and the reality that the 2011 WSER population of 310 competitors is of sufficient size to be representative of the entire “ultramarathon” population (and, of course, likewise for the 2011 London Marathon), the degree of competitiveness in this “most competitive” ultramartahon is not even close to the competitiveness of a large marathon. It is 36 times less likely to find a competitor within 10% of the winning time in the 2011 London Marathon than in the 2011 WSER. This is an enormous difference in competitiveness.
One might argue that if the ultramarathon population were to grow (and ultramarathon events allowed more competitors) a disproportionate effect would obtain on the distribution of finishing times in a way that skews the distribution toward larger percent back finishing times. I find that this is not founded as the 310 WSER competitors in 2011 are likely representative given the nature of the selection process and that there is no reason to assert that the distribution of those drawn to the sport would change in ability (performance) on average as the number of participants grow. But, as the sport of ultramarathon running grows, the probability of ultramarathon competitors at the very ‘sharp’ end of the distribution increases just as it has in the sport of road marathon running. If the 2011 London marathon data are indicative, these ‘sharp end’ ultramarathon competitors would be few and significantly better than the rest. If the 2011 WSER data are indicative, such does not currently hold in the sport of ultramarathon running.
To add another dataset let’s look at one of the largest and most competitive ultramarathons in existence today- the Ultra Trail du Mont Blanc (UTMB). Specifically the 2011 UTMB, which did manage to traverse over 160 km (100 miles) on a rerouted course that year. Here is the percent back distribution for the 1132 finishers of this race:
UTMB also has a cutoff time, in this case 46 hours, which led to a significant number of finishers in excess of 110% of the finishing time. Per the protocol adopted above we will truncate this population at 110% of the finishing time (813 of the 1132 finishers) for comparison purposes. Doing this and then calculating the ‘percent in bin’ as before the following is obtained:
The 2011 UTMB (truncated) definitely has a substantially ‘sharper’ competitive end than 2011 WSER but this is still significantly less ‘sharp’ than the 2011 London Marathon (truncated). In the 2011 UTMB (truncated), 1 in 270 competitors is within 10% of the finishing time compared to 1 in 800 competitors in the 2011 London Marathon (truncated) and 1 in 22 in the 2011 WSER. One in 45 competitors are within 30 % of the finishing time for the 2011 UTMB (truncated) compared to 1 in 74 competitors for the 2011 London Marathon (truncated) and 1 in 7 competitors for the 2011 WSER. Thus it is 3 times less likely to find a competitor within 10% of the finishing time in the 2011 London Marathon (truncated) than in the 2011 UTMB (truncated). This is a very large difference in competitiveness.
Update 9 April 2013: A table of the percent back from the winning time as a function of ‘percent in bin’ for the 2011 London Marathon compared with the 2011 UTMB and the 2004-2011 WSER is presented at the bottom of this post. As is clear in this table, similar results as that presented above obtain for the entire period of WSER studied.
A simple approach for analyzing and comparing trail ultramarthon competitiveness with road marathon competitiveness is presented. Comparison of two ‘large’ and ‘competitive’ trail ultramarathons with one large, competitive road marathon reveals that trail ultramarathons are substantially less competitive than road marathon equivalents.
It is expected that, as the sport of trail ultramarathon running grows, population of the expected ‘sharp’ competitive tail of the athletic performance distribution will occur and lead to significant gains in competitiveness within the sport. The relative early stage development of the sport of ultramarathon running and the predominance of competitors from other, allied, sports has lead to a blunt-ended distribution at the competitive end. This leads one to postulate that as the sport matures and competitors grow up within the sport (rather than being drawn to trail ultramarthon running later in their careers from other sports) the ‘competitive tail’ will be efficiently filled and the ’3+ sigma’ athletes will be a part of any competitive race.