Statistical Lap time & Racing category Analysis
12/06/2011 - Dale Riley
With the advent of Chip timing for Colorado Cyclocross races in the 2011 season, a large set of data is becoming available to do all kinds of geeky data analysis. Besides the basics of seeing how far off of the leader you are lap to lap, and watching your own lap times to see if you speed up, slow down or remain consistent over a race, the ability to look at the age & category times can provide a data driven look at how each racing category stacks up to the others, as well as if the progression up thru the category groups is in even steps, and even for that matter, if the progression of age slows you down in nice, even steps.
A set of data was taken from one race, the Excel Sports Monarch High Cyclocross from October 16, 2011. This race was chosen as it was a consistent course, it's mostly hard, fast grass did not change conditions much over the day. The first laps times were discarded, in order to minimize the effect that different size fields and first lap crowding would affect the lap times. The remaining "spread out" laps were used for the analysis. Also removed from the data set was any riders lap times who were lapped or only finished part of the race. This was done to insure full race lap times were represented. Only the men's groups data was done, as the field sizes of the women's groups were small enough to not appear statistically significant. The sample size in the open mens data set is not nearly as big, so may be a bit skewed from a smaller sample. Also, the differences in the overall time for each race was not addressed. times ranged from 40 minutes to 60, so the open groups did many more laps than the lower categories, and that could also somewhat affect the lap times.
Statistical analysis software was used to generate distributions of lap times for each group, as well as mean (Average) and Median (Mid point of the distribution, half the times slower and half the times faster...) lap times for each racing category. Histograms of the distributions, and plots of the Mean & Median numbers were generated. Besides the overall group data, Progressions up thru the ability categories for the senior men and 35+ men groups were done, along with the open age group categories.
Russell over at The Road to Cat 1 did a similar analysis on the same race, and you can see his approach to the data and conclusions over there. I had hoped to find another race to expand the sample size, but so far did not see any races with lap time data that would remain reasonably consistent. Plus, it's a lot of time to sort out & prep the data, so may never have gotten it done if there was more data!
Russell over at The Road to Cat 1 did a similar analysis on the same race, and you can see his approach to the data and conclusions over there. I had hoped to find another race to expand the sample size, but so far did not see any races with lap time data that would remain reasonably consistent. Plus, it's a lot of time to sort out & prep the data, so may never have gotten it done if there was more data!
Distributions for all groups are here:
(you may need to click on this image to blow it up to readable size...)
(you may need to click on this image to blow it up to readable size...)
The Mean and Average lap times for all groups is here:
Nothing unexpected about the data here, wider distributions in the lower categories and older age groups. Speed increases as the ability category increases, and slows somewhat as age increases.
Given recent commentary on adding racing groups in Colorado cyclocross, I split out the category "progressions" for the senior men cat 4, 3 and open, as well as age 35+ from category 4 to 3 and open. intent was to see if the resulting race groups showed "even" increases in overall speed, and have "reasonable" overlap in the distributions.
Senior Men Progression - Distributions
Median & Average Lap Times
Average & Mean lap times speed up by approximately 40 seconds from 4 to 3, then only speed up about 20 seconds to the open group. The distributions overlap fairly well, with the fastest cat 3 lap times landing about in the middle of the distribution of open lap times. Somewhat less overlap in Cat 4 to 3 times, but still the fastest cat 4 times are near the middle of the cat 3 times. This would indicate the fastest of the lower category times would be near the middle of the next fastest group. For the Senior men, the jump group to group seems pretty well spaced, with a bit of a larger speed jump from cat 4 to cat 3 than the cat 3 to open jump.
So how about the large Masters category groups, and how does the hotly debated 35+ cat 3 field slot into the ability progression? Seemingly pretty well, and very similar story is seen in the 35+ groups, compared to the Senior ability progression. There is a very large "tail" in the lower lap times for the 35+ 4 group, as could be expected by it's "beginner" category opportunity. Still, distributions show good overlap, with the fastest in each 4 & 3 group matching up well to about the middle of the distribution in the higher category. This data shows offering the 35+ 3 group does offer a reasonable intermediate step up to the 35+ Open. Average & mean lap times show about the same delta as the senior groups, with about 40 seconds form cat 4 to cat 3, and a bit more than 20 seconds From 3 to open. The new 35+ cat 3 group matches the progression in the senior groups pretty well.
35+ age group ability category progression distributions are here:
Average & Median Lap times
Distributions for the open masters age groups:
Age group average & Median lap times
I also found a similar lap speed analysis done by Colin of crossresults.com fame over on his blog. His was a similar analysis done on a race with little course condition change. Wile I think my graphs are a LOT cooler, and look far more technical, his data shows a very similar outcome on distribution as well as age & category progression. A side note, Colin was also an inspiration for me getting and running a GoPro camera on my cross bike, mostly from a seat cam compilation he did. Anyway, here is some East Cost data, stolen from Colin's Blog.
My Conclusions? Nothing earth shattering. Racing gets faster as you go from cat 4 to 3 to open. Racing gets slower as age goes up. Pretty standard. My take is that the categories as run today in Colorado allow "reasonable" progression for both senior and master riders to upgrade and not immediately start fighting to avoid DFL. It seems reasonable that the fastest times in the lower category hit pretty much in the middle of the next fastest group. allows for a reasonably linear progression in speed up thru the groups. There is a bit bigger jump in average & median lap times going from category 4 to 3 in both senior and master ranks, than the jump in average & median times from cat 3 to open. The cat 4 groups have a MUCH larger "tail" in the slower lap times, with 35+ 4 the largest. This indicates a lot of people far outside the "average" lap speeds are in these races, as you would expect. Also seems that the 35+ Cat 3 fills a needed slot, as the jump from fastest times in the 35+ cat 4 to the 35+ open would be much larger than any other ability progression without that group being offered.
Overall, ignoring any time constraints, philosophical arguments, or other constraints, the current age & category structure seems to provide a good split in ability and age capability. While it may not be Ideal, and there are other potential ways to split our racing population out, what is being run now seems pretty well split from a lap time perspective.
