OZmium Sports Betting and Horse Racing Forums - View Single Post

Quote:

Originally Posted by Lord Greystoke

Thanks Paul. Informative and insightful, especially where you start to reveal some of the 'inner world' of the professional and back this up with some hard numbers e.g. the starting point for where the serious analysis begins.

I had a quick look at the adaptive market hypothesis. That is one hell of a wormhole in which to get lost in for the 'mere mortals' amongst us.What is evident to me however is the need for a 'rolling data sample' to monitor when & where the risk v reward playoff changes, and to evolve with the moving target in real time.

Cheers LG

Hi LG,

That is a great point to take from this. It is also something that we build into Axis; practical and relevant data balance.

For instance our past 20 starts are a rolling 20 for each horse. We maintain 5 years worth of 28 day months (as horses race in 7 day weeks not Calender months) in our database before the program archives the data. For trials we keep 3 years of rolling data. The same does not apply though to our pars (benchmarks) which go back around 14 years. Everything we do in terms of program structure, is set up to display only relevant information with an emphasis on as much speed as possible achieved by controlling the amount of information the program has to access. Of course users have full control over the settings and can 'horde' everything if they want, but those are the defaults and the archive allows them to test as much data back as they want.

Other than for sentimental reasons there is very little point of having the runs of Sunline and Might and Power in your database. It wastes space and slows down the database. It is also irrelevant as the market has changed and in many ways the data is so unrepeatable that its misleading. On top of that, testing ratings and so on based on races that happened 13 years ago is asking for trouble, unless the ratings have not changed in that time (always a possibility with some). This is something most punters don't take into account. There is no point testing a ratings method that for all intense purposes no longer exists because they have been superseded by a better version of the same rating. This is another reason for using Axis. If (not when) we change and improve the performance of our ratings, we actually release a file that replaces all the old ratings data in the database with ratings that are calculated with the current algorithm. This is done 'as if it was calculated on a that day basis' and requires us to keep a snapshot of each day in order to do this properly. This consumes an incredible amount of data and time to do it correctly. Notwithstanding that, we do it so that our clients are actually testing past data using the most recent algorithm instead of having to wait months before seeing any emerging patterns that the new data produces. This sensible practice is not performed by any other provider in the marketplace.

We as punters would expect nothing less and require that information for ourselves so it only makes sense to provide it to our clients as well. After all we all view the data in a different manner based on our own personal views, which in turn makes the likelihood of saturation remote.