TacticAI: an AI assistant for football tactics

As part of our multi -year cooperation with Liverpool FC, we develop a full AI system that can advise coaches in corner kicks

“Corner quickly … origin!”

Liverpool FC has achieved a historical return in the semi-finals of the 2019 Champions League. One of the most famous moments was a corner kick from Trent Alexander-Arnold, who lined up Divock Origi to record what happened in history as the greatest Liverpool FC.

The corner kicks have high potential of goals, but a routine -based position depends on a combination of human intuition and game design to determine patterns in competing teams and responding to flying.

Today, in Nature Communications, we offer Tacticai: AI (AI) system that can provide experts with tactical visions, especially in corner kicks, through an artificial and successive artificial intelligence. Despite the limited availability of standard gold data on kicks, Tacticai achieves modern results using the deep engineering learning approach to help create more complex models.

We have developed and evaluated Tacticai with experts from Liverpool Football Club as part of multi -year research cooperation. Tacticai’s suggestions were 90 % preferred by human experts on the tactical settings seen in the practice.

Tacticai explains the capabilities of AI technologies to help revolutionize sports for players, coaches and fans. Sport like football is also a dynamic field for developing artificial intelligence, as it features interactions in the real, multi -agent, with multimedia data. Artificial intelligence progress can be translated into many areas inside and outside the stadium – from computer and robots, into traffic coordination.

Develop a game plan with Liverpool FC

Five years ago, we started a multi -year cooperation with Liverpool to analyze artificial intelligence of mathematical analyzes.

The first paper, the game plan, looked at the reason for using artificial intelligence in helping football tactics, highlighting examples such as penalty kicks analysis. In 2022, we developed the IMPUTER chart, which made it clear how artificial intelligence could be used with a preliminary model of a prediction system for the firm tasks in football analyzes. The system can predict the movements of players outside the camera when tracking data-otherwise, the club will need to send a scout to watch the game personally.

Now, we have developed Tacticai as a complete AI system with predictive and obstetric models combined. Our system allows coaches to experience the player’s alternative settings for each interesting routine, then evaluate the possible results of these alternatives directly.

Tacticai is designed to address three basic questions:

1. As for a certain angle kick, the tactical preparation, what will happen? For example, he is likely to receive the ball, and will there be a snapshot?
2. Once the preparation is played, can we understand what happened? For example, have similar tactics done well in the past?
3. How can we set tactics to achieve a specific result? For example, how should defenders to reduce the possibility of photography attempts?

Predicting the results of the corner kick with deep engineering learning

A corner kick is given when the ball passes over the secondary line, after touching a defender player. The prediction of the results of the corner kicks is complicated, due to the randomness in the way of playing from individual players and dynamics between them. This also represents a challenge to the Supreme Organization for Standardization due to the limited golden parking data data-about 10 angle kicks are operated only in each Premier League match every season.

Tacticai successfully predicts the running of the kick through the application of deep engineering learning approach. First, we directly represent the implicit relations between the players by representing the corner kicks settings as graphics, where the contract represents the players (with features such as place, speed, height, etc.), and the edges represent the relations between them. Then, we take advantage of the approximate corresponding to the football field. Our engineering structure is a variable of the group’s tafafi network that generates all four possible reflections of a specific situation (original, impressed, manuscript V, HV stored) and forcing our predictions on receptions and photography attempts to be identical in all four. This approach reduces the search space for possible functions that our nervous network can represent that respect the consistency of reflection – and results in more generalized models, with lower training data.

Submit constructive suggestions for human experts

By harnessing its predictive and twin models, Tacticai can help trainers by finding similar corner kicks, and testing different tactics.

Traditionally, to develop tactics and anti -tactics, analysts re -watch many videos of games to search for similar examples and study competition teams. Tacticai automatically calculates the numerical representations of the players, which allows experts to search easily and efficiently for the relevant former routine. We have also verified the validity of this intuitive observation through extensive qualitative studies with football experts, who found that the best 1 recovery of Tacticai was 63 % related to time, almost, as the standard doubled by 33 % in methods that indicate pairs based on the player’s position analysis directly.

Tacticai’s obstetric model also allows human coaches to redesign corner kick tactics to improve the chances of some results, such as reducing the possibility of a shot to prepare defensive preparation. Tacticai provides tactical recommendations that modify the positions of all players in a specific team. Among these proposed amendments, coaches can determine important patterns, as well as the main players of the success or failure of the tactic, more quickly.

In our quantitative analysis, we showed that Tacticai was accurate in predicting kicks reception devices in the corner and filming cases, and this player who was repositioning was similar to how real plays were revealed. We also evaluated these qualitative recommendations in a blind case study, as Raters did not know any tactics of real play and which of them was created tactically. Liverpool football experts found that our suggestions cannot be distinguished from the real angles, and were preferred over their original positions by 90 % of the time. This indicates that Tacticai predictions are not only accurate, but are useful and viable.

Artificial intelligence progress for sports

Tacticai is a complete AI system that can give coaches immediate, accurate and accurate tactical visions – a process in this field. With Tacticai, we have developed AI assistant capable of football tactics and achieved a milestone in developing useful aids in AI in the sport. We hope that future research will help develop assistants who are expanding to more multimedia inputs outside the player’s data, and to help experts in other ways.

We explain how artificial intelligence can be used in football, but football can teach us a lot about artificial intelligence. It is a very dynamic and difficult game of analysis, with many human factories to psychology to psychology. It is difficult even for experts like experienced coaches to detect all styles. With Tacticai, we hope to take many lessons in developing the broader aid technologies that mix human experience and artificial intelligence analysis to help people in the real world.