nothing more basic to making speed or pace figures than the accuracy
of the times used to make them. When I was working on the book, “Handicapping
Speed,” I did something I had never done before, and maybe no one had
done systematically from a handicapping standpoint: I traced the handling
of time and speed from the gates out on the track, through the entire
timing system, into the photo finish booth, onto the timing slips, into
the hands of the typists and information officers, to the point where
it was passed on by the tracks to (at that time) the Daily Racing
found where every step occurred in the time/speed process and hung out
with the people who managed it. I didn’t just interview them and ask
questions; I stuck around until I was part of the furniture and learned
the routine. I spent so much time with clockers, starters, assistant
starters, flagmen, and photo-finish operators, that most of the time
I smelled like either horse sweat or a chemistry set.
doing this, I found a few flies in the ointment, a few junctures in
the process where errors could creep in, a few less-than-interested
low-level employees but, I’m pleased to report, never a dishonest man
or woman. In a business where money rides on omissions and commissions
by dozens of people in a fairly loosely knit process, that’s quite amazing.
I’m sure that intentional irregularities have occurred and can at any
time, but in my experience of spending more time with some of these
people than their spouses over a year, I saw errors happen, but never
an intentional act that would affect us as handicappers.
the speed book, I talked about sources of errors, and how one of the
photo finish examples supplied to me for publication even had an observable
error in how the negative had been laid under the thread used to superimpose
a finish line onto the final print. Since the times of each horse’s
finish are read off the negative, here was a place where consistency
under pressure is important, or else times could be misread.
the last steps of processing the results before they leave the track,
there is also the chance of typos going unnoticed when times and lengths
are transferred from handwritten slips to printed sheets or computer
files. In spite of all of the different technologies and all the hands
that the data passed through, the greatest potential for errors affecting
speed and pace handicappers appeared to be the judgment calls that are
made on lengths back of non-winners.
appearance was because I was making a major assumption: that once the
times were recorded in seconds and hundredth of seconds—that was it.
From that point on, I imagined that perhaps a typo could creep in, but
what else could go wrong?
one day last winter, a large envelope arrived from a friend and fellow
handicapper. In it, were printouts of race results from tracks around
the country—in triplicate. Times had been circled in red pen
and there were notes and an irate letter asking why one of the three
racing data suppliers was providing different times at the finish and
points of call in virtually every race. Since Equibase now provides
all data, he asked, why was this one supplier wrong so
had been using raw data files from all of the suppliers for some time
and knew the raw numbers were essentially the same, but I had never
happened to print out all three suppliers’ own representations of their
data converted to fifths, and looked at them side-by-side. Sure enough,
there frequently was a difference of a fifth of a second between
times at any given split or at the finish. There was also no observable
pattern to it—this one supplier that was “wrong” was not consistently
fast or slow—the times seemed to differ at random.
who doesn’t deal with times and speeds in horseracing might shrug—a
fifth-of-a-second isn’t much. But a fifth-of-a-second is the fundamental
unit of measure in thoroughbred racing, originally because it was the
unit of time on the clocker’s analog stopwatch. Fifths have continued
as the unit of measure out of tradition—and the fact that they are very
represent approximately “one length” in thoroughbred speed (which I
went to great lengths to explain and correct in the speed book, but
nevertheless), for eye-ball, seat-of-the-pants figuring, they are extremely
convenient and I hope the convention never changes. This is
a major example of the types of information I talked about last week,
where “improvements,” in this case going to hundredths of seconds in
the paper printouts of our data (or the Form itself), would do
no one any good.
with this in mind, the idea that this data supplier was frequently and
randomly off by a fifth of a unit-of-measure in either direction—fast
or slow—was rather a shocking revelation. Since I had raw data on hand
from all three suppliers for the same races, it just took a few minutes
to see what was going on—and it was a far bigger shock. The one data
supplier, who differed from the other two—Handicapper’s Daily—was
was the only one that was right. All other data suppliers
were wrong in how the represented fifths. I was stunned.
bet I had ever lost while entering data by hand from paper flashed before
may recognize the previous sentence as a type often used when someone
is about to spring some kind of ironic twist—however, there is no ironic
twist. Those bets are still passing before my eyes.
my own computer programs, since raw time data has been available, I
have routinely used the same little algorithm I wrote at the very beginning.
It takes the raw time in hundredths and rounds it to the nearest
fifth. This is not ingenious. How else would any handicapper,
interested in speed or pace, do it? It never occurred to me to
do it any other way.
the rest of the world does it differently. As best I can tell, except
for me, Eric Langjahr (the horse player who wrote the Handicapper’s
Daily software), and no doubt some of you who are handicappers
and who have written your own programs, spread sheets, etc. —the rest
of the world doesn’t agree.
talked about this situation briefly in my presentation at the Daily
Racing Form “Expo 2000” last spring in Las Vegas. I have not listened
to the tape, but I think if you do, you’ll hear that I was almost apologetic—I’m
not sure why. You may also hear a comment from someone in the audience,
whom I’m pretty sure was a Form employee, to the effect, “The
Form has always said it rounds down.”
hate to break the news, but what they do is not “rounding,” it is truncating.
They simply delete whatever fraction-of-a-second is leftover, back to
the lower fifth of a second.
the few brief comments from the Expo audience, I got the impression
that some thought “rounding down” was OK, since it is consistently
applied. The practice itself may be consistent, but the result
upon times displayed for the public is not.
this means, in short, is this: a horse wins a 6 furlong race in 69.99
seconds, and another horse wins a different race in 70.00 seconds.
The times are 0.01 seconds apart—one one-hundredth-of-a-second—far
less than a heartbeat. If you truncate the first decimal fraction of
a second to the lower fifth of a second, it leaves 69.8, which shows
up in print as 1:09:4. The second time of 70 seconds flat doesn’t need
to be truncated, so it would print as 1:10:0. Two horses—not a nostril
apart—are now a length apart on paper (or, about a length-and-a-third,
by my figures).
effect upon printed times in fifths-of-a-second is totally random.
There is no way a handicapper can tell by simply looking at the fifths,
or manipulating them in any way, whether the raw data was truncated
a lot (maximum of 0.19) or not at all.
may have noticed that the Form has recently added the time in
hundredths-of-a-second, along with fifths, in their charts and in the
results displayed on the Web. This may be their way of coming clean,
but as you can see immediately by comparing a few times in the two formats,
they have not changed from the method of truncating to the lower fifth.
They are not alone.
might argue that this inaccuracy has not hurt handicappers, again because
it is “consistent.” However, it is not consistent. Horses randomly
run times of 0.00, 0.01. 0.02, etc., up to the next fifth at 0.2. If
you are a pace handicapper, and one horse runs to the first call in
21.99 while another runs the same distance in 21.80, yet both are represented
as “21:4” does this interest you?
you consider the performances represented by the actual times in the
table below to be the same? Does it interest you that a horse which
ran two lengths off the pace of “45” may actually be a two lengths off
a 45.19 or roughly a 45:3 rather than 45:2—more than a length’s difference—and
that the splits can “accordion,” longer and shorter, randomly?