Here’s a quick overview of speed ratings. My high school speed ratings are an adaptation of horse speed ratings and speed figures used for handicapping and betting thoroughbred races.
You can not simply sit down and start making speed ratings. It requires gathering and compiling LOTS of race data. You start local .. for my high school ratings, I started with the SUNY Utica course because it was the Section 3 championship course and it was going to be the NY State Meet course in 2000 (the year I started posting XC speed ratings for NY). That course was slow and difficult. I arbitrarily decided that 26 minutes would be zero points and every 3 seconds faster would be 1 point higher .. this was based on getting a 100 speed rating for median above average girls (I was a volunteer coach for the Tully girls).
The SUNY Utica course times became the original baseline. Then I worked backwards using Section 3 runner results from previous meets at different races and courses to determine the relative speed of individual runners on individual courses. This runner-to-runner comparison allows determining the inherent speed of courses on specific days .. and a seasonal compilation determines the inherent speed of individual runners.
Local data compilation progressed to NY Statewide data compilation .. NY runners also competed in a number of out-of-State races, so I complied data for those races which was a beginning for an unknown, but upcoming national evaluation to be called Nike Team Nationals in 2004 .. and this morphed and morphed into my current speed rating process on a national basis. Obviously, much cross-correlation between races occurs.
Speed ratings and race adjustments are determined by three methods in order of preference:
(1) Reference Runners .. meaning runner-to-runner comparisons of seasonal speed ratings of individual runners. The final times of races are correlated statistically with the known speed ratings of individual runners, and the correlation yields an adjustment for how fast or slow the race was compared to my baseline of zero equaling 26 minutes. The adjustment is applied to calculate the speed ratings. This is the preferred method.
(2) Graphical Method .. I always start by graphing the race times vs. finish positions to see the race profile (what it looks like on a graph). And whenever possible, I compare it on the graph to the same race from previous years. Each individual race has its own baseline, and I have compiled a library of many hundreds of baselines and race profiles. A graphical evaluation often yields a decent or ball-park race adjustment when using the same race from previous years.
I also maintain a library of race profiles for generic types of races .. average invitational, poor quality invitational, high quality invitational, and various others. I know the inherent speed adjustment of those types of races, and a race profiling method comparison may yield an acceptable race adjustment for the race in question.
(3) Known inherent course speed based on historical data (as compared to the SUNY Utica baseline) .. This is my last resort, but sometimes usable. Van Cortlandt Park 2.5 mile races are an example.
In actual practice, I start with the graphical method to approximate a race adjustment. I use that approximation as a starting point to statistically iterate the reference runner known speed ratings. All this requires a library of evaluated data.
Speed ratings are never perfect or precise .. It’s impossible because the data points (humans, horses) are variable, imperfect data points. Human (horse) performance is variable, but that makes betting interesting! .. When I have sufficient decent data, the speed ratings should be accurate enough for the purpose. Speed ratings are simply one handicapping tool … It’s up to the user to determine their use.