OZmium Sports Betting and Horse Racing Forums - View Single Post

Quote:

Originally Posted by evajb001

Hi guys,

Paul:
Certainly understand that PIR data isn't probably enough to put together a premium speed map. However what i'm attempting to do is at least put together a solid speed map with freely available data. I understand your position given you sell said data/maps so I'd expect that your speed maps would be more accurate than mine giving the data inputs you most likely have at your disposal.

I'd be interested in your opinion Paul (and others) on what an acceptable strike rate of predicting the leader would be? i.e. I looked at a very small sample of my speed map predictions and in races <1800m my predicted leader did lead almost 38% of the time, and settled in the top 3 approx 72% of the time. I'm just wondering what type of accuracy I should be aiming for at a minimum (and please don't say 100%). I see it as a decent advantage if you know your predicted leader sits forward that often AND you rate the horses ability/class highly as well.

Hi Josh,

So if we are talking about predicting the settling position using only the normal available historical PIR data as the basis for the prediction, versus the actual settling position then you need to be aiming for 30% for the predicted leader and around 63% in the top three. These are in fact the results long term from the studies we have undertaken so what you are seeing in your results at the moment is an aberration. Of course you may not even achieve the above percentages as it depends on what algorithm you are using to achieve the output using this data but regardless of that point you won't achieve better as the data is absolute in nature.

I can tell you that our betting model achieves an accuracy in the predicted leader versus the actual leader of around 44% and a much higher top three percentage to what is described above.

You are on the right track when you say "I see it as a decent advantage if you know your predicted leader sits forward that often AND you rate the horses ability/class highly as well" however there is a lot more at play here as you need to understand and interpret the data with an overview of both speed and pace which are of course very different, as well as track bias and its affect on the race outcome. Understanding bias properly can take years to learn and is a skill that is lacking for the majority of punters; but if learnt is extremely worthwhile. Understanding pace and speed can only be learnt and subsequently applied if you possess adequate sectional data and other tools. E.g. We recognise up to twenty different types of pace when establishing the type of pace for an individual race. I know this all sounds like a lot of work (it is) and that it comes at a great expense and isn't something I would recommend unless you have the time, resources and desire to learn at this level.

I wish you all the best with your continued endeavors.