Note
slept:3.0
The USOF rankings were recently updated and my rank on blue advanced from 61 (65.59 points) to 58 (69.91 points). This is deceptive however, because the total number of ranked runners increased from 73 to 84 due to the local density of A-meet events - six A level events were added for M21 since 7 April, while only one was removed. I also blame the recent Canadian invasion for the swelling.
In any case, I passed the 7 runners immediately ahead of my on April 7, and I'm just outside of striking distance of Neal Trump, Brendan, and Gerald. It clearly will take a lot of work to make a 10 point gain in rankings; I was assisted this run by the recent density of events from QOC, West Point, and CSU (I now have 20 ranked events, so my 8 lowest scores are dropped). This relatively large sample also gives me a glimpse into my performance relative to the field on the different types of courses. It's not a perfect measure because terrain types, environmental conditions, and personal mistakes are uncontrolled variables.
My data points are five sprints, seven middles, three longs, and five classic courses. Because of ambiguity of classification, I consider classic courses and long courses as part of the same class.
The ranking confirms that sprint is my best course, with a mean score of 74.9 and a standard deviation of 2.3 (ignoring the MP). This is unsurprising; it is disappointing that my best individual result is STILL the Team Trials sprint from 2008. I surmise that the standard deviation is small because the range of sprint navigational difficulty is relatively small, and the comparative ease of navigation reduces the probability of large errors. Also, my error rate on sprints has been approximately constant. I need to do more intervals to try to get my speed up and more armchair orienteering to practice quick decisions.
My middle performance is markedly improving but is highly variable; I have seven data points with a mean of 61.8 and a standard deviation of 10.4. If you fit a linear model of my score versus how many A-meet middles I have completed, the coefficient of the index of the middle distance course is 2.82. According to this crude model, each subsequent middle distance course I run will have a score of 2.8 points higher than the previous. I surmise this is due to improvements in my navigational abilities. My best middle so far was at the CSU meet, which is terrain I'm familiar with. I need to practice navigational choices (as my error rate is unacceptably high) and wood speed.
The eight data points for my long courses are illuminating; I have five typical values, two DNFs, and one outlier from the e-punch drop epic fail of May 2008. The DNFs reveal my great weakness - I have endurance problems. Ignoring the three anomalies, my long average is 65.1 with a stdev of 4.53. When I do finish a long, I am fairly consistent, but large navigational errors and inadequate maintenance before the race inflict insurmountable penalties. I think merely increasing my running fitness, I can improve my long course result by about five points. Otherwise, I need to be careful to eat well, hydrate effectively, and bring GU before my future long and classic competitions.
A promising statistic is that I have a 70+ point result in each of the three categories, and while my middle and classic results came from CSU, it's not clear what the systematic shift in my result was. While I was on home turf, I was preoccupied with meet organization, my training had been defunct the week of the meet, and I was running on very little sleep. Also, the temperature was extremely high, and while I do not have enough data to support this thesis, I believe that I respond more adversely to heat than most competitors.
At the end of May, once West Point '08 (at which I met with disaster) and Team Trials '08 are removed from the ranking, my remaining 14 scores give me a score of 71.0 (the top 9; admittedly the ranking method will adjust that slightly). My goal for the fall needs to be consistently running 70+ point runs. My (very ambitious) goal for 2009 was to reach the 67th percentile in the rankings (at the start of the year, that was Baltero). At present, that would require an unattainable score of 88 points. I expect to attend 7 A-meet events this fall: the ROC meet, UNO's Boulderdash, and the individual champs in WI. Ignoring the dynamics of the model, if I ran 80 point runs at all 7 events, I would end the year with a ranking of 78.2. That's probably impossible (since my PRs are 76 point runs), but it's a nice goal to reach for.
Note
A note on modeling:
I find the trend in my progress rather interesting. The signal is extremely noisy - the variance on my courses is extremely large. Discounting my two long DNFs and a sprint MP, my average score is 64.8 with a stdev of 10.1. I decided to model this with a linear regression using two different methods - one which fit score to the ordered index of the course (i.e. the first A-meet of the past 12 months has index 1, the second has index 2 and so on), one which fit score to the date.
I then applied these two models to the 7 A-meet events I expect to run in the fall. A more thorough model would consider the course types (4 classic, 1 middle, 1 sprint, 1 ultralong), but I don't really have enough data to record anything meaningful. The linear model is only really useful for examining the projected average because of the high variance.
Modeling score versus index gives the following score projections:
69.7, 70.2, 70.7, 71.2, 71.7, 72.2, 72.7; i.e. an average of 71.2. Modeling score versus date gives the following: 71.8, 71.8, 72.1, 72.2, 72.6, 72.7. I consider the latter model more reasonable because it accounts for the long time interval between now and the next A event. The latter model also has a slightly smaller error, of 9.2 vs 9.37.