Neural factor Analysis

Back to Neurals

Racing and Sports announces a revolutionary, alternative race form assessment methodology called NEURAL FACTOR ANALYSIS.

Artificial Neural networks have been around since the 1950's, but only more recently have they been applied to solve an ever increasing number of complex real world problems dealing with such areas as medical research, law enforcement, mechanical system diagnostics and the environment.

With this as a background, Racing and Sports set about applying the neural networking to race form analysis.

Neural networking, without going deeply into it, is a tool essentially used to analyse information in a structured way by collecting numerous mathematical models to emulate actual observances.

The advantage of Neural Networks is their ability to even out distortions in input data. They are also capable of "learning" the process to which they are applied.

Neural Networks are generally good for solving problems which tend to be too complex for traditional analysis tools such as clustering algorithms, multivariate statistics, fuzzy logic etc.

In many respects, Neural Networks are very well suited to problems that humans are able to solve but other more rigid, traditional methods are not.

That's why we decided to investigate them for an application to find solutions in race form analysis.

Racing and Sports has investigated the various types of Neural Networks which would best suit this purpose and again without getting too technical opted for a variation of the supervised algorithm.

In other words, the Racing and Sports modified neural network takes the input data and supervises it through the network system to produce a set of Neural Weightings as an output.

These Neural Weightings assigned to each horse are an indication of the best (higher) to worst (lower) based on the factor input values decided upon by the user.

The input information is processed through the Neural Network by a mathematically derived "engine" which drives the system. This engine is composed by a large number of interconnected processing elements which receive and process the input algorithms.

As you know, the analysis of race form is a very complex task. A single weight rating , a la Don Scott is not the only way to solve the pre race riddle.

Traditional weight ratings are but one of these factors and instead of adding a kilo here and taking one off there for other inputs like barrier, consistency etc, the Neural Networks offer an alternative approach to form analysis by combining all the factors mathematically.

In the development of the "neural engine" the Racing and Sports computers have run at least 2.5 million simulations based around the various input algorithms.

It is these simulations the "engine" draws upon to generate the output weightings of the network.

With so much racing, the time now required by conventional form techniques would drive one insane, if every day was devoted to "doing the form" old style.

This is where the power of computers, technology and mathematics come into play.

Racing and Sports has a big investment in this "state of the art" Neural Factor Analysis and is certain that all our development in fine tuning has now delivered a form tool like no other.

At this juncture, Racing and Sports has incorporated just 12 input form factors into the Neural Network.

In our trialling period, we have been astounded by the consistency these factors have delivered.

The Neural Weightings derived have shown an excellent record in identifying the main chances.

Of course there is no "magic bullet" in form assessment. At the end of the day we are dealing with 500kg animals with emotions and feelings like the rest of us.

They do not, and never will, always conform to a mathematical formula or a single numerical figure, no matter what.

However, as a guide as to identifying who the main group of horses are likely to be, in the outcome of a race based on the input data, we believe Neural Factor Analysis will give an excellent starting point.

It will provide a "new edge" in form analysis and offer a robust framework within which to assess each runner.

Early evidence suggests this is an alternative way of producing consistent results from your form assessment.

The neural factor analysis is NOT a system.

It takes the traditional class/weight correlation algorithm, adds a new race time algorithm and then integrates the other peripheral elements like jockey, trainer, course, distance, barrier to give a final list of weightings for the main chances.

This final listing depends of course on the users preference of the algorithms (Preference Scale ) and should serve as a very helpful guide to punters desiring to "fine" a field down to the main chances for closer examination.