I first wrote this blog post nearly a decade ago based on an experience I had with Dave Schwarz, someone who had a profound impact on the way I look at horse racing, and who’s advice and ideas I’ve often taken into my own approaches.
It began with this blog post about making a power rating using a fuzzy logic method.
In this blog post I’m going to use the same four factors of Speed, Race Preference, Class and Form that was used in the first blog post.
It doesn’t matter whether these are raw factors, rankings, difference from top, impact values or some other method of measuring a horses performance, the following method will work for them all.
Power Horse Racing Ratings: The Different Ways To Create Them
Let’s start with a five horse race. Using a race with just a few runners makes it super clear to see the process of creating power ratings.
Here our are imagined runners using a ranking for each of our criteria
Our next job is to assign weights.
WTF ARE WEIGHTS?
When we talk about weights in relation to horse racing ratings, we don’t mean the weight a horse is carrying. We mean the importance of each of the ratings we are using.
The more important we consider a rating in predicting the outcome of a race, the bigger the value we give it’s weighting.
In order to calculate this weight, there are a number of methods we can use. These are:
Linear regression is one of the most well-known methods for calculating weights amongst stats folk.
If you’re not a stats person it’s kinda confusing (at least I think it is), and there’s no need to know exactly how it works.
What’s important to know, is that it’s not a good solution for horse racing.
MULTINOMIAL LINEAR REGRESSION
Yeah, I know, it’s a mouthful.
I like to say this is an improved method of linear regression, but a stats person would almost certainly tell me that it’s just different and they each have their place.
This is the method that Bill Benter famously used to make millions in profit on Hong Kong racing, and it has huge popularity in horse racing because of that.
As well as the two methods we’ve already looked at, there are a number of other methods for calculating the weight, or importance, of a horse racing rating.
These include Decision Trees, Random Forests, Genetic Programming, Neural Networks and more.
They’re all pretty complicated.
They all require a lot of knowledge of maths and statistics.
They all can only be done using computer programs that need a fair amount of coding.
In other words, they’re not very friendly towards horse racing fans who are looking to improve the performance of their betting but don’t have a maths or statistical background!
SO THIS IS WHAT WE DO…
We use the thing that separates us from the maths experts…. our experience of horse racing.
You already know what factors are likely to be most important, because you look at racing every day.
We can use that knowledge, and combine it with basic weights, to give us a starting point. And then we test.
Start by using the weights 1, 0.75, 0.50 and 0.25 for the four horse racing ratings the most important rating gets the weight of 1 and the least important gets the weight of 0.25.
Ideally you want to keep the range of your weights between 0 and 1, but you could go up as high as 3 if you wanted to.
Here’s the table from above, but with weights assigned to each horse racing rating at the top.
All we now need to do is raise each rating to the power of the weight. You can do this on a calculate or in a spreadsheet using the =power() function.
Doing this would change the above scores to:
|Horse||Speed||Race Preference||Class||Form||Power Rating|
You’ll notice the Speed column has stayed the same because we are raising the rating to the power of 1. The other columns have all reduced because their weights are less than 1.
It’s very important to remember whether your horse racing ratings have the highest number as the best or the lowest number as the best. Because we’re using rankings in this example, it means the lower the number the better.
In the above table, you’ll also notice that there is an extra column on the far right called Power Rating. I’ve already made the power rating for these horses.
And I’ve done that because once you’ve used the weights to adjust each rating, making the power rating is very easy.
To create your power rating multiply all the weighted ratings for each horse together.
In our example, we would calculate the power horse racing rating for horse one with this sum:
3 x 2.83 x 1 x 1 = 8.49
Giving the horse a power horse racing rating of 8.49.
ADJUSTING YOUR WEIGHTS
The final step is find the best weights for your ratings. To do this perform this power rating calculation on your ratings across a number of races. I’d suggest using at least three months of ratings, but ideally a year or more.
If you haven’t got ratings to use, register at the Race Advisor and use ours.
Create your stats to measure:
- How often the first, second, third and fourth rated horses win
- The ROI from betting on the first, second, third and fourth rated horses
- The A/E of the first, second, third and fourth rated horses
Once you have this information setup, make a note of the results for the current weights, then change the order of weights and record the results.
There are 16 possible different ways you could have the weights across these four ratings, and go through all of them recording the results.
Pick the order that has the best results based on the three stats we’re measuring.
But don’t expect them to make a profit out of the box.
If you’re looking for them to find you contenders, then go for the highest strike rate, balanced with the best A/E.
This will give you the perfect jumping off point to take your contenders and then look at them in more detail to determine which are going to be the strongest to bet on.
A LOOK AT WHAT WE’VE DONE
In order to create our power horse racing ratings, this is what we’ve done…
- Choose the four most important ratings that we are measuring a horse race on
- Assign a weight of 1, 0.75, 0.50 or 0.25 to each rating, giving the most important rating 1 and the least important rating 0.25
- Checked the strike rate, return on investment (ROI) and A/E for the first, second, third and fourth best rated horses.
- Change the order of weights and repeat the process of all 16 possible combinations.
- Choose the best performing combination.
It’s taken us five steps to calculate a power horse racing rating. Having a lot of ratings can be confusing, and being able to combine them into a single rating which only you have the key to can give you a strong edge in the market.
Have you ever thought about using power horse racing ratings before? Let me know in the comments.