An extra analysis weapon employed here over the jumps season is my custom National Hunt Neural Network. This page is not the time nor the place to go into high technical detail. In simple lay man's terms a Nueral network is simply a computer program that tries to put weight of importance on various factors in order to make a prediction about the future. Stage 1 typically involves an initial period of Training. A decent sample size of historic old data provides the initial teaching to the neural network. After that one feeds it daily fresh data and it uses it's learning from history to predict the future. The system should be continually learning as new data is fed in each racing day.

Weather men these days will use a similar core idea to predict the weather. Their systems will be trained on and eat a daily diet weather data such as wind speed, wind direction, pressure and temperatures. I am more interested in data related to horse racing.

My own network under went it's initial training in 2016. Since then it was a private thing for productive private use till the service here came about in 2019. I do not follow the computer blindly. More so I see it as one additional extra tool to be used in conjuction with other information and analysis. Client messages from me are not full of computer output techno babble. The Neural Network is more of a private back end thing that will have a degree of influence on which horses I note to clients or those I avoid.

Here are some explanations of the factors that are utilised in the Neural Network that powers the bets advised in the messages 

In total we have 8 factors that we assign a numerical measurement to, and, that the network “learns” from.



This is provided by the most accurate collateral form ratings that are then adjusted for today’s race conditions (primarily ground and distance suitability). It is a numerical rating that is considered the most likely number that the horse will run today.



This is a numerical assessment of the horse’s best form. It is calculated from official handicap marks, class of races contested and the likelihood of finishing position in those races suggested by in-running prices. We are utilising the knowledge of the marketplace.



We are looking for any Positive Pace Bias or Negative Pace Bias given today’s opponents previous running styles. It is essentially pace prediction.



The last run is the most “predictive” run and we look for factors that identify the run as better than the literal interpretation of the form may suggest.



These are trainer stats that have a strong chance of being relevant (most stats are not.) These are generally race specific or type of runner specific.




These are jockey stats that have a strong chance of being relevant (most are not) These are generally associated with running style or are course specific.



We try to identify runners over specific course and distances that have a high percentage chance of front-running, which confers a large advantage.



This can come from a variety of sources, such as, stable information, trainer information, professional judgement/information. It is entered as a numerical factor into the neural network.


The network has been “trained” on these factors and numerically evaluates/weights them based upon this. We can then compare the output with the current odds available and identify potential value bets where the NN suggests a greater chance of winning than the odds suggest. There are caveats regarding the size of margin required at different odds (the greater the price, the greater the margin demanded to qualify for a bet) but in simplistic terms, that is how the bets are identified.


The Neural Network has proven very successful in the years that myself and a betting partner have used it. We will continue to utilise the output ourselves.