0

I have a very large dataset containing horse racing results. What's intriguing to me is the variance across the time taken across the various race tracks. It has led me to look into causal factors.

The one that I'm completely stumped on is how to interpret each result. For example, let's say that we have a result for Horse A;

Distance: 1100 yards

Track: Ascot

Track conditions: Soft

Class of race: 5

Time Taken: 71.02 seconds

Now, we might assume we can run a lookup of the 71.02 seconds for previous race conditions so we are comparing like-for-like. That is fair enough, I think.

However, let's assume that Horse A is next running at a new track with the same conditions, but the topography is wholly different. Is there a way I can build a catalog of averages/standards, etc. that would allow me to make a meaningful inference from the 71.02 in Horse A's previous run?

Graham
  • 1
  • 1
  • You would probably want to factor in such variables as longest decline, longest incline, sum of elevation changes, etc. - this is a very in-depth project! – The Chaz 2.0 Oct 31 '20 at 14:53
  • An alternative to The Chaz' comment is to let the results of a large group of other horses do all the heavy lifting. For example, suppose that you simplify the question to ignore whether the given horse is a front runner. Further suppose that you have looked at 20 other horses, and found that on average, their time will drop 0.03 seconds from Track A to Track B. Then that is what you can approximately expect from your horse. – user2661923 Oct 31 '20 at 14:57
  • The question as @TheChaz2.0 pointed out is in the quality of your dataset. Does it contain all relevant factors that are predictors of the outcome? If it does, then you can proceed to make an inference. Otherwise, you will get a correlation of outcomes to the inputs. But, as we all know -- Correlation is not causation. – vvg Oct 31 '20 at 16:18

0 Answers0