How Artificial Intelligence(AI) could improve the way $4.4 billion is spent on football?

$100 million. That is how much money FC Barcelona lost out because of one transfer decision. FC Barcelona bought Antoine Griezmann in the summer of 2019 for a transfer fee of $120 million. After a performance below expectations, Griezmann was sent to Atlético de Madrid for a two-year loan before being sold for $20 million.

The total money spent by all the clubs in Europe’s big five leagues (the Bundesliga, La Liga, Ligue 1, Serie A, and the Premier League) for the year 2022–23 was $4.4 billion. This article talks about how clubs make transfer decisions, and how exactly AI could improve this process.

The goal of any club is to win trophies and they want quality players who can help them achieve that, so they spend millions of dollars on players. However, questions like, who to buy, how much to spend, and when to buy are decisions made by the club’s leadership. Transfer decisions are really hard for any club to do perfectly as there are so many different factors in play. This is why I believe that AI could certainly augment this process and help clubs make the correct decisions.

Artificial Intelligence

The term Artificial Intelligence (AI) is being thrown around a lot these days and if you look at mentions of artificial intelligence online, you can see a clear exponential trend. Beyond all this hype, what is Artificial Intelligence? I mentioned that Artificial intelligence could improve how football transfers are made, how exactly can AI do that? What is the impact of AI going to be in this field?

What is Artificial Intelligence (AI)?

Artificial Intelligence (AI) is generally regarded as something that makes computers think like humans (and some are scarily close enough). In this context, I feel that Artificial intelligence (AI) is better defined as an extensive tool that enables people to rethink how we integrate information, analyze data, and use the resulting insights to improve decision-making.

Some really cool applications of AI are computer vision, speech recognition services (like Siri), web search engines (like Google), recommendation systems (like Netlfix), etc. One of the most recent innovations in AI is the semi-automated offside technology used in the FIFA World Cup 2022 in Qatar. The AI-based technology solution uses 12 dedicated tracking cameras to create three-dimensional models of a player’s skeleton to determine whether any part of their body is offside.

AI has started developing in the football industry and only a matter of time before its impact spreads to other areas of football.

Artificial Intelligence is a really broad term and contains various branches of knowledge that are actually used to implement them in real life. The branches of AI that I believe will be implemented in this solution are Machine Learning (ML), Deep Learning (DL), and Natural Langauge Processing (NLP).

There are quite a few different branches that might be involved in the future but I believe that these are the key branches to be used in the solution. I will explain these concepts briefly, then about the current methods of transfers then talk about that how these concepts are going to be used in the solution.

Part 1: Going deeper into AI — the concepts

Machine Learning

Machine learning (ML), a subfield of artificial intelligence, is broadly defined as the capability of a machine to imitate intelligent human behavior. For example, the classification of spam email is an application of training a machine learning algorithm to understand the differences and predict if an email is spam or not.

Machine Learning algorithms use training data (sample data from past experiences) to build a mathematical model that helps in making predictions without being explicitly programmed.

There are four main types of machine learning algorithms:

• Supervised learning: Models are trained using well “labeled” training data, and on basis of that data, machines predict the output. The labeled data means some input data is already tagged with the correct output.
• Unsupervised learning: Models are trained using an unlabeled dataset and are allowed to act on that data without any supervision.
• Semi-supervised learning: An algorithm that represents the intermediate ground between Supervised and Unsupervised learning algorithms. It uses a combination of labeled and unlabeled datasets during the training period.
• Reinforcement learning: Feedback-based techniques in which an agent learns to behave in an environment by performing the actions and seeing the results of actions. For each good action, the agent gets positive feedback, and for each bad action, the agent gets negative feedback or a penalty. This method learns from unlabeled data.

Deep Learning

Deep Learning is a subfield of machine learning concerned with algorithms inspired by the structure and function of the brain called artificial neural networks.

Deep learning and machine learning both are playing crucial roles in today’s world. ML models are good for small and medium-sized datasets. On the other hand, deep learning models require large datasets to show accurate results. Ultimately, the use of machine learning and deep learning totally depends on the use case. In cases where there is large amounts of data, deep learning can prove to be particularly useful.

How is Deep Learning implemented?

While machine learning is implemented through different models and algorithms (like sci-kit learn, TensorFlow, etc.), deep learning is implemented by using Neural Networks. Neural networks, also called artificial neural networks or neural nets, are computing systems inspired by biological neural networks. The neurons are connected by links and they interact with each other. The nodes can take input data and perform simple operations on the data.

Neural Networks use the architecture of human neurons which have multiple inputs, a processing unit, and single/multiple outputs. There are weights associated with each connection of neurons. By adjusting these weights, a neural network arrives at an equation that is used for predicting outputs on new unseen data. This process is done by backpropagation and updating the weights.

1. Input Layers: It’s the layer in which we give input to our model. The number of neurons in this layer is equal to the total number of features in our data (if it is an image, it is the number of pixels).
2. Hidden Layer: The input from the Input layer is then fed into the hidden layer. The number of hidden layers depends on the model and the size of the data. Each hidden layer can have different numbers of neurons which are generally greater than the number of features. The output from each layer is computed by matrix multiplication of the output of the previous layer with learnable weights of that layer. Then by the addition of learnable biases followed by activation function which makes the network nonlinear.
3. Output Layer: The output from the hidden layer is then fed into a logistic function like sigmoid or softmax which converts the output of each class into the probability score of each class.

Types of neural networks:

Recurrent Neural Networks (RNN)
A recurrent neural network recognizes data’s sequential characteristics and uses patterns to predict the next likely scenario. It is commonly used in speech recognition and natural language processing.

Convolutional Neural Networks (CNN)
The convolutional Neural Network (CNN) works by getting an image, designating it some weightage based on the different objects of the image, and then distinguishing them from each other. CNN requires very little pre-processing data as compared to other deep learning algorithms.

Natural Language Processing

Natural Language Processing (NLP) is the application of computational techniques to the analysis and synthesis of natural language and speech. In simpler words, it's how computers understand human speech. It's the fundamental principle behind how Siri and Google Assistant work.

NLP algorithms are based on machine learning algorithms. Instead of hand-coding large sets of rules, NLP can rely on machine learning to automatically learn these rules by analyzing a set of examples (i.e. a large corpus, like a book, down to a collection of sentences), and making a statistical inference.

Part 2: Understanding Football transfers

What are transfers?

Transfers are a central component in football, with the biggest clubs in the world exchanging mind-boggling sums of money to sign the players they want. A transfer in football is a business transaction between two clubs that sees a player move from one club to the other. The player signs a contract with the club ensuring his services for an agreed-upon price. So to get the player to play for a club, they have to pay a transfer fee as compensation to the player. The transfer fee takes into account several factors, including the perceived quality of a player, duration of the existing contract, commercial value, future potential, amount of future salary owed, and the willingness of clubs to name a few.

Although the process of a player joining a new club seems pretty straightforward, transfers involve a lot of discussions and behind-the-scenes movement, despite an official ruling stating that transfers can only be dealt with once clubs have official consent for the player to speak to other clubs.

The process behind transfers

The success of transfers by a manager hinges on his scouts, and the process is not taken lightly. While the time between an approach for a player and his signing can be days, everything that goes before can take months, even years. The process behind this is actually pretty huge, there are several steps before the money gets involved.

This is the transfer process from a scouts perspective(from Bleacherreport)

Step 1: Requirement
The first team managers and coaches come together to tell the scouts about their requirements for the players to be bought for the season. The scouts have to make sure that they identify players who will be able to fit into the manager's vision (players who don't match the club/manager's philosophy usually don't produce ideal results).

Step 2: Identification
The identification is the most important part. The scout now has to go to find a player to match the requirements set by the manager. The scout can identify players in numerous ways these days either by using video platforms like Wyscout or InStat or even from DVDs and website links from football agents themselves. However, only relying on these sources is not feasible as clubs are expecting the best to spend millions of dollars on.

Step 3: Planning and Travel
This is mostly an administrative job, where the club arranges for the scout to travel to the match to watch the player identified in person. The schedule has to be ideal and the scout would have to manage his time well.

Step 4: Viewing the player
So the scout goes to the player's matches and watches them play in person. This gives the scout an overall idea of the player including minor details like the mentality of the player, off-the-ball movement, attitude toward people, etc. They can also observe the player's attitude on and off the ball.

Step 5: Report and Discussion
The scout has to document his findings as soon as he can and he has to tell the club his opinion by providing them with a scout report. If the club feels like the player is the right fit, they will have further discussions with the scout to get more information about the player.

Step 6: Further Monitoring
After having discussions, the club may send the scout to view the player again or might send another scout to get another opinion about the player. They consistently monitor the play to ensure that he is performing consistently.

Step 7: Bid and Conclusion
After all these steps, we finally reach the stage where the club places an official transfer request for the player and they end up signing the player.

Part 3: AI in football transfers (assisting scouts)

The potential impact of AI in the stages of transfers

I think that the impact of AI can be really seen in Step 1, Step 2, and Step 5 and it could potentially be used in 3 as well.

Step 1: Identifying the requirement of a team
Managers and scouts look for a particular area of the team and identify weaknesses to improve upon. They would then look into the requirements of the players they would need to improve those weaknesses. AI could analyze tons of data from various sources and identify the ideal requirements for a player the team needed

Another point, AI could analyze the data of the team much more intensely and identify much more intricate weaknesses and more specific problems. AI would also be able to identify the exact measures required like understanding areas of friction, areas where possession is lost, areas of goals most conceded, etc.

This can be done using machine learning algorithms that look at all the data collected through various sources. At the same time, deep learning algorithms for image classification would help in collecting data such as player movement, shots taken, and advanced metrics like facial expressions, and gestures to understand the players' mentality

Step 2: Identifying the player to fit the requirements
Combining the intricate data identified by AI, and using the new & relevant metrics developed (or even developing new metrics relevant to the team) like pre-assists, chances created, work rate, etc. to identify players best suited for the team. Adding the knowledge of scouts to the AI, clubs can buy the ideal player for their squad and the chances of him performing well would be really high.

In the future, the AI would also factor in different factors like the psychology of the player, fan appreciation, loyalty, etc. to make predictions as some of the recent transfers have failed because of these reasons (Reference to this)

Step 5: Report and Discussion
Using Natural Language Processing(NLP) to understand the scouts' reports could also lead to faster processing of the data. This could be used to create some basic insights about the player that helps all the members of the teams leadership to easily understand the important details about the player.

Conclusion

The impact of AI has been monumental in several fields and I feel that its impact in this industry (where $4.4 billion was spent last year) is going to provide some fascinating results. There have been several transfers that haven't provided the desired outcome for the clubs and I believe that the integration of AI into the existing results will greatly help in achieving those results. A deeper analysis of these factors could have saved over $100 million dollars for FC Barcelona

Griezmann transfer news — https://www.insider.com/barcelona-griezmann-deadline-day-sale-caps-awful-summer-2021-9