Art and Errors
(Part 5 of a Series)

Charles Carroll

Article Library




All Beyer-like speed figures have a foundation of a theoretical class-par-time hierarchy.  As described last week, simply listing the classes of races at a track in ascending order of speed is no easy trick.  Somehow, bending and fudging times in fifths of seconds for Thoroughbreds seemed less of a problem than doing the same with decimal times, so it took a run at Quarter Horses to make me blink and realize that, if the Emperor wasn’t totally naked, he was at least in a G-string, and it wasn’t a pretty sight.

Since I suggested that a good alternative to The Beyers is to make your own Beyers, the G-string is important.  That is the thread of truth that ties all this together.  Beyer figures are pretty good—certainly the best speed figures that have been published for the world at large.  But, since everyone has them, the question becomes can you do better?

Eventually, I’ll explain an alternative way to calculate and compare speeds in horse racing, but to understand the crowd’s Beyers, you can’t beat experimenting with the steps that are used to create them.

I am an unabashed Beyer fan for a number of reasons, the main one being the level of journalism he brought to this sport—another is the fact that in his approach to making speed figures, there are no magical numerical factors; he derives and describes every step.  The only cockroach in the salad is that underlying all of this is the 1960’s man-to-the-moon belief that the universe—and horses’ times—are orderly.

Pick a track, any track.  Get a stack of Forms or other data source from last season.  Paying no attention to times or surface, simply make a list of every class of race run at that track.  At cheaper tracks, the list will be relatively short.  There may be no graded races at all, and claiming and allowance conditions may be straightforward and simple.

At other tracks, depending upon the inclination of the managers and/or Racing Secretary, conditions can be very complicated, and very situational, sometimes written for a particular handful of horses the Secretary knows is in the barns.  After you have written down every different set of conditions for races at the track, put them in order from “highest class” to “lowest class.”

Right here you run straight into the oldest debate in horse racing:  what is “class,” and “is class speed?”  If you ignore times and deal only with race conditions, there are still numerous internal decisions to make, even at a cheap track.  Is an Allowance $4,200 N1 higher or lower than a Clm $5,000 N2?  Is purse value the common denominator of all racing?

As someone who used to haul friends' losing horses to the track to run in Allowance races because they were pets, I can testify that the Allowance classes are not automatically listed above Claiming at the same or similar purse values.

If you make the list and end up with something like last week’s tables for the entire track, the traditional next step is to make a table for each distance (separating dirt and turf), and go back through the race results, marking down final times run within each class.  This is where you start “fudging.”  You’ve decided that the Allowance class above is higher than the $5K Clm, but there are only three Allowance times to average for 7 furlongs on the turf and none for the $5K Claimers—so what do you do?  You make up a time that is appropriately lower.

As I said early on, the great value of the original Beyer approach was that is was do-it-yourself.  If you did it, you could not help but learn a great deal about times and running conditions at a track.  It is hard to separate which is more valuable to your ultimate profit—the learning—or the figures.  If you are going to do this, I’d suggest a couple of steps to make the figures more valuable.

First, do not make a blanket class hierarchy for track and then apply it to each dirt and turf distance, making up numbers to fill unraced classes or changing numbers to fit preconceived notions.  Instead, use a spreadsheet.  (If you are reading this on the Internet, you almost certainly have one.)

Make a heading for each dirt and turf distance, and under it—without worrying about class order—list only the classes of races run at that distance and surface.  Enter every time recorded on a fast track for that class and distance, and write a little formula to average them in the last column.  Most spreadsheets have built-in formulas, which make this very easy.

This may sound like a bigger task than you want to bite off—and doing an entire track history is work—but if you want to have an interesting hour or so, try it for just a couple of moderately run distances at one of your favorite tracks.

Your spreadsheet should look something like this (abbreviated for space and not all classes and times shown):

Distance: 5.5 Furlongs      Surface: Dirt

Class (no order)

Average Time

Clm $6.2K





Alw  NW 1





Clm $4K





Mdn Spc Wt





One of the first things you should notice while entering numbers, even in small samples, is that times within the classes overlap.  You will also find that for many cases, only a few classes actually run the distance, and when they do, there may be only a small number of races each season.

In all situations, wild-flyer times (fast or slow) are not uncommon.  The conventional wisdom is to throw out flyers, but the computer doesn’t know that and, for this exercise, let the chips fall where they may—leave them in.

Now, for the moment of truth.  This was very difficult to visualize with the pencil-and-paper technology that originated these methods, but it just takes a couple of clicks of a mouse button now: Sort the table by the "average time" column.

Sometimes sorting from fastest to slowest times will rearrange the rows such that “Class” makes pretty good sense.  At least as often, the order of “Class” will be almost as garbled as it was at random to begin with.  Most often, there will be a general trend that the highest Class will be faster than the rock-bottom lowest Class—with a bizarre mix of classes in between.

If you want to take this beyond just an interesting exercise and actually make the foundation class-pars for improving upon the method, the next step is to break with tradition and either lump classes, or accept an unconventional hierarchy.

Suppose “Maiden Special Weight” turns up the fastest class on the track at some seldom-run distance?  Should you throw out the data and give it an artificial time to force it into the hierarchy pattern for the track as a whole?  Two factors could be operating to make it appear the fastest class.  First of all, maybe it is the fastest class.  Some tracks may run hotly contested short-distance races for two-year-olds early in the season, and only a few races at that distance for plugs that can’t stand up a full 6 furlongs.  OR, it could be a function of pure flukes and flyers.  (Understand that these are not just rhetorical questions for making your own numbers—they are built into the speed figures fed to the public.)

If you can recognize a true pattern like the first example, or the absence of a true pattern, like the second, you will make better speed figures and variants if you follow the advice of Tommy Chong and "...go with it."  Let the upside-down classes stand—or, if there is no valid pattern, lump them into several, or even one group, and simply use one or two times as pars for the distance.  This creates a very complicated scheme for comparing speed figures across distances, but it can be done.

The approach that I use makes all of these issues moot, but the purpose here is to understand the published figures the public is using, and understand where art and errors can enter—as well as how you might do better on your own if you wish.


Copyright © 2009 iCapper.Com