How neural networks are helping bookmakers

Table of contents:

• What are neural networks?
• Why do bookmakers need it?
• Pros and cons of neural networks
• Bottom line

What are neural networks?

Neural networks are one of the areas of research in the field of artificial intelligence, based on attempts to reproduce the human nervous system. More precisely, it reproduces the nervous system's ability to learn and correct errors. This makes it possible to simulate, albeit not exactly accurately, how the human brain works.

Our brain is a sophisticated biological neural network, which receives information from the senses and processes it (face recognition, sensations, etc.). An artificial neural network, on the other hand, is a structure that receives a signal, transforms it (in much the same way as real neurons do), and passes it on to other neurons (which do the same).

But that is where the similarities between biological and artificial neural networks end. The structure of the human brain is much more complex, so it is not possible to reproduce it, even if less accurately.

A key property of neural networks is the ability to learn. "Learning" consists in changing the "strength" of synaptic connections between neurons.

How does this work? Recall the experiment of Academician Pavlov with a dog. He rang a bell when he ate, and after a while the dog learned to associate the sound with food intake. All because the synaptic connections between the areas of the brain responsible for hearing and the salivary glands became stronger.

As for artificial neural networks, they are a system of connected and interacting simple processors (artificial neurons). The learning of a network consists of finding the coefficients of connections between neurons. During learning a neural network is able to detect complex dependencies between inputs and outputs and also perform generalisation.

Why do bookmakers need it?

As told by an independent business consultant, neural networks and machine learning algorithms help BK to solve a variety of problems associated with the analysis of large amounts of data. The more data there is, the harder it is for a specialist, even with a lot of experience, to find dependencies. Usually, a strong analyst can find correlations between three to five statistical parameters, but more data requires third-party software.

"Obviously, neural networks and machine learning algorithms allow us to solve a wide range of tasks: they help us determine the odds, track unwanted players at an early stage, including forging, predict user behaviour (when they stop playing, how much money they can bring, what events they will bet on in the future)," he stressed.

According to the expert, implementation of such technologies allows reducing the amount of manual work and automating business processes as efficiently as possible. In addition to reducing costs, it also allows minimising the possibility of "human error", i.e. errors due to inattention.

"As far as I know, a bookmaker loses considerable money on unwanted players. The earlier you identify the next pro and restrict him, the more money you save. In the classic game of cat-and-mouse, in which some people try to figure out other people, there is a constant evolution in which undesirable players change their tactics. Algorithms usually adapt more quickly to new strategies, which allows for faster reactions and, again, savings. Of course, mandatory player verification makes life easier, as it is now much more difficult to register an account. "

According to the former chief operating officer of Betconstruct, an international betting and gambling software provider, at the moment neural networks are implemented in the betting business at a very rudimentary level.

"Big Data, Machine Learning, Artificial Intellect - these are very buzzwords that everyone is using in sales in the market, but in fact in most cases such statements make me personally smirk.

According to the specialist, the only really working cases are marketing cases: predicting the player's category in terms of marketing and its key parameters, and a recommendation system which, based on the actions of the player, recommends him to make a bid with certain criteria and parameters which suit him best.

Pros and cons of neural networks

The main advantages of neural networks are accuracy in predictions, elimination of human error from business processes, and cost optimisation.

"Neural networks help to see what the human eye cannot see. Almost everything that is common practice in betting has grown out of some successful case studies or the opinion of certain respected experts. This opinion has then developed into dogma, but it is not in fact dogma. In order to verify these statements and assumptions, you need to collect a huge amount of statistical data and carry out a lot of expensive analytical procedures, which for most operators is simply not economically feasible.

According to the expert, the human mind and brain cannot see very many of the dependencies and patterns that are revealed by processing huge amounts of data and applying machine learning.

As for the drawbacks, they are related to incorrect integration of the solution or a lack of understanding of how to use it.

"If you look at traditional e-commerce, you can see that companies are investing millions of dollars in automation with the help of neural networks and ML algorithms. It is solutions like this that are the future. Every year, the tasks that these technologies solve become more advanced. Whereas previously the computer brain was trusted with simple tasks like selecting recommended movies or books, today autopiloted cars are already showing that they are safer on the roads than ordinary drivers," emphasized the business consultant.

He also added that over time, the dependence of business success (its financial performance) on the adoption of new technologies will become increasingly apparent.

How are bookmakers using neural networks now?

Right now, neural networks are only being used by betting companies in marketing and by The legality of betting sites in India.

"There is no doubt that in the future bookmakers will use neural networks in compiling live and pre-match odds, in risk management and player categorisation, and in anti-fraud (behavioural factors), and in customer support," said the expert.

It is no longer news to anyone that the success of reactivating an inactive client is directly related to how much time has passed since he left. The success rate of reactivation can be divided into three stages: in the first, very short stage, the probability of successful reactivation decreases gradually. The second stage is when the probability drops sharply downwards, i.e. each day of delay significantly reduces the probability of success. And then there is the third stage, when the probability again decreases gradually, but it is at a negligible level.

It would seem simple - as soon as a player has stopped playing, you should immediately interact with them to reactivate. But this approach doesn't work. After all, every reactivation has a cost: it's either the manager's time spent or the cost of the bonus. If you reactivate every client after four days of inactivity, the business will not be profitable, moreover, it will turn out that not every client needed to be reactivated.

Every player has a different life cycle. One person bets only on Mondays and then rests all week. The other one bets every day. In this example, offering the first person a bonus because he hasn't bet for a few days is obviously a hasty conclusion. Chances are he would have placed another bet even without the bonus the following Monday. If you offer him a bonus, he would accept, but his business margin would decrease.

In the case of the second player, four days of inactivity is probably already a lot, as the player may have started betting with another bookmaker. The person bets every day, for him four days of inactivity is an eternity, so such a client should have been contacted after one day to keep the probability of a successful reactivation high.

"This example clearly illustrates the need for an individual approach to each client. Just imagine how much manual work it is for a manager. And as practice shows - people don't really like to work. Accordingly, routine work will be done, at best, only once, and the company will be losing money".

Bottom line

As of today, neural networks are not yet widely used at bookmakers. But companies are aware of them and are trying to implement them. Every year the extent of its implementation is increasing, and that means only one thing. In the near future it will not be easy, if at all, to survive without such a development in the betting business. And the players will have to "compete" not with humans, but with artificial intelligence in a bid to bankrupt the bookmaker.