At the end of 2013 I made the Monte Carlo Manifest as a gift to you for reading my blog.

If you haven’t had a chance to download it yet, then you can grab it here.

The name of this software is the key to what it does. It uses a technique known as a monte carlo simulation to determine what chance a horse has of winning the race.

Before I get into a method of using this software successfully, it’s important to first understand how the simulation works.

A monte carlo simulation requires you to have a variety of ratings for each runner in the race. And, as with anything to do with data, the better quality the ratings you put in, the better quality the results will be.

By default the software uses the Official Rating (OR) and the Racing Post Rating (RPR). But, if you’re a user of the Racing Dossier, then you also have access to all those ratings inside the software as well.

You can also add your own personal custom ratings in for each race.

For the purposes of explaining how a monte carlo simulation works, I will use the two default ratings.

If you haven’t got the software yet, then you can download it for free from here. Start it up and follow along.

When you open a race, by double clicking on it, you will get a screen similar to this:

You can see each horse in the race and the OR and RPR ratings.

At the bottom notice that there are two boxes, one for each rating, with 15 in it. This means that these ratings have a 15% variance.

Every rating is only an estimate which means that the true rating could be a little bit either side of the rating that has been given. In this case we are using the default setting of 15% for the amount either side of the rating.

Take Talkin Thomas as an example. He has an OR of 110. If we take 15% of 110, which can be done on your calculator using the sum:

110 x 0.15 = 16.5

We would say that with a 15% variance this rating could be between:

110 + 16.5 = 126.5

and…

110 – 16.5 = 93.5

While it can be between those two figures, we expect it to be closer to 110.

The bigger the variance for a rating then the bigger this range will be, the smaller the variance for a rating then the smaller this range will be.

It is possible to change the variance on a rating-by-rating basis, but for the purposes of this article leaving it at 15% will be accurate enough for us.

Now when you press the right arrow on the bottom right hand side of the screen you will see a page similar to:

This is the simulation running. By default it will run 1000 times. Each time it runs we take a random rating for each horse within it’s variance for each rating.

This means that for Talkin Thomas, on each of the 1000 simulations, we will take a random rating between 93.5 and 226.5 which is the range of variance we calculated above.

We do the same for each horse, for each rating.

Once we have this, we then take Talkin Thomas and compare it to each of the other horses individually and count how many of them it beat (had the higher rating). And…

We do this for every single horse in the race!

The numbers in the rating columns represent the number of times each horse has won against the others for each rating, with the percentage in brackets).

In the far right column you can see the Total, which is a combination of all ratings together.

So, using the image above, you can see that Talkin Thomas is currently the best horse (after 154 of the 1000 simulations) and has won 1034 times or 22.38% of the time.

Now that we know how the Monte Carlo Method works, the question is how are we going to use it.

Of course, there are many different ways of using this tool and I’m going to be focusing on just one of them.

I am going to be using three ratings in this example. These are…

Cst10 = Consistency of horse over last 10 starts

Contender = A contender rating taking into account a number of factors

SHorPro = A speed projection

We will also be using the Racing Post website once the simulation has finished.

First of all we need to choose a race where the majority of runners have all the ratings. This is **very important**.

A monte carlo simulation will be significantly skewed if horses don’t have ratings. If a number of horses don’t have ratings, then you can use an average rating for those runners if you want to run the simulation on the race.

As you can see above, in this race all runners have a rating except for Module who is missing a Cst10 rating.

After the 1000 simulations have finished, this is what the results look like.

Sire De Grugy has the best overall total with 16.53%, followed by Sizing Europe with 11.68%. Close behind are Captain Conan at 10.74% and Arvika Ligeonniere at 10.4%.

Before we make any decisions however we want to look closer at the individual rating columns to see those who are significantly better than others in each of these.

Sire De Grugy and Sizing Europe have the highest Cst10 simulation results with 17.23% and 16.19% respectively. Arvika Ligeonniere follows in third place with 15.3%.

Sire De Grugy has by far the best Contender rating result with 18.18%, followed by Baily Green and Hinterland at 15.47% and 15.43%.

In the SHorPro rating Sire De Grugy comes out on top again with 14.17% with the next best being Somersby at 13.79%.

The first thing that we notice is that Sire De Grugy came out on top for all the ratings we used. Sizing Europe in fact only came in the top three for one of the ratings, but it was a strong enough result to push this horse into second place in the totals.

That reduces the claim of Sizing Europe on the race.

Next we need to consider the race conditions. This is the Queen Mother Chase at Cheltenham and we know it’s a festival race, which means we can expect a lot of competition, and it’s over 2 miles.

The distance means that the speed projection is going to have slightly less importance than the other factors in this race.

We can look at the two ratings independently from this result but, to make things clear, I’m going to re-run the simulation removing the SHorPro rating so we can see what kind of difference this makes.

This change our view of the race. Sire De Grugy is still the top rated, but it is now Hinterland who comes second due to high Cst10 and Contender. If you go back and take a look at the previous results you can see that he was let down previously by the SHorPro rating where he only won 2.5% of the simulation.

In third place we have Sizing Europe and Arvika Ligeonniere comes in fourth.

We now have a much better picture of the race. Sire De Grugy looks to be the one to beat with Hinterland and Arvika Ligeonniere being strong possible bets. Sizing Europe is still being boosted by a high Cst10 rating (16.1%) compared to a very average Contender rating (8.21%). And the same is true for Arvika Ligenniere, although to a lesser extent.

So who do we bet on?

Well the first thing to do is to get an idea of odds. You can use the percentages from the total column to calculate these. But…

Please don’t use them as gospel.

They are only rough figures to give you an idea of whether you are getting a value bet or not.

If we convert the percentages to decimal odds, which is done using the sum:

1 ÷ (Percentage ÷ 100)

For each of the runners we are considering we get…

Sire De Grugy – 5.64

Hinterland – 7.45

Sizing Europe – 8.22

Arvika Ligeoniere – 8.52

Let’s take a look at the current market.

Straight away you can see that Sire De Grugy is lower than we may expect, although not hugely so that is still within the possibility of a good bet although the value may be low.

All the other runners seem to be offering value, particularly Hinterland.

Now, it’s a good idea to get an analysis overview of the race from a few different websites. Three sites you may want to consider are the Racing Post, Sporting Life and Betfair’s Timeform.

I have put the analysis from each of these websites below.

RACING POST

SPORTING LIFE

BETFAIR TIMEFORM

Looking at these three different analysis’s, we can see that they share certain information. They all agree that Sizing Europe is no longer the horse he used to be and we had our doubts about him anyway, so we can remove him from the contender list.

Sire De Grugy looks to be the one to beat, but it’s interesting to note that in the Racing Post they mention Cheltenham doesn’t seem to be his favourite course.

Two of the reviews mention Arvika Ligeonniere and all mention Captain Conan who we didn’t put in our contenders list but did perform well.

Going back to look at him, his total percentage was 10.44% which would convert into decimal odds of 9.58, making him possible poor value.

None of the analysis’s mention Hinterland, this isn’t necessarily a bad thing. It just indicates that they don’t think it is a primary contender, although our own simulation would indicate otherwise.

We are left with Sire De Grugy, who we can’t ignore due to the strong chance this runner has of winning the race, Hinterland and Arvika Ligeonniere.

Of course, the big question is how do we now bet on the race.

As this is a festival race, and I advise a different process for these race types, we may want to consider just looking for the most likely winner instead of focusing on value.

However, as part of your regular betting strategy you should always be considering value as the most important focus when your betting on selections.

Hinterland and Arvika Ligenniere are the two who look to be offering some value whereas Sire De Grugy looks to have the strongest chance of winning the race.

There are many ways to structure your bet in this race. Some examples are:

A saver bet on Sire De Grugy to break-even if this runner wins with each-way bets on Hinterland and Arvika Ligienniere.

A win bet on Sire De Grugy and each-way bets on Hinterland and Arvika Ligienniere.

A win bet on Sire De Grugy and place bets on Hinterland and Arvika Ligienniere.

Each-way bets on Hinterland and Arvika Ligienniere.

Place bets on Hinterland and Arvika Ligienniere.

A dutch bet on Sire De Grugy, Hinterland and Arvika Ligienniere.

A dutch bet on Hinterland and Arvika Ligienniere.

My personal preference in this race would be to place a win bet on Sire De Grugy and each-way bets on Hinterland and Arvika Ligienniere.

Hi Michael

A good analysis on what the Monte Carlo can do, I think you got your sums a little wrong though as 110+16.5=126.5 but I’ll forgive you. It’s a method that’s got great potential, keep up the good work.

Hi Richard, thank you for the comment. You’re right, a bit of a typo there on that sum. I’ve changed it now, thank you for pointing it out.

Hi!,

Probably the questions below highlight the fact I am a newcomer to this side of race analysis.

The article I find most interesting having worked through the example you show but

could you shed a little more detail regarding:

a. Cst10. Is this just %wins over the last 10 races or something a little more sophisticated.

b. Contender. Is this based on some trends info? Not met this one before.

and finally

c. SHorPro. Again not met this.

Little surprised the speed factor is ignored in a 2m Chase. Does this indicate you would advise not using a speed factor in any NH racing?

Any help regarding the above would be appreciated and apologies if the queries are a bit basic for the site.

Warm Regards

JayKay

Hi JayKay, thank you for your comment. The Cst10 is a rating I use along with the others, and is available in the Racing Dossier, which measures the consistency of a horses performance over the last 10 races, it isn’t a win%. The Contender rating rates horses based on a number of factors into how likely they are to “contend” in todays race. SHorPro is a speed projection rating so I did use speed

Is it possible to remove from or add to the ratings with which the software comes? I’d like to add, for instance, the Puntology ratings and the RP Betting Forecast, expressed as a percentage, and possibly remove the last 6 form figures.

Thanks.

Absolutely, you can see how to do this at the end of the instructional video at http://www.themontecarlomanifest.com/using-the-monte-carlo-manifest/

Hi Michael,

Could’nt get this to work first time around & again this time, I’m on Firefox with AppleMac & repeatedly asked to re submit Name & Email. I do have more problems with trying to benefit from Excel but can’t get a handle on that either.Sorry to sound so negative,

Regards Roger

Hi Roger, it sounds like the software has become corrupted. You need to find your app directory which you can do by searching for %appdata% in your library navigation and then removing the folder com.anonymousginger.montecarlo This will reset your software licensing and allow you to access it again.

Hi Michael,

I’ve enjoyed reading the articles and experimenting with the simulator. The object of my interest is to adapt it to racing in the USA by inserting “ratings” which would cause it to perform Sartinesque Match UP simulations. Another possibility is to produce simulations using Pizzolla’s fulcrum and related pace rating system.

Any advice on how best to accomplish this laundry list?

Appy

Hi Appy, thank you for the comment. First of all I would suggest choosing one of those possibilities to start investigating rather than all of them. Which one holds your interest the most?

Matchup and the Pizzolla rating system are closely related, but at this time I am focusing on Pizzolla’s ideology.

It appears to me that I will have to perform all the calculations and enter those results as ratings in Monte Carlo to digest. I was hoping to find a way the software could perform the calcs as well as the simulations.

Ah, yes you would have to perform the calculations first and then enter them in. There is no way to get the software to perform those calcs for you as well as the simulation.

Hi Michael,

interesting analysis. It is also possible to load the betting speed evolution ratings into the monte carlo manifest?

That would be fine.

Regards

Fabi

Hi Fabi, thank you for your message. The BSE software uses past speed ratings to predict a trend in horses performances over specific sets of conditions which means that the ratings used can’t be put into Monte Carlo software. However if you use BSE to rank your contenders and then use this inside the Monte Carlo software, this could be very powerful.

HI Michael,

thank you for your answer. I planned to load the recent rating into the monte carlo rating. With the recent rating I mean the rating that is shown in brackets after the horse when you click into a race.

Could you load this rating into the Monte Carlo software?

This rating is calculated on the fly by the software dependent on the conditions you have set to view the graph in so we can’t pre-calculate this and put it into the Monte Carlo software. You would have to add them in as a custom rating.

Michael,

what would be the best way for using US race card.

The folders Montecarlo and RCViewer do contain data folder, I assume there is where the data should be. I’m sure that your race cards are in different format than US race cards. It would be easy for us from this side of the pond, to just click the calendar and have the horses name as you’re showing in your video, and I would not mind to manually input my ratings.

Thanks for a well thought app, and as you guys says:

Cheers

G.

The data is provided directly from our databases so you would want to open the US races we have inside the software and then manually input your ratings like is shown at the end of the video. You could also simply find a race with the same number of runners and assume the top one is horse 1, the second top is horse 2 etc… and then enter your own ratings for them. The horse name isn’t important, it is the calculations on the ratings.

Michael,

excellent little program. I realize that I had to use the same number if horse. It’s a bit of time consuming and I wish the export feature wold be in Excel rather than CSV. My ratings are based on the first 3 fractions, speed ratings, rating from a new pace system and finally my own composite ratings. So far in 48 races tested the results are promising, the ROI was a positive .13 nothing to brag about, only time will tell. More data is needed, as much as 500 races.

Thanks,

G.

I believe that you could improve the software by adding a weighting for each rating that is entered. You might trust one rating category to be more reliable than another and, therefore, rate it higher. That way, in the example you gave, instead of removing SHorPro completely you could leave it in but with a lower rating.

The challenge is then to have a reliable weighting.

Thank you for the feedback, weighting would indeed add another level of accuracy to the software. You can do this to a degree by making the variance on the more important ratings smaller and/or by converting your ratings into impact values so that everything is on the same scale and weighting your ratings before using them in the software.

Would entering the Racing posts T.S. be any advantage

In my experience there is very little value in using the TS as it’s used by so many people that the market has already absorbed it into the odds.

Hi Michael,

thanks for your efforts and detailed explanation. Could you confirm for me if this will work with Australian racing?

Kel

Hi Kel, thank you for your message. The concept will work with Australian racing but unfortunately the software won’t as we don’t have an Australian racing data feed at the moment.

Thanks Michael,

Please keep me on the mailing list for when the software is able to meld with an Australia data feed. For now though, I am I able to import a CSV of Australian data I have on hand to run simulations?

Kel

You can’t import an entire CSV but you can enter the selections manually for each race.

There is no information being downloaded to my copy of Puntology or Monte Carlo Manifest.Are you having server problems ?

We have had a few server issues today. Please restart the software now and it should be working.

Hi there Michael.

Could you tell me please how I can access the ratings dossier?

Much appreciated.

Kind regards, Steve

Hi Steve, you can get access to the Racing Dossier at http://www.racingdossier.co.uk/

Can you explain how you came to the default variance figures. The problem I have with things like this is the variance figures appear to be subjective. Is there a ‘correct’ mathematical way of identifying the variance? I.e using standard deviation

We used the average standard deviation to determine the default, but as you say it is quite generic and is much better to work it out on an individual basis. There are a few ways of doing this mathematically, such as standard deviation, but it can also be done subjectively which can be just as effective if you really understand how a rating works.

I have been trialling the Montecarlo software and adding my own ratings and the speed ratings from the racing post. It is a bit time consuming and error prone so is there anyway of cutting and pasting the stats from the RP into the spreadsheet? I have tried but you have to cut and paste the whole race and then it does not match with the columns.

Unfortunately there is no easy way of copying and pasting from the Racing Post website anymore.