How To Create Simple Punting Systems

Learn how to create a profitable betting strategy

Horse Racing Systems

If you’ve ever thumbed your way through a form guide on a Saturday afternoon, chances are you’ve been looking for horses that fit your favourite betting strategy.

Even though your approach may have been a rather simple one, in many ways you were using your own betting system.

Recently we touched on how syndicates develop complex betting strategies that analyse a huge range of factors. These factors are combined to form a set of ratings that are compared against the odds on offer prior to the jump.

While highly effective and profitable, these types of betting systems take years to develop, require specialist expertise and come at a considerable cost.

Creating simple betting strategies might not be quite as robust, but they still have the potential to make a decent profit.

It could be argued that using video analysis or analysing trials are effectively their own betting systems, but for this example we’ll examine largely data driven betting systems.

In essence a system is just a strategy that we are able to test.

Coming up with the factors to test is fun and interesting. However the most important part is setting up the right workflow to ensure that when you come up with factors and test them – you’re getting accurate results.

Getting The Right Data

To test out any system you first need to have access to the form and results data. In some countries like Hong Kong, most of that data is freely available and made public.

However in Australia, getting the right data to help get you an edge is more difficult and at times quite costly. To get the data to the point that it can be used to test you need to take it and put it into a database.

If you’re going to download or scrape the data from free websites, the most common and easiest database to use is MySQL. You’ll be able to easily access wins, places, starts and all the basic information that you’ll be able to get in a form guide. A more important and significant factor at this point in time in Australia are sectional times. Sectionals are expensive to generate as they involve a lot of man hours, by quite literally watching replays. As technology improves, so will access to this type of information.

There is also a lot of value to be had from video form and the notes on how a horse ran. You can use this information to add more depth to any model.

There are some websites in Australia that provide the ability to test different factors and how they might have hypothetically performed. There are a number of providers in Australia that give access to comprehensive form guide databases. Some of the better known providers are Southcoast, The Rating Bureau and Ratings 2 Win.

But it’s important to understand that there is no real shortcut to success in trying to build winning horse racing systems.

The Right Odds

But it’s important to understand that there is no real shortcut to success in trying to build winning horse racing systems.

However if you’re using the SP, or even the Betfair SP (including commissions), then you’re getting a far more accurate picture of how good your system actually is.

Sample Data

Now that you’re setup you can start testing different factors. However it’s important to understand exactly how to do this.

You need to break the data up into two different samples. The first is called the in-sample and this is the data that you’ll use to analyse and build your model. Then there’s the out-of-sample, which is the data that you’ll be testing your model on.

For example if you have 10 years of data, you might want to build your model based on the first 6 years – then test it on the final 4 years. More advanced databases can chop the information up in more scientific ways. For example, analyse/build on 70% of the available data (randomly) and using the remaining 30% for the out-of-sample test. Or split the data into non-consecutive years.

Overfitting

There’s a very good reason we do reserve the out-of-sample data.

That is because in trying to optimise our betting system, many punters quickly run into the issue of overfitting.

Overfitting occurs when we start testing too many factors and finding relationships that aren’t really there. They are more just random chance and are unlikely to continue occurring going forward.

For example we might build a betting system that is marginally profitable. However we find that we consistently lose on Wednesdays. In this case many people would consider that if we simply didn’t bet on Wednesdays we would have a powerful betting system.

However, it’s a clear mistake to think that by removing Wednesday’s results, you’ve built a solid system. In truth the system is flawed and the factors that you’re using aren’t significant enough to show a long term profit.

Fundamental Factors

Most important of all when building your system is to use factors that you feel are fundamentally important to determining the outcome of a race.

If you build a hypothesis and then simply test it on the data, then you are not going to fall into the trap of overfitting your data and coming up with factors that aren’t relevant.

Putting It All Together

Once you’ve created the right workflow it’s a matter of coming up with simple ideas that you feel might lead to a long term profit.

This might be as simple as how fast a horse ran its closing sectional after coming from a wide barrier – or if a horse got checked in the straight after starting as the favourite.

Often with simple betting strategies we are only looking for these factors that have sound fundamental reasoning and then testing them to determine if we can make a profit not just in our in-sample data, but far more importantly in our out-of-sample test.

Even building simple horse racing systems can be a lot of work, but no one ever said that punting was meant to be easy.

Some of our Tips and Ratings membership rely heavily on data, but most use very complex strategies. Take a closer look at each package for more info on how each analyst goes about finding value in the betting market.