Nessie
I apologise like i said to chaz it must be the way my writing is interpreted i really meant no harm it is quite a common saying where i live and i just put it up without thinking sorry mate,

Mtoto

On the matter of sfs, I'm claiming no expertise whatsoever, but my starting assumption is that, because they are based on time, there must be a relationship between sfs and distance beaten. That is, if horse A runs a mile in 1m 40, and horses B and C finish 1 length and 2 lengths, respectively, behind A, they palpably took longer to run the same distance, and whatever the times on the stop watch for B and C, the interval between B's finishing time and 1m 40 would be approx. a half the interval between C's finishing time and 1m 40.

On that initial assumption, a quick comparison of sfs and distance beaten, using races where the same weight is carried (ie the Derby), suggests that, broadly speaking, in VDW's day horse A, beating horse B by a length, would receive a sf 2 points higher than horse B. A comparable comparison using current Form Book sfs suggests that the difference would now be around 0.75 points. (I accept that I have made these calculations from just a very small sample of races, eg a couple of Derbys from each period, cross-checked with a couple of other races, and a more extensive survey could well produce different figures.)

I've also noted what Split Second says in his introduction to "Timecheck 2003", namely that the only adjustments made to, presumably, the time each horse took compared with the standard time for the course in compiling the ratings are those to take account of (a) the going and (b) the class of horse running. The first of these seems straightforward in principle, though I am unclear about the second - ie if on the same track, on the same going, a horse in a maiden race runs a mile in say 1m 40 seconds, and a horse wins a Group 2 over the same distance in the same 1m 40 seconds, why should they not be given the same sf?

Given the above (admittedly very limited observations), on 16 April, Fayr Jag finished 3.5 lengths behind Pic Up Sticks, who was given a sf of 113. Given that Split Second quite explicitly ignores weight, on the apparent "going rate" of about 0.75 of a sf point difference per length beaten, I'd have expected Fayr Jag to be assigned an sf of about 110, but in fact he was assigned 107. In other words, from my provisional perspective his sf was, if anything, under-stated.

By comparison, on his next run on 15 May, Fayr Jag finished 1.25 lengths behind Border Subject, allocated an sf of 114, and I'd therefore expect FJ's sf to be around 113 - it was actually 112.

Overall, therefore, if anything I'd have expected FJ's sfs for the two races to show a slightly smaller increase than the five points actually recorded. Perhaps just three points, but nevertheless an increase.

But then comes the factor of weight. IF one takes the view, as Split Second used to and the evidence suggests VDW did, that achieving a given time under particular circumstances carrying 10 stone is a better performance than achieving the same time under the same circumstances with 9 stone, some standardisation is necessary. The question then becomes what factor to use in that standardisation.

In VDW's day, on the Flat the standardisation was to 9 stone, with 1lb being regarded as the equivalent of 1 sf point. On that basis, Fayr Jag's 16 April run, when he carried 9.8, would return a standardised sf of 115 (or, if I am right in thinking his actual sf for the race, vis-a-vis Pic Up Sticks, was a little understated, perhaps as much as 117. By contract, on 15 May FJ's 112 (which on my line of thinking may have been worth a point higher), when carrying 8.9 would be, at most, a standardised 108, ie markedly lower.

If one halved the adjustment factor to 0.5 of a sf point to a llb, using the returned sfs of 107 and 112 respectively, one would sill have the earlier performance the higher rated - 111 compared with 109.5.

Frankly I am not at all sure where this is getting me, but at the moment I have four very tentative conclusions:

* there must surely be a relationship between time (and therefore sfs) and distance beaten;

* if so, the relationship as expressed in Split Second figures may have changed significantly between VDW's time and now;

* again if so, there seem to be inconsistencies between how SS currently assigns sfs;

* one either believes that weight makes a difference, or one doesn't. If the former then, PURELY ON SFs, I can't at present see a case for regarding Fayr Jag's run on 15 May as an improvement on his run on 16 April.

However, I am sure those with expertise in sfs, like you and Crock, will be able to pinpoint the flaws in my reasoning.

Have you read Bellingham's piece on Raceform SF's
I could put the article up but it will cover 3 pages

Boozer

I haven't. If you would please post the source of the article, I'll track down a copy.

It was published in the Handicap book a while ago
Cant remeber when
I will put it up later

Boozer,

I would be most grateful if you could post it, as I'm extremely interested but unfortunately don't have any access to such sources.

Many thanks

Oldtimer

DAVE BELLINGHAM, Race-form's time expert on Split Second speed figures - how they are compiled and how to use them



AT the end of 2000, the Race-form speed figures underwent a fundamental change that saw the weight factor dropped from the equation and a change to the way the standard times, now called median times, were calculated.
As a result, the speed figures you see in Raceform publications now are completely different to those you would have seen in years gone by. The scale has changed for one thing, one point on each scale meaning something very different.
First of all, though, it is worth going back to the basic thinking behind the concept of speed figures, which has not changed at all.
Speed figures can tell you things that performance ratings cannot. Whereas performance ratings are totally based on an individual's opinion, speed figures make use of a constant called 'time' which, if measured accurately, cannot be distorted no matter how much you try to squeeze or stretch it.
If a five-furlong race is run in 60 seconds flat, that is a fact and no one can say differently provided, of course, the method of recording the time is accurate.
The real challenge is what you do with the information once you have it.
One thing you definitely don't do is to compare winning times with one another over the same trip on different tracks or on different days.
If one horse runs one and a half miles in 2m 35s on one track, and another horse runs the same trip in 2m 38s. somewhere else, the first horse is not necessarily the better of the pair.
All British racecourses are unique, which means the times taken to run different distances on them are also unique. With that in mind, the benchmark you use to establish just how good a performance was on the clock also has to be unique to that track.
The method used by speed handicappers is to establish a set of standard times to serve as the benchmarks.
The method used to calculate these standard times varies from person to person, but in the case of the Raceform median times it is pretty straightforward.
The median times are updated on an annual basis, the Flat median times during the winter and the Jumps during the summer so they do not change during the peak season of either code.
The reason for updating them
on a regular basis is twofold. Firstly, track alignment is changing at racecourses all the time. The moving of a starting position by a few yards here or the re-siting of a fence there can significantly alter the time taken to get from start to finish, even though the advertised race distance does not change.
The second reason is that new race distances are coming into operation at different courses all the time and they will need to be picked up when the median times are next updated.
For obvious reasons, if a certain race distance has not had a reasonable number of races run over it at the time of the update, a median time would not be reliable.
Under those circumstances, the median time is left blank until there is enough data to go on and no speed figures are generated for races over that trip.
The median time is precisely what it says. It is the median time of all winning times for a certain trip over a five-year period, eliminating races confined to young horses or fillies.
Standard times tend to be an average of the faster times run over a certain trip, after which adjustments may be made for class and weight.
Consequently, the Raceform median times are slower than their standard counterparts. Unlike standard times, however, median times are not theoretical and have been achieved by at least one horse in the history of mankind.
That is not to say that the class of horse is ignored. Adjustments for class are made in the calculation of the Split Second speed figures, but they are carried out at a later stage.
Once median times have been calculated, it is then possible to calculate speed figures for any meeting.
Firstly, the winner's time for each race is compared to the me-
dian time for the trip and a raw unadjusted speed figure is allocated using this comparison in its calculation.
There is an obvious question here. Does this not mean that this figure will always be high for races run on fast ground and low for those run on soft?
Yes they will, but they are not the figure that is published and there is one more very important stage to go through before the speed figures are allocated to each horse.
We have to work out how conditions on the track have influenced the final times and, of these factors, the going is but just one.
Other factors that may affect the time taken to run a race include moving the rails inwards or outwards, especially around a bend, or the omission of fences or hurdles. Even forces of nature like wind or driving rain can have a big effect.
To compensate for all of this, a set of 'par' speed figures have been established for each class of race and have been arrived at after plenty of research.
These par figures are the speed figures you would expect to be earned by the better winners within each class and would, in theory, be earned if track conditions had no effect either way on their final times.
Incidentally, the research showed there to be an 18-length gap between a top-class Group 1 horse and a winning selling plater over a trip of one mile.
The raw speed figure allocated to each winner at a particular meeting is compared to the par for the class of race and an average difference for the meeting is calculated.
This average difference becomes the going allowance (or track variant) and is applied to the raw speed figures earned by each winner on the card.
Separate going allowances may be established for races run on the straight and round tracks at the same meeting if it's obvious the ground was different on each, as is also the case with races over fences and hurdles at a National Hunt racing.
Once the winner's speed rating has been established, the speed figures for beaten horses are calculated with the distance
of the race having been taken into account.
Soft ground will result in a positive going allowance and when applied to the raw speed ratings will have the effect of adjusting them upwards, while fast ground will have a negative '-' going allowance and will ad-just raw speed figures downwards.
Rather than being shown in terms of 'lengths per mile,' which is what one point on the current scale means, the going allowance in Raceform publications has been adjusted to 'seconds per furlong' for the sake of continuity.
At this point many speed handicappers will make adjustments for weight, but not in the case of Split Second speed figures.
This is because research suggests there is little evidence of a link between changes in weight carried from one race to the next and a change in the horse's speed figure, at least not one that can be applied globally to every horse in training.
Add to that the fact that we do not know the combined weight of the horse and jockey, and that horses are not a standard size and shape, and it can
be seen that adjusting for weight has some major pitfalls.
All in all, it was decided that it was better to leave the question of adjusting for weight up to the individual and not factor in some theoretical weight adjustment that can have the effect of drastically altering a proper speed figure before you see it.
What do the Split Second speed figures actually show us? As previously stated, one point on the scale is equal to one length per mile, which is a much more significant margin than one pound per point on some other scales would be.
A genuinely high speed figure represents a fast performance irrespective of conditions on the day, but one high speed figure should not be used as evidence that a horse really is fast unless the horse in question has run only a couple of times.
Backing the top-rated horse blind is never a good idea even though the odd long-priced winner will turn up.
Much better to look for horses that earn consistently high figures compared to its opponents, or whose recent figures are on an upward curve.
Also worth looking out for are horses that have fairly recently
earned speed figures that are significantly higher than the par for the class they will be running in today.
Better still if they have done it more than once. Obviously, it is much more significant if a horse has earned high figures under similar conditions he or she will be encountering today, such as ground and trip.
A disciplined approach to betting is the real key to being successful but, when used properly, Split Second speed figures can help make a positive impact in the betting fortunes of the shrewd punter.
For use with the speed figures shown in Raceform publications, the par figures for each class of race are listed in the table at the bottom.
The allowances shown in the table at the foot of columns one and two should be deducted from the par figures for races confined to younger horses depending on the month.
Any race confined to fillies or mares should have three further points deducted in addition to any of the
allowances.


Class A Group 1 117 Class C 107
Class A Group 2 115 Class D 105
Class A Group 3 113 Class E 103
Class A Listed 111 Class F 101
Class B 109 Class G 99

2yo 3yo
Jan-Feb -6
Mar-Apr -11 -5
May-Jun -10 -4
Jul-Aug -9 -3
Sep-Oct -8 -2
Nov-Dec -7 -1

Boozer

Thanks. Very kind of you to post the article.

Bellingham says that one point on the Split Second scale equals 1 length per mile. And I must say that Fayr Jag's sf on 16 April reflects this: he was beaten 3.5l in a 5f race which, I suppose, SS assumes is the same as being beaten by about 5.6l in a mile race, and thus worthy of the six point difference in sfs duly recorded. Similarly, Fayr Jag's 1.25l defeat over 6f on 15 May would, on the basis Bellingham suggests, warrant the two point sf difference recorded. Thus my concern about inconsistency would seem to fall.

One of Bellingham's colleagues at Raceform, David Dickinson, makes the assumption that a length is worth 1.875 pounds over a mile, or 3lb over 5f and 2.5lb over 6f (page 10 of his "How to Compile your Own Handicap"). On this basis, at level weights Fayr Jag would have finished 2.5ls behind Pic Up Sticks on 16 April, and received an sf of 109. On a similar basis, at level weights on 15 May F Jag would have finished more like 5l behind Border Subject and then been allocated an sf of about 107, ie still a downturn on adjusted sfs.

Overall, it seems to me that the key issue is not how the sfs are compiled, though that is certainly not unimportant, but whether one takes the view that weight should be factored in. VDW did take that view.

Boozer,

Many thanks indeed, much appreciated.

Oldtimer

Hi Fulham,

I’ve just noticed that Boozer has put the article up (thanks Boozer) so this may be redundant,but it’s taken me ages to type out so I’ll put it up anyway.

I’m not an expert on speed figures but I hope you won’t mind if I comment.

I may have misunderstood you but I think it’s very risky assuming a figure of 109.5 is a downturn from 111.There are so many ways that inaccuracies can creep into the calculation of speed figures and there is more to ratings than just the bare figure.

VDW made the point about the variation in form ratings from different compilers.I think the scope is there for at least as much variation in speed figures.It also begs the question of whether improvement is necessary all the time.Some horses only have to repeat previous performances to be worthy of consideration.

I seem to remember our old friend Jock Bingham saying that one of his examples had run within 7lb of it’s best rating.I know JB may not be everyones cup of tea, but he had obviously spent a lot of time on research.

Re the weight issue .When I did my own figures I never took weight into account at all.I was heavily influenced by the Beyer and Mordin approach though.As I concentrated mainly on handicaps in those days I was really looking for horses that had run faster than I expected for that class and then seeing how the handicapper treated them and what the trainer did with them.

I suppose if I had started to factor the weight carried into the rating ,in the belief that it makes them more accurate,I would also have had to factor in things like unfavourable draw, bad luck in running, poor jockeyship and a 101 other things. Ironically, I think ,these unquantifiable things would have introduced even more potential for inaccuracy. Once I start adding extra factors into the equation where do I stop and do they lead me into a false sense of security about how accurate the figures are.?

I agree with a question that I think Mtoto raised previously.Why use weight in the calculation of the base figure but ignore the weight carried in todays race? I don’t understand that.


Re your example of the maiden and Group race winners.My own thinking is that they should have the same figure. Indeed if the Group2 race was run in the time expected it gives us a good clue to the merits of the horses involved in the finish of the maiden.

JIB,

A devastating indictment of the weight principal ?

Whether you or I or Dave Bellingham think weight matters isn’t really the point is it? I think it matters to a lot of trainers and that in itself makes it an issue worthy of consideration.

All the best.

The old Raceform figures used to be an excellent guide to the class of a race, but I stopped using them when DB started compiling them, because his reasoning is patently flawed.
It is wrong to assume that a class C race at Pontefract will attract the same calibre of horse as a class C at Ascot, something VDW indicated often, and I have proved to my own satisfaction, that whatever rating is used, that is not the case.
Similarly with Fayr Jag, (Disregarding the obvious class difference), it is wrong to compare a s/f achieved at Beverley, to one achieved at York, simply because of the calibre of horse each course attracts.

Bream

Thank you. I agree with your point about comparisons. I think the only point I'm really making re Fayr Jag is that, for those who assume a key element of the Roushayd method is the achievement of higher sfs in races of successively higher class, at the very least it is open to question whether that element was present with Fayr Jag.

On the more general issue, there are clearly two very different views, reflected in those of two of Raceform's employees, and indeed two of their ratings. Split Second" says weight doesn't matter, and reflects this in his ratings. Dickinson, one of the Raceform Private Handicappers, takes a different view and presumably reflects that in the form based ratings.

For myself, I'm happy to follow VDW on this matter. His approach (when properly understood and applied) seems to have been successful for over 25 years, and is still to my personal knowledge successful today, so why change? To others who may have found approaches which pay no heed to weight and achieve comparable levels of success, I merely say a sincere good luck - no reason why they should change, either.

JIB

Now you have found out that an increase in weight makes no difference whatsoever will you be upping your anti from £3
a point?

FAYR JAG,

Been away walking in the middle of know where since very early Sunday morning but before I left I managed to have a very quick look at this race.

Whatever VDW says Fayr Jag was the only horse I wanted to back at when I found out it had won I spent several hours spitting my dummy out at missing another good winner.

I haven`t read the past 5 - 6 pages of the thread in any great detail as yet although it is very clear FJ has come in for much discussion.

One VDW quote springs to mind, ie - what is form if it isn`t that one performance is better than another.

In my opinion when looking at the merits of FJ and SOA last runs the run of FJ was superior.

*** the bare ability rating is a guide only and I believe this race is ideal to show that it shouldn`t always be taken literally ***

FJ`s overall form to date when looking at the class of race he`s competed in shows him to be the class horse. 2 runs this season, one over the wrong trip and then one on a flat track ( prefers stiff courses ) against good class horse brings him to the boil.

Is then dropped in class when the prize is big enough and bingo.

JIB

Avoid you? As if ...

Since returning from holiday, apart from the post to which this is a reply I can't recall you addressing any post to me. My policy is to reply to any post addressed specifically to me, whether from those interested in VDW's approach or from sceptics, but to follow up other posts only when I have something I wish to contribute.

The SF Debate

Firstly, and I appreciate this may sound strange from somebody who spent the best part of 15 years compiling SF's, I don't think they have relative merit in deciding how fast a horse may run TODAY. They are purely and simply an indication of how fast a horse may run given a unique set of circumstances. Races are by and large won during the last 1 or 2 furlongs, speed figures are invariably decided in the first 1 or 2 furlongs. SF's give us an indication of the PACE of the race.

VDW was quite specific in how he used SF's, as a measure of CLASS for those horses where it may be difficult to gauge class purely on races won.

It's relatively pointless today to try and compare SF's from Split Second in VDW's era to those today. The base is entirely different, most SF compilers today try to produce a rating figure that equates with the official handicappers scale. This wasn't a consideration for Split Second in those days and even it was, then it would bear no resemblance to today as the handicapping scales have changed considerably, a 0-70 handicap in VDW's day was a good race not a bunch of selling platers like today.

Of course any speed compiler will give the same rating to any horse achieving a specific time under identical conditions at Ascot, whether it was a maiden or group 1 race, the class of race is immaterial. Where the class of race is vital is in deciding the going allowance. All we have is usually 6 winning times over varying distances and on different courses (straight/round). To isolate how much the going was speeding up or slowing down horses is usually taken from a standard compiled from a particular class of race. If we have a standard of 2 mins based on 'C' class handicaps and a seller clocks that time we will assume the going was fast by a certain degree, if a group 1 performer clocked 2 mins then we would assume the going was slow.

This calculation of the going allowance is THE critical element in compiling a SF and by it's nature is the one overriding flaw that has never been solved in races that are not run flat out from start to finish.

We might have a list of 6 races that after allowing for class broke down as:
-80 secs per mile
-60 secs per mile
-40 secs per mile
-40 secs per mile
-60 secs per mile
-80 secs per mile.

Most SF compilers will estimate that the going was slowing horses down by an average of 60 secs per mile and will base their going allowance on that figure.

But what if all those 6 races were run in diabolical times with absolutely no pace! and the track was actually riding quite quick. Result EVERY single horse running at that meeting gets a vastly inflated SF. Imasellingplater running in a 'D' handicap suddenly gets an 80 when in reality its probably run a 30! Dropped in class it looks a good thing next time.

Speed figures are more about the compilers skill in calculating going allowances than they are about form or class.

I could give countless examples of C & D handicaps that were run in a faster time than a group 1 or 2 on the same day due to the tactics and lack of early pace in most pattern races. Would you really want to bet the handicapper against the group 1 winner next time?