ric flair woo wrote:
So Tully Runners has Josh Methner's 14:11 at detweiler this year as a 188 speed rating. They rated Liam Anderson's NXN win as a 201. And if each point is worth 3 seconds, they are basically predicting that Liam Anderson could run a 13:32 on detweiler, which would beat Craig Virgins record by 18 seconds. Am I understanding how speed ratings work correctly?
NO ... That is NOT how speed rankings work.
Speed ratings are a runner-to-runner comparison and NOT a runner-to-course comparison.
My typical example ... Any invitational meet is raced on any course and the individual runners are speed rated ... the following week, the same runners race on the same course and everybody runs 15 seconds faster ... everybody expects the speed ratings to go up ... BUT the speed ratings stay exactly the same because on a runner-to-runner comparison basis Nothing Changed ... Many Track coaches hate this concept or can’t understand it.
Speed ratings (derived from horse racing) want to know who to bet based on where runners finish relative to each other (time difference between individuals) ... Speed ratings are a measure of relative speed.
The "relative part" is based on statistical derivations and cross-correlation that includes:
(1) seasonal speed ratings of the individual runners
(2) historical statistical profiling on the individual race
(3) historical statistical profiling on the individual course
(4) a cross-correlated library of known race adjustments for the current season and prior seasons
My speed rating process started with one "baseline" race in Section 3 of New York (sectional championships that includes a good cross-section of large and small schools such as Fayetteville-Manlius, Liverpool, Tully and others) ... From that race, I decided that a final time of 26:00 would equal 0 speed ratings points and every 3 seconds faster would add 1 speed rating point ... That’s where the speed rating numbers are derived from.
For every race I speed rate:
(1) I apply the statistical correlation noted above to derive a "race adjustment" ... the "race adjustment" is the number of seconds in time the race was faster or slower than a "baseline" ... the baseline is usually for that race specifically already determined from historical data ... all baselines in my library have been cross-correlated backwards so that applied race adjustments yield times where 26:00 will equal 0 speed rating points.
(2) I take the race adjustment and add it to the actual final times of the race in question ... I then take the adjusted final times and calculate the speed ratings for the race (26:00 will equal 0 speed rating points).
Determining the race adjustment is a time-consuming process when building of databases and data retrieval is included.
The overall process is part statistics and part "art" ... some believe the "art" part means it’s nothing more than Meylan’s opinion ... But most of the "art" part and guessing I have talked about involves statistical iteration which is a well known process in statistics where guesses are made to help fit the available data to find an acceptable solution. The "art" part is also necessary for other situations where anomalies occur in the data and human interaction is required.
Subjective opinion of hard and easy courses has no place or inclusion in speed rating ... speed rating is based on fast or slow and the final times alone are all that is needed to determine if the race was fast or slow ... the final times of the race derive the speed ratings (not weather, mud, hills, flat or any other factor).
The race adjustment for NXN Nationals was 60 seconds for both boys and girls (I added 60 seconds to the final times and calculated the speed ratings) ... Last year, the adjustment was 54 seconds for the boys and 39 seconds for the girls (the boys mucked up the course for girls and made it slower).
This year, 9 of the top 10 boys set speed rating PRs and bunch guys behind them set speed ratings PRs (looked more like a Foot Locker race, but that’s what happens when top competitive runners are in the same race under decent conditions ... the top guys can distance themselves from most runners and that results in higher speed ratings) ... Many girls set speed rating PRs (some by a lot).
Detweiller is a fast course (I suppose 3 miles in distance and being flat helps explain it) ... and it ran quite fast for Detweiller at the Illinois State Meet (the race adjustment was 144 seconds) ... Josh Methner has improved since the State Meet and running against top competition has also improved his speed ratings (Meylan's opinion) ... Also, the top boys at IL State Meet had lower speed ratings than normal (some boys are usually in the 190s), while the girls ran fast (Naperville North girls similar ratings to NXN and Yorkville much higher) ... I don't know how fast Detweiller ran in 1972 when Craig Virgin ran 13:50 (compared to 2018), but 2nd place was 14:20 (so I assume it ran much slower in that 1972 race).