TullyRunners -  Article

Can You Make Your Own Speed Ratings??

by Bill Meylan (August 17, 2011)

Every year I get inquiries from people (coaches, parents, runners, interested observers) who want to make their own speed ratings. Most inquiries come from out-of-NY State since I post only a few selected speed ratings from other States during most of the cross country season. In general, the interested people have read some of the speed rating articles posted on this web-site, and believe they can make speed ratings for their own State ... After some initial attempts, they discovered it is more time-intensive and complicated than original thought ... so they are asking for advice ... others ask for advice before attempting it on their own.

This article discusses this issue and talks about several speed rating concepts that are important to understand.

Can Anybody Make Their Own XC Speed Ratings??
I believe many people are capable of making XC speed ratings (meaning they can do it if they really want to and are willing to spend the time required) ... And that’s exactly how I respond to most inquiries ... But interested people need to understand the following if they want to mimic the process I use:

• There will be a steep learning curve for most people

• It will be a "Lot More" Time-Intensive than you thought possible

• Starting out is one of the hardest aspects ... It initially requires major background work and practice

• Good computer skills with appropriate software is absolutely necessary

Background Requirements ... I was fortunate when I started ... I had years of experience in making horse speed ratings and understood the basic concepts ... There is no better teacher for "understanding the concepts" than losing money at the race track.

When I started examining the possibility of making high school cross country speed ratings in the late 1990s, I began the process exactly the same as if I was moving to a new horse track in a different location and needed to collect background information ... Initially, my "new horse track" was Section 3 high school XC in NY State ... The background information needed was race results (as complete as possible) ... I spent about four weeks collecting Section 3 race results from a variety of sources ... Electronic Internet results were scarce at the time, so I had to get mostly hard-copy results ... I spent a lot of time in libraries in Syracuse, Utica and Watertown and several sports desks at newspapers ... several meet directors and coaches were very helpful ... People starting out today can usually get sufficient results from the Internet (a great time saver, but still requires time and effort).

I collected Section 3 race results from the late 1990s, so now what?? ... Speed numbers for horse racing (thoroughbred + harness) rely on three concepts:

• Horses can be classified into groups of specific "quality" ... these groups include claimers (horses available for sale with a specific price tag), stake-grade horses, and a variety of other "horse" classifications ... A mountain of data from many years had shown that each specific group run races at a known speed relative to each other (meaning the final race times were correlated by class ... better horses ran faster and lower quality classes ran slower at a known rate).

• As a season progresses, each horse will have a series of known speed ratings (one speed rating for each race) ... Collectively, the known speed ratings for all horses running on a particular day are used to derive the "speed" for that day and the corresponding speed ratings ... This is the essence of speed rating and how it is best applied.

• Individual race tracks generally have an intrinsic or inherent speed (this is primarily a harness racing concept) ... When the same horses (or classes of equal quality horses) race at different tracks over a period of time, a compilation of final times can be used determine how fast one track is compared to another track ... This is a concept of comparative times that already existed in high school XC (e.g. course conversion charts).

Adapting the Concepts to High School Cross Country

For starters, here are two useful speed rating articles on this web-site:
 ... (1) Early Season Speed Ratings and a Brief Overview of Speed Ratings
 ... (2) Speed Ratings - Response to Some Common Questions

My first adaptation concern was developing a speed rating number table that corresponded to actual XC race times (the table is shown in article #2 above) ... I arbitrarily decided that one speed rating point equaled three seconds for most XC races (4K races and shorter can be an exception) ... the table numbers were initially developed for girl's XC (boys were an after-thought).

High School Groups of Specific Quality:
This was a trial-an-error process because horse-racing and high school XC are classified very differently ... But it quickly became apparent there were four basic varsity groups within Section 3:

 ... (1) a general overall group as defined by merged results from sectionals
 ... (2) low-quality invitational
 ... (3) medium-quality invitational
 ... (4) high-quality invitational

The "general overall group as defined by merged results from sectionals" became my first "base-line" (but it is only one of many base-lines I currently use) ... Section 3 (in NY State) has an excellent statistical sampling of small, medium and large schools of varied ability making it applicable (at times) throughout NY (and many other States as well).

NY State has eleven sections ... So I have base-lines for all those sections as well ... this was necessary when I expanded beyond Section 3 ... And at the State Level, I have base-lines for:
 ... NY State Meet
 ... NY Federation Meet

Base-Lines & Profiles ... These are basically the same thing ... I maintain a large "library" of profiles and base-lines ... Every race I speed rate gets a "profile" in either an Excel spreadsheet or plain text-based file ... The profile includes: (1) the finishing position for each runner (sometimes merged with races from the same meet), (2) the corresponding actual race-time, (3) the speed rating correction factor in seconds (the time added or subtracted from the actual race-times) and (4) the derived speed ratings ... These profiles of "known speed" are used statistically or graphically for comparison to other races to help determine their speed rating correction factor.

When I first started this process in the late 1990s, I decided the merged sectional results (from Section 3) were the base-line by definition ... no correction or adjustment needed ... all speed ratings were taken directly from the table ... period! ... Then I started working backwards with league meets and invitational races one race at a time using the general profile AND individual runner ratings to speed rate as many races as possible with "reasonable" accuracy ... After a fair amount of time and effort, I had speed ratings and profiles for an entire season of Section 3 racing ... NOW I could try it on the upcoming season and hope it works ... This background work is necessary.

Example Use of Profiles ... I'm discussing profiles here because each profile corresponds to a specific group or specific quality of race ... The profile for a State Meet has a different quality than a profile for an average invitational ... The first season I tried this process on live-racing at the Tully Invitational, I used the profile from the previous year's Tully Invitational and the general Section 3 base-line profile as the statistical comparisons (could not not have done it accurately without them at that point of the early season).

The Essence of Speed Rating - Using Individual Runner Ratings to Speed Rate a Race

A re-occurring example in Section 3 ... Leone Timing posts the results a few minutes after a race finishes ... I'm looking at the results and one or more coaches approach and ask "What do you think the speed ratings will be??" ... One coach points to his or her runners and notes their "typical" or recent speed ratings plus a few other runners they may be familiar with, and based on that, the coach suggests the ratings will be roughly "xxx" at the top with his runners at roughly "xxx" ..... Yikes, people are starting to understand these ratings ... I should have kept the process cloaked in secrecy and mysticism.

That's the general idea of speed rating a race based on the known ability of individual runners ... I have a computer program that matches every runner in my database to the results of a race ... The program appends each runner's overall seasonal speed rating (and individual ratings) ... This can then be used statistically or graphically to derive speed ratings for the race ... Yes, I take it to extremes because sometimes it's necessary ... some races require special attention to speed rate correctly ... and all of this takes time.

Something to note in the example ... This method can be be applied without knowing the race course on which a race was run, and that is a significant concept ... Speed Ratings are meant to compare runners to each other.

Inherent Speed of a Race Course

This concept has been used by XC coaches for many years ... The two articles linked above talk about "Course Conversions Charts" for high school XC ... For example, results indicate that one race course may be 10 or seconds faster or slower than another race course on average ... Although this useful information, it can lead to significant inaccuracies when applied to specific races "as speed ratings" ... It needs to be used with caution and supporting data whenever possible.

When I first started making speed ratings for 1999, I was given a course conversion chart for most XC courses in NY State run on by Section 3 runners ... Jerry Smith spent a lot of time compiling the chart (Jerry Smith was the Fayetteville-Manlius coach who preceded Bill Aris; he coached for a time at Cornell and is a current coach at Westhill-Ludden) ... The chart was accurate in 1999, but over time, many of the courses are no longer used or have been altered significantly ... This is a reminder that course conversion charts need to be updated frequently.

 
                                  Median 2010
  Cross Country Course            Adjustment    Normal Adjustment
                                 (in seconds)   Range (in seconds)
  -----------------------------  ------------   ------------------
  SUNY Utica NY                       0             standard (Section 3 sectional 1999)
  Bowdoin Park NY                    -21         -12  to  -27
  Sunken Meadows NY                  -18         -12  to  -24
  Van Cortlandt Park NY (2.5-mile)   -235       -230  to  -245
  Saratoga Park NY (3.02 mile)       -84         -75  to  -90
  Holmdel Park NJ                    -39         -33  to  -42
  Woodward Park CA                   -72         -63  to  -78

Course conversion charts do not factor in (1) weather-related things that change the speed of a course or (2) alterations made to a course that change the speed ... These are important considerations.

Courses get altered a lot more than I realized ... "construction on-site" is common culprit (e.g. Van Cortlandt Park) ... Virtually every week last season, I got an e-mail from a coach or meet director describing a course change at an invitational (they thought it might be helpful in making the speed ratings).

Another problem that's come-up in recent years is the number of different 2.5-mile courses at Van Cortlandt Park ... Three different 2.5-mile courses were used one year (by the Manhattan Invite, by the CHSAA, by the PSAL and by the AIS).

This season, NY Section 3 sectionals will be held at the Jamesville Beach course ... This course has been used for a number of championship races in recent years ... Unfortunately, the course was changed a number of times and is expected to change again this season ... Any course conversion number would be meaningless at this time.

A Recent Request ... I was asked by a coach (outside NY State) if I could send my complete course conversion chart for every course I have speed rated ... He wanted to use it to speed rate races in his State and other States ... I do not maintain a "formal" up-to-date conversion chart ... I told the coach that I speed rate races outside-of-NY strictly by "profiles" as described above ... Course adjustments are available from the profiles on an individual race basis.

I believe the MileSplit Network is looking at the possibility of using their large results database to derive course conversions for every high school course in the nation ... and then using the conversions to rank runners and teams ... It would certainly be better than distance-based ranking, but would lack the accuracy of speed ratings.

Time Requirements

I was going to describe my process in more detail, but it depresses me when the amount time spent gets totaled ... Here are some steps:
... Get results, convert to usable format (sometimes easy, sometimes a time-consuming pain-in-the-butt)
... Retrieve race profiles (previous years, similar races)
... Apply & Evaluate race profiles with race of interest. This typically requires a statistical program or spreadsheet. Evaluation done both graphically and with numbers (and compared).
... If sufficient individual runner data available, Apply & Evaluate via database comparison. Compare with profile results.
... Generate speed ratings

Since I post stuff on a web-site:
... Add speed ratings to results and post
... Upload results with speed ratings to a database; database will later be used to make web-pages for viewers to see; database will later be used to rank individual runners for people to see.

It really is a full-time job over the course of a week (and sometimes more) ... Average of roughly 16-20 hours on a typical Saturday and Sunday (does not include going to the meets) ... Why do I do this?? ... There really is a reason I want to quit every year.

It is possible to "Short-Cut" the process and sacrifice some accuracy ... might be "good enough" for most people ... But I can't do it.

Developed and maintained by wmeylan@twcny.rr.com