How accurate is your form provider?
 

Recently we had some correspondence with a website member that proved to be quite interesting as it exposed the possible level of blissful ignorance and acceptance amongst the general punting public regarding the accuracy of form providers.

We are the first to admit, that to correct everything and maintain a database that is free from all error is a near impossible task given the daily avalanche of errors we ALL face resulting from faulty timing methods right through to human input errors and everything in between. It is also true that there are many that claim to do OK, at least for the short term, using the so called official form. Frankly, what else is anyone to do anyway, official form is all there is.

This last point is not quite correct, official form is NOT all there is. The serious professionals have for the longest time now produced their own form, usually starting with the so called official form and then replacing just about everything. These pros invest hundreds of thousands a year in time and money to make sure that the data they are using is correct.

That is one extreme and the other is the official form, sadly there is little in between that ordinary punters can use. There are corrected race times available such as those provided by Daily Sectionals produced by Vince Accardi) and whilst these can be a worthy investment, the ordinary punter still faces the task to actually make use of that data in a meaningful way. This usually requires a fair amount of programming skills and financial investment which is for the most apart not within reach of the average punter. Those that have access to this type of information know and enjoy the benefits that it provides. I should point out that Axis is capable of importing and making great use of all the Daily Sectionals data in a way that no other software can. This takes shape in the form of reading the data in a structured manner all the way to calculating pars so that comparing times across different tracks and distances actually makes sense.

It is difficult to convey the scope of errors without an example. Below you will find a table of 29 race time corrections that we (ratings2win) had to change. These occurred on 5/12/2013 while processing the form for next day. You might look and think, “Well that is not much from the many hundreds of past starts of each horse, 29 races is acceptable”. You might be right, had that been the case, but it’s not, each day a new list is generated often much larger; this list only deals with NEW errors, thus far not corrected. So those 29 errors will be corrected today and then again without a notice every time they come up. Over 98% of all races have incorrect times recorded whether they be overall, sectional or both. This sample includes errors ranging from 4 lengths to 0.1 lengths. If we accept that a horse is 3m long and travels 0.16sec per length, trying to consistently and accurately calculate ratings becomes erratic. We are dealing with discrepancies of up to 12 meters here and most race finishes come down to lot less. What is more, each error affects all horses in the race, and since generally most punters use at least the last few races to assess form, each timing error has the capacity to affect the form study and ratings calculations for a great deal of horses over a lengthy period of time. Mix them all together and you get ..... well work it out for yourself.



Let’s have a look



Ok if time was all that was wrong, we could all just correct it right? But there are other issues that affect the form. Margin errors, track conditions designations errors and SP prices and so on occur with great regularity, down to the distance and class designations which pop up almost as frequently. As stated above, each error plagues each horse with decreasing importance over at least 5 starts and then lingers for the next however many starts that you have. The industry can’t even provide a single uniform name convention for Jockeys, Trainers and sometimes even track names.

So what does all that mean for the Axis clients? Simply we endeavour to fix all these errors. We face the daily frustration of fixing race times, missing/incorrect sectionals, incorrect margins/track conditions, trials assigned to the wrong horses, missing settling positions, incorrect settling positions, names, distances, class descriptions and others on our client’s behalf each morning. I would be first to admit that the never ending volume of errors and inconsistencies are such that some we do miss some, but believe that our tracking error rate is less than 2%.

The value of this is such that we do not provide race or sectional times and protect most of our corrections from being exported and shared around so Axis is not for everyone! The data is available within the program to be used as intended. This protects the integrity of the data for us and our clients and as a result we all enjoy the unique edge that our corrections provide well into the future. Coupling this with intelligent market research produces a superior level of performance long-term. Having the luxury of not having to second guess the many (good or bad) performances of individual horses, wondering if they are real or if the time was out by a second or more, goes a long way towards maintaining a clear mind so that we can then undertake more detailed form study or simply build systems which in turn allows all of us to execute our bets with more confidence. The same can’t be said for most other racing software. The final proof of course lies in the industry leading results, our client’s subscription take up and rate of renewal and the fact that we offer transparent daily result summaries that anyone who is inclined can view before checking them against their data.

How do you know yours are right and theirs are wrong?

Hi jazzy,

Our results using our data v the official times data suggests that the significant investment is worthwhile. In addition experts such as Vince Accardi (who also uses our software) know like we do the level of timing errors that exist both in overall and sectional times.

I have no doubt that there are errors, (some very glaring - not even basic checking done by data suppliers?) but a few of the examples you gave are not much different - and my question remains unanswered - how do you know?

Cheers

Hi jazzy,

What we are talking about in the example cited relates to just 1 Friday's racing. Often the instances are greater. Generally speaking the more meetings scheduled the more instances apply. I don't see how I can explain it any more clearly that we already have. If you are not relying on SPR's as one of the critical and important references then you would be concerned by the magnitude and regularity of these errors.

My question is why are you using Times at all in this country when even the Ofiicial Race Course Handicapper still stick to Weights.

Cheers.

But how do you determine the correct time? - do you time a video, pull a number out of a hat or what?

And how do you even know the distances are correct?

I don't want to answer for Paul.

But the short answer, from my point of view your Standard deviation will be wider with raw data, once you make adjustments /corrections to your database information your std deviation has to narrow, that's the true measure.

Hi darky,

The OHR is certainly relevant in particular types of races. Other than for these, IMO it has no relevance other than it provides the handicapper with a trigger from which it assigns the horse weights. The OHR significantly under performs all of our proprietary ratings by a large margin.

SPR's have the ability to identify talent in a horse far sooner than traditional weight/class based ratings. That is why IMO SPR's when used and understood properly are the most accurate measureof a horses ability.

Hi Jazzy,

Timing videos or vision that is recorded directly from the TV is problematic. This is due to things such as frame loss, compression issues etc. The only way to get it right is to have the technology which is substantial, that enables a live stream feed that is then digitally measured and verified from different points in the race.

In terms of getting the distances correct, we have these independently verified. Regulation distances are affected sometimes by different rail positions and other factors. Needless to say we are confident that we have these situations covered.