There are many ways to analyse a horse race and one method is by using head to head analysis. This is a simple technique that can be advanced if desired, however it is seldom used and the principles are easy to understand. We should have the knowledge in our arsenal of tools for both of these reasons and this is the purpose of the article.
The concept of using head to head analysis is this:
- We take each runner in the race and estimate how likely it is to perform against 1 other runner.
- We repeat this process for every possible combination of two horses in the race.
- We add all the probabilities up that each runner achieved to get the final probability of each horse winning the race.
A more practical example would be a 3 runner race with horses A, B and C racing against each other. We would estimate the following probabilities:
- A beats B
- A beats C
- B beats A (this has already been done in A beats B)
- B beats C
- C beats A (this has already been done in A beats C)
- C beats B (this has already been done in B beats C)
Each horse has now got two probabilities that we add together (and then normalize) to give their final chance of winning. Of course it doesn’t have to be probabilities, you could perform the same process with ratings.
I am going to take an example from the todays races in the members area of the Race Advisor and use the process with ratings.
The VDWF rating is the one I am going to use in this example and I have sorted the runners so that the highest rated horse is at the top. The VDWF is the top figure of the two ratings each horse has in this column (the bottom being the VDWA).
In this race there are 8 runners and so there are quite a few comparisons (8 x 8 = 64 combinations). To determine the edge a horse has over another runner we are going to simply use a method of division. Again this is easier explained with an example.
Cordillera has a rating of 100 and Lady Chloe has a rating of 100. This gives:
100 + 100 = 200
To get Cordillera’s chance of beating Lady Chloe we take her rating and divide it by the total for the two runners:
100 / 200 = 0.5
Cordillera has a 0.5 (or 50%) chance of beating Lady Chloe based on this rating. We perform the same calculation for Cordillera against all the other runners. This gives her the following chances of beating each runner based on this rating. I have put the chance in brackets after the runners name to show each horses chance
Cordillera (0.5) vs Lady Chloe (0.5)
Cordillera (0.5) vs Melancholy Hill (0.5)
Cordillera (0.5) vs Symphony Of Love (0.5)
Cordillera (0.5) vs The Absent Mare (0.5)
Cordillera (0.51) vs Freedom Flyer (0.49)
Cordillera (0.52) vs Red Zeus (0.48)
Cordillera (0.71) vs Fine Style (0.29)
This gives us the chance for each of Cordillera’s head to head possibilities in this race. We then add all of her scores together:
0.5 + 0.5 + 0.5 + 0.5 + 0.5 + 0.51 + 0.52 + 0.71 = 3.74
Her final score is 3.74, we repeat this process for each of the runners and then we can either use the final score as its own rating or we can normalize them and create an odds line from it.
A good rating to use this with in the Race Advisor members area would be the RL figure which predicts the horses speed in the race.
This will get you started thinking about the possibility of head to head analysis and when you are comfortable with using it then you can take it to the next level. The next level would involve using a monte-carlo simulation. This takes the horses rating and the standard deviation and then “runs the race” as a simulation as many times as you want, maybe 10,000 or 100,000 times and each time it records the winner. At the end of the simulation you have a probability based on the amount of times each horse won the head to head race. This process is again repeated for each combination and then you can combine the results to give a final odds line.
We will look more into the monte-carlo version of this methodology in the future but for now I wanted to get you started using this racing analysis strategy.
Great Michael this is what i want i know it might be time consuming for you but no point me looking at these cards and they mean nothing maybe add a beginners section to the site where you can guild people in the wright direction like above explainning and giving a break down of what it is,i will be out most of the weekend but when i have some free time i will have a good practice of calculating the ratings
A good idea Mario, I shall look into making a beginners area inside.
Would it be easy to design a spreadsheet where the runners and their rating could be entered manually and the resulting calx made automatically?
Hi Brian, yes I can’t see any problems with creating a spreadsheet that does that.
I am sorry but I cannot see what this achieves. I have used Microsoft excel to work out the figures for 50 races using the Racing Post OR ratings and all it does is to list the horse with the highest OR rating at the top of the list with the highest figure at the side of it and the rest of the horses in the order of the OR ratings.
If you convert into odds, it might help you find value, which is the main point over the long term,
Hi Joe, yes it will list the horses in order of rating but when you normalise the figures you will have a much better figure to create your odds line from due to the differences that are now represented between each horses figures.
In the next article I am going to right on this subject we will introduce using the monte-carlo and giving each runner a deviation. e.g. each a rating of 80 means that the horse may have actually achieved somewhere between 75 and 85, for example. We use this in combination with the head-to-head approach to get better figures yet again.
looks like a sound way of assesing, my question would be, have you tested it over time and does it make any difference to the outcome of the race , compared with the original top rating anyway?
There are many different ways that you could adjust this to put it into practice. I do use a collateral rating every day in my racing to great effect.
i understand what your saying but i dont have much spare time.would be interested to see the out come for a days racecards
cheers john.
I shall look into doing a few race examples.
Dear Sir
The monte carlo simulation method appear to be very interesting. I would appreciate if it would be possible to have some examples how to proceed or direct me to a link so that I can check for it.
Best regards
Ally
Thank you for the comment Ally. You can see our free Monte Carlo software at http://www.themontecarlomanifest.com