You are on page 1of 20

Betting Strategy: Simon Rowlands on time

analysis, part one


Betting Strategy / Simon Rowlands / 24 March 2009 / 13 Comments

Simon Rowlands is going deep, very deep, into the complex but rewarding world of
calculating timefigures in racing. This is part one...

There have been a few requests on betting.betfair.com for guidance on calculating


timefigures and going allowances, which, with a new season just around the corner, I
provide for turf Flat racing here.

The subject is a complicated one that requires more explanation and examples than are
suitable for a one-off weekly blog. As a result, the guidelines will be spread over a few weeks
and include spreadsheet examples for those wishing to have a crack at producing their own
figures.

First, though, a bit of theory.

Several factors affect the time that a horse records and can be expected to record,
including the distance of the race, the course, the weight carried, the slowing effect of the
surface and the horse's ability. A standard time, adjusted for these factors and therefore
directly comparable between different courses and trips, should be used as a benchmark. One
of the chief issues thereafter is to put the difference between this standard time and a horse's
actual time into its proper context.

Do not listen to people who maintain that times are best left as "pure". That misses the
point entirely and is usually an excuse for leaving out some vital component of information.
A time in itself is meaningless: 120 sec would be abysmal for 5f but is usually good for 10f,
for instance. Times assume significance only when put into context, and that means adjusting
the raw time for the circumstances in which that time was recorded, among other things.
One component that some choose to leave out is the effect of weight, which can be difficult
to pin down. "Difficult to pin down" is not the same, however, as saying that weight has no
effect. In this, I am with Sir Isaac Newton and the laws of the universe, rather than with those
who take the masking effects of other factors in horseracing as evidence that weight is
inconsequential.

The effect of weight is also intrinsic to horseracing analysis because it is the means by which
differences in ability are offset for handicapping purposes. There are theoretical alternatives
- such as starting horses at time intervals or from staggered starts - but, for what I hope are
obvious reasons, they remain impractical. We are stuck with weight as a component, whether
we like it or not.

The weights that horses carry are measured precisely in pounds, the apparent abilities of
horses are also often expressed in pounds (ratings), so if you can express time accurately in
pounds you have more or less cracked the problem.

You could, if you were so minded, express your findings in a number of other ways -as time
or as lengths, for instance, though both involve a different set of assumptions - but weight
itself should not be ignored.

Longer distances, softer going and more demanding tracks lead to slower times; less
weight carried and greater ability lead to faster times. A vital part of the following is the
effort taken to normalise for these effects, from which an allowance for the apparent effect of
external factors like going, wind and precipitation can be deduced.

Some of the information from this example, taken from Dewhurst day at Newmarket in
2008, is self-explanatory. The shaded areas require manual input, while the equations in the
non-shaded areas should be copied down for the other cells in that column. These are the
calculations and inputs that may be unclear:

* Cell D5 is standard time (my own), of which more at a later date.

* Cell E5 = C5-D5 (time above/below standard time)

* Cell F5 = =E5*(1400-(900/B5))/(AVERAGE(C5, D5)), converts time above/below


standard time into pounds. This is a slight simplification of an equation which fits pounds per
length to the margins between horses in standardised conditions.

* Cell G5 is the weight the horse carried less than weight-for-age. For example, a fully
mature horse carrying 9-0 would be given as +14 (14 lb less than 10-0/wfa). There is an
argument for ignoring weight-for-age (though not weight) in time calculations, but ability
ratings would have to be adjusted accordingly. If you do use wfa, do not use the official scale,
which is hopelessly inaccurate in places.

* Cell H5 is the form assessment of the performance in question, expressed in pounds.


Whether you use your own or valid alternatives (Timeform's are shown in this example), you
should use the same ratings consistently. Some people use so-called "class pars", often
expressed as times, but this way of addressing the influence that ability has on times is crude
unless adjusted substantially. If you recognise that ability is an important component then you
should recognise that the more accurately you assess ability the better.

* Cell I5 = SUM (F5:H5) normalises for time compared to standard time, weight carried and
ability.

* Cell J5 = going allowance. Start with the minimum value of Column I (=MIN(I5:I11), but
consider the possibility of higher or lower going allowances which will in turn result in
higher or lower timefigures in general. In particular, scrutinise the performances of horses
whose resulting timefigures are good and see if they could be/should be higher or lower. The
use of the minimum value in Column I as a starting point is based on the likelihood that at
least one of the winners on a card will have run in a truly run race and recorded a time in
keeping with its ability. This assumption holds true in the majority of cases - sectional
analysis supports this - but becomes more questionable the smaller the sample you deal with
(such as when using different going allowances on different parts of the course). The going
allowance is essentially an expression, in pounds, of the level of ability that a horse would
need to show to equal standard time, carrying wfa in a truly run race. The lower the going
allowance figure, the quicker are the conditions.

* Cell K5 = H5+J5-I5 is the winner's timefigure deduced from the preceding information.

* Cell L5 = K5-H5 is the amount in pounds by which the timefigure is better or worse than
could be expected given the ability of the horse in question.

Next week, I will demonstrate how timefigures for beaten horses can be calculated.

ShareThis

Comments (13)
1. Charlie | 25 March 2009

Hi Simon
Just to let know

Cell F5 Formula is written as (C5, D5) when it should be (C5:D5)

Look forward to the next article

Cheers

2. Ian | 25 March 2009

Hi Simon
I've enjoyed your introduction to calculating timefigures and look forward to future
instalments, however, I do feel a bit disappointed. I understand that the ability of the
horses obviously has to be accounted for but surely the results of this method are far
too heavily influenced by the "ability ratings" to achieve a true time figure?

3. Simon Rowlands | 25 March 2009

Hi Charlie.

Good spot, though you will find that the instruction I gave works for two (rather than
more) values, which is the case here.

Hi Ian.

Glad you have enjoyed it thus far. Timefigures do not have to be dictated to entirely
by ability ratings - I acknowledged this with the comments in describing the J5 Cell -
but a horse's ability to run a fast time WILL be limited by its athletic ability: that is a
simple fact. Even those who advocate using "class pars" are acknowledging this,
albeit in a more basic way.

If you do not place recorded times in the context of the times that can be expected of
horses with those abilities you have no way of establishing a meaningful going
allowance. You could over-rate or under-rate an entire card to a sizeable degree,
which has indeed happened frequently with people who ignore this vital component in
horseracing performance.

You could, for instance, rate a workmanlike winner of an ordinary handicap at


Goodwood the equivalent of a Group performer, as I have seen happen, when the
likelihood of that being the case is exceedingly remote.

There are other instances where time analysis can shed a great deal of light on areas
where form analysis may get it wrong.

The instance I particularly recall is the 2-y-o maiden at Sandown in 2002 won by
Rimrod from Oasis Dream. A form handicapper (which is what I was at the time
myself) could easily have rated that anywhere between 80 and 100, and was likely to
have rated it somewhere near the middle of that range.
But the time was outstanding and suggested strongly that a positive view - a very
positive view - was justified. Events showed that it indeed was: the winner ended up
being rated 115 by Timeform and the second was one of the best sprinters of recent
years.

I think there is a greater danger of the level of time ratings being misleadingly
dictated by the assumption that at least one of the races has been truly run on a card, a
method which many other time analysts use but which I treat with a pinch of salt. It
usually holds true, but not always, and I use analysis of overall times in conjunction
with analysis of sectionals to guide me wherever possible.

It is a false dichotomy to regard time analysis and form analysis as two entirely
separate disciplines. Time and form are inextricably linked (let's not forget that
margins between horses in the UK are a direct conversion of time, besides anything
else), and one requires an appreciation of the other in order to work properly.

Simon

4. Ian | 26 March 2009

Thanks for the reply Simon. I see what you mean, I prefer to use previous time ratings
of the first few finishers in each race in order to factor in ability, while being aware
that previous time ratings may not represent the definitive ability of each horse. What
is your opinion on second per furlong going allowances?

5. Ian | 27 March 2009

One further point. I understand your reasoning that time and form are inextricably
linked and I know that not every race is truly run, however, when I've looked at
Timeform timefigures, which I think use a similar method, there is regularly a
disproportionate amount of runners running a timefigure as good as their ability
rating, even at the high class meetings.

As I've acknowledged, I know not all races are truly run, sometimes a whole racecard
will contain falsely run races but it can't be right that so many runners, even over their
entire career, never run a timefigure to match their ability ratings.

The example you give of Oasis Dream seems to happen all too rarely, very few horses
run Timeform timefigures better than their ability ratings, even by smaller margins.
You might not agree but it seems to me that this method of using time ratings is
mainly to substantiate the ability ratings, which makes them less valuable in their own
right.

6. Simon Rowlands | 27 March 2009

Hi Ian.

Sorry for the delay in responding, but I have been away for a while.
Seconds-per-furlong going allowances achieve the same job as pounds-based going
allowances but express the concept in a different way.

As mentioned above, "...You could, if you were so minded, express your findings in a
number of other ways - as time or as lengths, for instance, though both involve a
different set of assumptions - but weight itself should not be ignored."

A time-per-distance going allowance will place the raw time in the context of the
apparent slowing effect of the surface and a weight-based going allowance does
precisely that as well.

The choice is yours, in other words. However, at some point you will have to convert
the former into a weight measure or ignore weight altogether. The latter is not an
option, and you might as well avoid the former by dealing in weight throughout, in
my view.

I am not in a position to comment officially and definitively on how Timeform


compile their timefigures. The methodology is not the same as expounded above, but
there is, as you would expect, plenty of common ground.

As we both acknowledge, athletic ability limits the time in which a horse can run a
race. But our knowledge of a horse's athletic ability can sometimes best be defined by
that very process: by the time in which a horse runs a race.

I suspect that one reason why Timeform form ratings and time ratings coincide so
often is not so much because the time ratings are mimicking the form ratings but
because the form ratings are being guided by the time ratings. Rightly so, in my view.

If, in the instance cited above, I had not bothered to consider the comparative time of
the Rimrod/Oasis Dream race and just rated the race on form-handicapping principles,
I would have rated it in the 80s on race standards and about 90 on breeding standards.
But I rated it higher because of the persuasive evidence of the time. If I had not then
the time rating would indeed have exceeded the form rating to a sizeable degree.

Both form and time can help in identifying the ability a horse possesses, and it is
wrong therefore to favour one to the exclusion of the other.

The spreadsheet example above was, by necessity, somewhat simplified. I


recommend a more nuanced approach to deciding upon a going allowance than one
which assumes at least one race has been truly run, and I allow time ratings to exceed
form ratings more often than do some others.

But I allow time ratings to exceed form ratings to what I consider to be a realistic
degree rather than to ignore form altogether and risk having time ratings that are
divorced from reality.

Simon Rowlands

7. Ian | 27 March 2009


Ok, thanks very much for explaining all that. I'll leave you in peace now, until the
next instalment!

8. Charlie | 27 March 2009

You should be on racing tv explaining this subject Simon in my opinion

9. Simon Rowlands | 28 March 2009

I didn't think it was quite THAT bad, Charlie!

More seriously, I have never really seen myself as a natural for tv work, and anyone
who witnessed my few attempts on RUK a couple of years back is likely to agree. I
am also not sure that time analysis in this detail would translate to the screen at all
well.

Broadcasting is a very different discipline to writing, which is what I am comfortable


with. In particular, you need to have instant recall and facts and figures on the tip of
your tongue, as well as to be able to present those facts and figures both coherently
and informatively/entertainingly.

Too much Stella over the years and the more recent habit of looking most things up
on a computer rather than carting it around in my noggin seems to have dulled my
ability in this area.

Radio is similar, but you can at least refer to copious notes on air (the rustling sound
you may hear in the background is not always just another packet of crisps being
devoured).

I had a lot of respect for many of the people on racing tv before I had a go at it, and I
have even more respect for them now. The likes of Nick Luck, Graham Cunningham,
Richard Hoiles, Gary O'Brien and Lydia Hislop (to name but a few) are tremendous
assets to broadcasting on the sport in my view.

Unfortunately, people seem to regard such individuals as "fair game" when they go on
telly - as distant and as unlikely to be aware of hurtful remarks as a Piers Morgan or a
Jade Goody - rather than as people closely connected to the sport who are trying to
present it (and themselves, admittedly) in as good a light as possible.

A thick skin seems to be nearly as much of a requirement of broadcasting as an in-


depth knowledge of the subject matter and articulacy. That, as much as questionable
editorial values, is why we have some of the second-raters (fill your own names in
here...) on our tellies that we do.

Still, if your comment was meant to be a compliment, rather than an invitation for me
to relive my personal hell in front of the cameras, then thank you very much!
Simon

10. Charlie | 28 March 2009

Hi Simon

It was both a compliment and a suggestion

Something like this does not have to be done live, so you should probably be able to
refer to all the notes you need

US handicapper/owner Jeff Siegel did a very good series called Handicapping 101
and you could be the next Jeff Siegel :)

Cheers

11. Simon Rowlands | 28 March 2009

As it happens, RUK have just stated that they intend doing a "guide to handicapping"
on Monday. So perhaps I was wrong, and doing that station a disservice, to suggest
that discussion of matters like these may not translate to television.

I just hope that they acknowledge that handicapping (in the UK sense of the word)
needs to be a lot more sophisticated than guessing which horse "ran to form" and then
basing the assessment of the whole race around that assumption.

Believe it or not, that is what a prominent pundit on the channel once claimed!

Simon

12. Charlie | 29 March 2009

Simon

I listen to pundits all time on tv, i have read and read articles by them all time

Only a small percent offer anything of value, the rest imo, would receive their P45's
Betting Strategy: Simon Rowlands on time
analysis, part two
Betting Strategy / Simon Rowlands / 01 April 2009 / 15 Comments

Our racing expert continues his sermon on the value of time analysis.

The following article is a continuation of last week's, which dealt with timefigures and
going analysis. It should be viewed in conjunction with the guidelines given in that article.

Having settled on timefigures for the winners of the respective races, as in the example given
last week, those for beaten horses need to be deduced. This is not a straightforward matter
due to the fact that times for those individual horses are not known (at least in the UK) but
need to be calculated from other information.

To make things worse, the BHA recently changed their methods of converting individual
time lapses into margins, making this process far more complicated than once it was.

The following spreadsheet and annotations take this into account.


Cell C18 is the lengths-per-second conversion used in official results (see instructions
below).

Cell E4 = SUM($D$4:D4), then copy down and fill column. The figures in the D column are
the official margins between individual horses converted into numbers, so this calculation
leads to cumulative margins. The official conversion in use for margins beaten is: nose = 0.02
lengths; short head = 0.05 lengths; head = 0.1 lengths; neck = 0.25 lengths; all others are self-
explanatory.

Cell F4 = $C$19+(E4/$C$18), then copy down and fill column, which takes the winner's
official time in Cell C19 and calculates individual times from it and from the cumulative
margins above.

Cell G4 = (1400-(900/$C$20))/(AVERAGE(F4,$C$21)), then copy down and fill column.


This is pounds-per-second calculated not just for the winner but for each runner. The
difference in pounds-per-second between the winner and the also-rans is never likely to be
large, and you may prefer to use the winner's figure (in this case 15.02) for all runners in the
race instead.

Cell H4 = ($C$22+(($C$19-F4)*G4))+(C4-$C$4), then copy down and fill column. This


extrapolates individual timefigures for beaten horses from the winner's timefigure, the
cumulative margin beaten, weights carried (Column C, taking weight-for-age into account in
this example) and pounds-per-second.

A central calculation in this week's and last week's spreadsheets is the pounds-per-second
one. This is a slight simplification of one derived empirically from the margins between
horses in thousands of standardised races.

Gaps between horses in such races become a reflection of the degree to which circumstances
string the runners out. It follows that the theoretical poundage allowances needed to bridge
those gaps are in direct but inverse proportion to those gaps.

As a result, you will need a different calculation when dealing with fibresand than when
dealing with polytrack and turf. The last two can be treated as the same for these purposes,
whereas pounds-per-second need to be multiplied by 0.8 for fibresand, on which horses get
much more strung out.

As mentioned above, the calculations involved are complex in no small part as a result of
the BHA's well-intentioned efforts to make official margins between horses more reflective
of the actual physical margins involved.

The BHA would, however, be doing punters a big favour if they took the simple step of
issuing individual horse times alongside conventional margins as part of the official result.
This is what happens in Ireland and South Africa, to name but two.

Whereas once a length in the UK was a length in visual terms (whatever that is), now it no
longer is. A length has been a conversion of a time lapse in the UK for well over a decade
now, so pounds-per-length (conventional form handicapping) and pounds-per-second
(conventional time handicapping) have long since been inextricably linked.

Between 1997 and early in 2008, the official conversion in place was five lengths-per-second
on the Flat and 4 over jumps, regardless of the fact that horses will finish at vastly different
speeds under different circumstances.

The following conversions have been in place since 15th June 2008 and are needed to prompt
the figure in Cell C18 in the above spreadsheet:

* Good or quicker turf, and polytrack at Kempton, Lingfield and Wolverhampton = 6


lengths per second

* From good (good to soft in places) to Good to soft (Soft in places), and polytrack at
Great Leighs = 5.5 lengths per second

* Soft (good to soft in places) and softer, and fibresand at Southwell = 5 lengths per
second.

Next week, I intend touching upon standard times and on some of the more difficult
circumstances under which to interpret going allowances and timefigures.

Before then, I have a few days away in Prague, a city of beautiful architecture, cheap beer
and lax morals. Or so I am told. When I return, I will, like millions of others, put my mind to
trying to crack the Grand National, which is due to take place at 4:15 on Saturday at
Aintree.

I am not going to prejudge the matter - which will depend on things like the ground - but I
have to say that I was very taken by the run of last year's winner Comply Or Die at the
recent Cheltenham Festival. There is less guesswork involved with him than with many of
his rivals. Last matched at 22.0 on Betfair, he could well tempt me in.
ShareThis

Comments (15)
1. Ian | 02 April 2009

Great article Simon. I know that Ireland publish the times for each individual horse
but do they still use the 5 lengths per second conversion for the flat?

2. Simon Rowlands | 03 April 2009

Hi Ian.

No they don't. A recent study I did found lengths-per-second in Ireland ranging


between 4.38 (13f heavy) and 5.94 (6f polytrack), with the average at 5.30 l-p-s. This
makes handicapping by form and by time a bit of a minefield, though fortunately
Irish-racing.com not only provide the times by horse but do so promptly and (as far as
I can tell) reliably.

Simon

3. Ian | 04 April 2009

Ok thanks. Going back to that Dewhurst meeting, do you think it's possible there was
a significant decrease in the wind strength from behind after the fourth race? It's not
an occurence that I regularly factor into my figures and I know that the last two races
were quite slowly run anyway but it still seems very possible, particularly using the
Rockfel as a guide.

4. Simon Rowlands | 05 April 2009

That is a possibility, though I would not state it more strongly than that. Timeform,
who had a man on the track, noted that the wind was almost directly behind and quite
strong but not that it died down later in the day.

I deliberately chose an example on a straight track, on which the effect of the wind
was likely to be uniform and differentials in going unlikely to be a major factor, as I
wanted to keep things fairly simple in this example.

Like many time analysts, I suspect, I find the effect of the wind difficult/impossible to
gauge, especially when it is gusting. I usually ignore what seem to be small effects
and estimate a poundage adjustment for bigger ones, using a crude means of allowing
for whether the wind is with, against or across the runners (or a combination of these
depending on where on the track the horses race).

In my experience, a wind that slows the runners down by 25 lb into their faces (quite a
strong wind, probably similar to on Dewhurst Day) may speed them up by about 10 lb
when behind and slow them down by about 5 lb when side on. I know that Timeform
uses vector analysis in arriving at their wind-adjusted timefigures, and in this they are
a long way ahead of me.

It is a subject I would not mind looking into more, but it is impossible to get away
from the possibility that a wind reading might have been very different at the stage at
which a race was run than otherwise during the afternoon. A degree of educated
guesswork is unavoidable imo.

5. Ian | 05 April 2009

Yes I agree, it is that uncertainty that is used by some to ridicule the merits of time
analysis, which is foolish because there is often a degree of ambiguity in any method
of evaluating form.

As split times are readily available for American racing, have you ever used your
knowledge of sectionals to pit your wits against the Sartinists in the U.S. markets?

6. Simon Rowlands | 05 April 2009

Wise words.

I have done little more than dabble in US racing and not at all in terms of sectionals.
Sectionals are of limited use unless you can place them in context, such as provided
by form and outright time.

As it happens, however, I am likely to be taking more of an interest in global racing in


the future, with the help of a few automated processes, and US racing may well be on
the agenda a bit further down the line.
I look forward to taking on the Satanists if that comes to pass. :-)

7. Simon Rowlands | 07 April 2009

As a footnote to the above, a correspondent has pointed out a couple of examples


where the lengths-per-second used officially is by no means clear, such as at
Folkestone on March 31st, when the ground was officially "Good" on the straight
course and "Good to soft, good in places" on the round course.

6 lengths per second was used throughout the card despite the fact that the round
course was predominantly Good to soft and many punters are likely to have assumed
that 5.5 lengths-per-second was applied. It transpires that, according to the BHA, "if
there are variations, the straight course calculation is used."

While I stand by my belief that the BHA should be providing punters with race times
for individual horses, I have to agree with my correspondent that in the meantime they
could make the situation more user-friendly by the simple step of stating which
conversion has been used.

Perhaps someone from the BHA would care to comment.

Simon Rowlands

8. Robert | 07 April 2009

These and many other issues were pointed out to BHA when they were undertaking
their "consultation" last year for the changes. No one asked the end user / customer of
this data, ie the punter, for their views. BHA mean well, but have made things more
complex for the more aware punters and for no gain. The old time length method,
once explained (and it wasn't), was simple and reliable. All flat racing, irrespective,
was 0.02 and jump racing 0.25 seconds per length.

Now you have to check to see exactly what values are being used and that the clerk
has remembered to make any changes, or not.

There was a case on the Newmarket July course last year where the official going
changed 3 times in the one meeting. For none of the races at the meeting did the timed
lengths per second agree with the arbitrary official ones. So what is the point?

9. Simon Rowlands | 07 April 2009

Thanks, Robert.

I am not absolutely sure whether your concluding question is a rhetorical one or one
that you would like me to answer! If it is the latter, then I would just say that
occasional deficiencies in the data are bound to happen, and as long as they are
occasional then I for one still see the point. I must emphasise, however, that I see time
analysis as part of a wider, multi-dimensional, approach to horseracing, rather than as
an end in itself.
Even with occasional deficiencies in information, time analysis can help you identify
good performances that might otherwise have slipped past unnoticed (as well as
possibly putting you away at times), it can help you gauge the effect of going/wind etc
on performances, and it can definitely help you to understand why what has happened
in a race has indeed happened.

It is, nonetheless, frustrating that the BHA, which seemed to mean well with its
amendments to the means for returning official margins, choose not to go the whole
hog and provide the basic information on which everything depends. I am led to
believe that this has been proposed, and may even happen some time, but the mills at
Shaftesbury Avenue do seem to grind exceeding slow...

Simon Rowlands

10. itooamtrading | 18 May 2009

Hi Simon,

you say the following:


"The official conversion in use for margins beaten is: nose = 0.02 lengths; short head
= 0.05 lengths; head = 0.1 lengths; neck = 0.25 lengths; all others are self-
explanatory."

Where did you get this information from as I have also encountered the following
distances on the RacingPost website:

snk = shortneck;
dist = distance;

The confusing thing about the RP codes is they also use the following symbol
¼ which is presumably the same as a neck ("nk")? Would be useful if you have a
reference which explains the exact unit for each of these?

11. Simon Rowlands | 18 May 2009

Hi itooamtrading.

I got that information from BHA handicapper Matthew Tester in March. I had been
using something fractionally different myself until that point.

I suspect that "snk" and "1/4" (rather than neck) are used only outside UK: let me
know if you find differently.

"Distance" is an indeterminate amount over a certain threshold. It is (or at least used


to be) > 30 lengths in UK but > 25 lengths in Ireland.

As this could cover anything from 31 lengths to infinity in UK, I do not attach ratings
to it unless I can check the time lapse myself.
Different means are used for converting time lapses into margins around the world,
which is something which anyone looking to analyse time and form elsewhere needs
to be aware of.

Simon

12. itooamtrading | 18 May 2009

Hi Simon, thanks for the quick response. I have until now wasted much time trying to
discover what these units amounted to (many unanswered emails to the BHA). So
thanks for these. I, like yourself, had used rough estimates - I have just updated my
own stuff to these values you have provided. It is argued that converting back from
units back to times causes extra error... But I suppose you just got to use the best of
what is available. I wish they did record actual times as that would make life so much
easier.

I have just done a quick search and found an example of SHD, take a look at Newbury
on the
16 May 2009, the 17:30 race (horse: Ryker).

If you want any more examples I can send you a list, just email me at my name at the
"moc liamtoh" bkwards.

13. itooamtrading | 18 May 2009

...my above post (when it eventually appears...) should have read SNK not SHD but
I'm sure you worked that out :)

14. itooamtrading | 19 May 2009

it appears I really was asleep when writing that post. I meant to write "many
unanswered emails to the RacingPost". I have actually found the BHA to be most
helpful, I don't want to wrongly discredit them!

15. David W | 14 October 2009

Hi

I have read with interest your three articles on time analysis.

The pictures that go with the examples are missing or have been moved.

Can you bring them back to life?

Many Thanks
David
Betting Strategy: Simon Rowlands on time
analysis, part three (subtitled: When the
going gets complex)
Betting Strategy / Simon Rowlands / 08 April 2009 / 8 Comments

In the third of his series of articles on time and going analysis, Simon Rowlands explains
a complex process to get you on the road to value.

The following article is a continuation of those about time and going analysis from the
previous two weeks and should be viewed in conjunction with them.

The importance of standard times must be obvious from what has come previously.
Unfortunately, deriving accurate standard times is a complex process, not easily explained,
though there are possible shortcuts.

What you need is a sizeable body of past results for winners' times at various courses and
distances, which should then be adjusted in reverse for the various factors - such as age,
weight, ability and going - that will have affected those times. Provided you have a big
enough sample of these normalised times, you should be able to derive meaningful standard
times from them.

At this juncture one crucial characteristic of times should be pointed out, namely that
they cannot be expected to be normally distributed. It is far easier to run slowly than to
run fast, and races are a test of which horse gets from A to B first rather than which horse
records the fastest possible time.

The upshot is nearly always a skewed distribution, and your methodology should allow for
this. For instance, average times and record times are inappropriate for calculating
standard times, and there is plenty of evidence of the folly of using them.

Sounds difficult? Well, one of the shortcuts possible is to take a set of existing standard times
- Timeform's are the best I know of and can be back-engineered from their timefigures with
a little effort, though others are more openly available - and to apply them to past results.
Anomalies, if they exist, soon become apparent, though it is again a lengthy process.

The importance of the going allowance must also be obvious from what has come previously.
To a degree, this is necessarily a subjective matter on account of the small samples (usually
only six or seven races) available. Having at least one winner running a time in line with its
form is a convention that is likely to be right more often than it is wrong (due to the
proportion of truly run races). But sectionals can help to identify when exceptions have
occurred.

It is important when choosing a going allowance to consider the possibility of there having
been different types of going on different parts of the course. Unfortunately, different going
on the straight course than the round course will mean even smaller samples and more
subjectivity in deciding going allowances and timefigures than would otherwise have been
the case.

It may also be unclear where precisely the different goings apply, though guidance can be
sought from going maps as provided by TurfTrax and from going stick readings.

If the divide seems to be clear, such as between the straight course and the round course, then
that should be reflected in the going allowances chosen. In particular, the going allowance for
races started on the round course should factor in what proportion of the race took place on
the straight course.

For instance, if the correct going allowance for the straight course appears to be 80, and that
for the round course 120, then the going allowance for a 10f race started on the round course
in which the final 4f takes place in the straight is: ((120*6)+(80*4))/10, which is 104.

For a 16f race under the same circumstances the calculation is ((120*12) + (80*4))/16, which
is 110.

This may all seem complicated and, in places, suspiciously subjective. But with the help of
spreadsheets and with practice it becomes both easy to grasp and quite intuitive. Practice
may not make perfect in this case, but it does move things in the right direction.

As I mentioned, when writing this series of articles on time analysis was first put to me,
interacting with results in a structured and methodical way has benefits over and above the
immediate figures that are produced. I know of no better way to get on top of horseracing
results than to rate times, or form, or ideally both.

It is time to go forth and multiply (and divide, subtract, add, sum, average, and the rest of
it...).

***

Readers with long memories may recall I took an interest in Betfair's First Season Sire
market (found under "Horseracing", "Breeding & Bloodstock") this time last year. I proved
right to oppose One Cool Cat, but wrong to side with Exceed And Excel, though the latter
came up not all that far short.

I am going to stick with my reasoning from 12 months ago and throw my lot in with a couple
of freshmen who have got plenty of named two-year-olds already and a certain amount of
speed, precocity, or arguably both. This year's fancies are Avonbridge (two placed runners
already) and Footstepsinthesand.

If nothing else, it should be a fairly cheap way of taking an interest in the action on a daily
basis over the next seven months or so!

ShareThis

Comments (8)
1. Jack Houghton | 08 April 2009

A superb series of articles Simon.

And whilst those who produce their own form/speed ratings might not agree
completely with everything you say, I'm sure all would agree that, had we had these
articles as a starting point when we first started trying to produce our own, it would
have saved us a lot of time, effort and wrong turns.

Brilliant.

2. Ian | 08 April 2009

Agreed, the best in depth explanation of time analysis that I've ever read (and I've
read most of what's out there), well done!

3. Charlie | 08 April 2009

Well done teach ;)

4. Simon Rowlands | 10 April 2009

Wow! Many thanks.

Those are possibly the nicest things that anyone has ever said about me, though the
music teacher who described me as "the next Mozart..." at the age of seven runs it
close.

I just hope you lot prove to have been more on the money than she was: I gave up the
piano in favour of football a few months later... :-)

Simon

5. Ian | 19 April 2009

Simon, have you come up with a standard for the Kempton 1m1f distance yet? It
hasn't been used much so there obviously isn't a lot of data but something doesn't
seem right about it. I read James Willoughby say he thinks in reality it is probably
short of 9f and at this stage I agree with him, in fact it seems to be considerably
shorter.

6. Simon Rowlands | 21 April 2009

Hi Ian.

Sorry for the delay. I have not come up with a 9f standard at Kempton for the reasons
you note. Any estimated standard would need to be viewed in the context that 9f races
take place on the inner loop, whereas 1m races are run on the outer loop.
The time of Mafaaz's race on March 18th is not instantly indicative of a short
distance, but that was a falsely run race. If you can point me in the direction of any
clear anomalies that support your contention then I will look into it.

Falsely run races at beyond 1m were a big problem for analysts of overall race times
at Kempton in the early days, and they resulted in some miscalculations regarding
standard times (for reasons spelt out in the above articles).

For instance, Racing Post at one stage had over 70% of the fastest times on a card
coming at 10f+, whereas less than 30% of the races had been run at such distances.

I believe that RP standard times have been recalculated since and are considered to be
a better guide now than once they were.

Regards

Simon

7. Ian | 21 April 2009

Apologies Simon, I'm beginning to confuse myself now! I was writing about that
Kempton distance from memory, which obviously isn't what it was. I should have said
JW thinks it is actually a LONGER distance and I think he may be right.

I take your point about Mafaaz's race being slowly run and I also think that the few
other races that have been run over the trip haven't been particularly fast, but even
allowing for that, the times seem a bit too close to the 10f ones on the same inner
loop.

Campanologist's race on 15/03/08 seemed to be run at a reasonable tempo but again


the time seems very slow for the distance. I suppose it's possible that the inner loop
was riding much slower that day, there is sometimes a difference between the two
circuits, but I still remain sceptical for now.

8. ian | 24 January 2010

Hi simon,

You mention back engineering Timeform standards. Do you have some pointers on
this? Also how do you adjust for different class of animal (ie class pars)? Would this
differ from Mordin/Beyer etc? I'm looking to examine different ways of calculating
figures for comparison and am looking for different standards than RP....I ve tried
calculating my own but always obtain illogical results, probably due to sample size or
the way ive accounted for cleass/weight etc

Thanks for your help it is a subject i spend far too long thinking about!
Cheers Ian

You might also like