TullyRunners - Article
High National Speed Ratings from the Past
by Bill Meylan (July 15, 2015)
Recent inquiries prompted this article ... Since TullyRunners focuses on speed ratings for cross country, some viewers are asking about the highest speed ratings at the nation level in past years ... In particular, they are asking about the speed ratings for the following three performances:
Dathan Ritzenhein - at the Michigan State Meet 2000 (MI State
Followers of TullyRunners probably know that the speed ratings for the Footlocker National races are generally the highest speed ratings I have derived nationally ... Each year I post the historical Footlocker results with speed ratings (including the highest).
The following table lists my highest speed ratings from both Footlocker and other XC races (including the three performances mentioned above) ... Please Remember -The lists are limited to races I have evaluated ... and that excludes most non-Footlocker races prior to 2000.
interested, this article also
discusses the method used to derive the speed rating for Dathan Ritzenhein's
Michigan State Meet record ... Plus some background information on Cathy
Schiro (see why she was capable of high speed ratings).
Why are the Footlocker National Speed Ratings So High?? ... Footlocker Nationals is unique ... A small field of many of the best elite runners in the nation compete head-to-head on a course that is generally in good (fast) condition with good weather ... Runners come to run fast and win ... The best finishers often elevate their performance level to new heights (by necessity to compete well) ... and it is reflected in the speed ratings by the process I use to derive Footlocker speed ratings ........ Analogies in track & field might be races like the Arcadia Invitational 3200 meters, the Brooks PR Invitational, the adidas Dream Mile, professional & college races and others where fast national track times occur due to the nature of the race and the competition ... High school cross country has far fewer opportunities, and Footlocker National stands out.
German Fernandez and Cathy Schiro:
Dathan Ritzenhein & the Michigan State Meet 2000:
Graphical Interpretation of Cross Country Results
My BS degree is in chemical engineering, and engineers use graphs and plots for all kinds of things ... My career as as an environmental scientist involves manipulation of large data sets of chemical and physical properties for various estimation techniques ... So I am comfortable with graphical solutions for data evaluation ... and I have applied some graphical techniques to evaluating high school cross country results.
Important Note about Data Points ... Humans are not perfect ... Not only are humans imperfect, human performance is variable; some more variable than others ... So statistical (or subjective) results from imperfect-variable data will never be perfect or precise ... BUT the results can be accurate enough for the intended purpose ... That is the goal of speed ratings - to measure the relative speed of runners at different races, different venues and different dates with acceptable accuracy ... "relative" means relative to all other runners.
I graph the results of nearly all races I evaluate ... any anomalies become visible as does part the race profile itself ...... I maintain a library of race profiles for most races I evaluate ... a Race Profile includes:
... (1) the data points
(finishing position with corresponding race times and derived speed ratings)
All race evaluations are NOT equal in quality ... Some are just better than others ... some have low uncertainty and low margins of error ... some have high uncertainty and may be useless in future statistical application ... I use higher quality race profiles to help derive speed ratings whenever possible.
Michigan State Meet 2000 - Graphical Evaluation
In my speed rating process, the graphical profile of the Michigan State Meet 2000 needs to be compared to existing race profiles of known speed in my library so I can approximate a race adjustment for the 2000 State Meet based on known adjustments from other races ... Using just any race for comparison will NOT work ... Experience has shown that using races of approximate equal overall quality general works if applied properly, and that the best comparisons commonly come from the same race from previous years.
So my first choice for graphical evaluating the 2000 Michigan Boys States Meet is to compare it to other Michigan State Meets ... Michigan used four divisions at States in 2000 and are still using four divisions with approximately the same number of runners in each division ... this really helps in maintaining overall quality for my comparisons ... The graph below includes merged results from the Michigan State Meets in 2000, 2013 and 2014 for the top 200 runners:
The straight red lines are my eye-determined manual placement of time comparisons (that's how I do it) ... I devised my process to read the difference at the Y-axis based on population segments of the races ... As can be seen, the top runners are ignored ... Dathan Ritzenhein and Grant Fisher's speed rating is determined by how fast they ran relative to other runners, and NOT by how fast they ran relative to the course.
Michigan 2000 and Michigan 2014 look very similar in terms of speed ... only a single red-line is needed on the graph ... In contrast, the graph shows Michigan 2013 running about 9 seconds slower than Michigan 2014 (I like multiples of three since 3 seconds equals one speed rating point).
In addition to the merged results for both boys and girls, I also look at graphs of individual division races.
My race profiles for Michigan 2012, 2013 and 2014 have established race adjustments:
... Michigan Boys 2014 - 75
That is the amount of time I add to the actual final race times to determine the speed rating ... 2013 was about 6 seconds slower than 2014 ... and 2012 was 6 seconds faster than 2014 ... Note that the recent Michigan race adjustments above include a separate method of determination using known speed ratings of individual runners from a database and statistical correlation with final times ... I don't have that for Michigan 2000, so I'm relying strictly on a graphical solution with Michigan 2000 ... I looked at other years as well.
Overall - Michigan 2000 had a race adjustment of about 78 seconds (or it was about 3 seconds faster than 2014) ... If somebody wants to argue it was 3 seconds faster or slower than my number, I would not complain because it is well within any margin of error.
Comparing Michigan 2000 to Washington State 2014 and New York State 2014:
I also looked at other States ... I tried the New York State 2014 Meet because it uses four race divisions at States ... I also tried the Washington State Meet because I thought the Washington course might be a good comparison to the Michigan International Speedway (actual site of the Michigan State course) ... Here is the graphical comparison:
The Washington State 2014 Meet has a known race adjustment of 81 seconds (good quality) ... the graph above shows a difference of about 6 seconds between Michigan 2000 and Washington 2014 ... The overall quality of the Washington State Meet may be a bit higher than Michigan 2000 if the slopes of the red lines are appropriate (I excluded the lowest of 5 divisions in WA which could explain it) ... But it generally agrees with the race adjustment of 78 seconds for Michigan 2000 based on other Michigan State results.
The overall quality of New York State Meet can differ a bit from other States meets because its qualifying procedure ... The graph shows a difference of about 27 seconds slower than Michigan 2000 ... Again, nothing to contradict the race adjustment derived from other Michigan State results ..... I tried a number of other State Meet results as well including older results going back a number of years ... Bottom-Line - a Michigan State 2000 race adjustment of 78 seconds seems accurate enough for the intended purpose.
Important ... The graphical solution method I use requires experience ... Where to draw the red lines and believability of the solution is something I had to learn over time ... The examples above are fairly easy and straight-forward, but many early and mid-season races can be a challenge ... I decided not to post speed ratings for about two dozen out-of-State races last year because I knew the solution was too uncertain at the time.
I use three different methods to derive race
adjustments for speed ratings:
As said in previous years - "Only a crazy person would spend the time doing what I do in high school cross country" ... But deriving speed ratings for Dathan Ritzenhein and Cathy Schiro was an interesting exercise ... and I actually believe the results.