After Tuesday's post about 2014 UCI Race Lengths, and the conclusion "men's races have gotten noticeably longer even though the rules haven't changed," I was rightfully called out for not using 2013 race lengths as a control group.
So, I started digging up that data, which got Matt interested again, which led to Matt making a really nice spreadsheet summarizing everything.
Here is the "real" 2014 data, which is a bit different than what was tweeted. The initial data came from "the UCI calendar as far as we knew it" and this data comes from "the UCI results that were reported to the UCI."
The women's race has been sub-40 five times in 35 events (14%) and the men's race has been sub-60 seven times (20%).
If we assume a women's race with a winning time of 39:XX is actually close enough to a 40 minute race to be within the spirit of the rule (if not the letter of it) then there have only been three incorrectly run women's races this year:
HPCX Day 2 (0:38:20)
Cycle-Smart Day 1 (0:38:16)
Baystate Day 1 (0:34:47)
However the men's distribution still looks quite biased toward the race running long, with 80% of men's events lasting longer than an hour. How does this compare to 2013?
In 2013, 18/44 Men's elite races were less than an hour (41%), and the average race time was 1:21 shorter than in 2014.
The women's races were also significantly shorter (2:37 minutes shorter, 43% under 40 minutes), but this is expected since the rule about the duration of the women's race changed between 2013 and 2014.
In summary, I think it is safe to say that the women's rule change of "greater than 40 minutes" has created a bias toward "greater than 60 minutes" for the men's race as well, even though that's not the rule.
Comparing the 2013 Men's race length dataset to the 2014 Men's race data via a Student's T-test gives P less than 0.05, which is to say there's a 95% chance that this change in the year-to-year data was due to an outside factor (like a rule change) and only 5% that it is just random noise, and that the men's race length hasn't actually changed.
The full data is here.