TacticAI, the artificial intelligence that perfects corner kicks by predicting their outcome

AI also enters the world of football with a straight leg and does so with TacticAIa system developed by Google DeepMindfamily business Alphabet (the holding company which also controls Google) in collaboration with the Liverpool FC. The model based on generative and predictive artificial intelligence has already demonstrated a good degree of accuracy in predict the outcome of corner kicks indicating the player who will touch the ball and the probability that he will be able to shoot on goal: the experts of the English team have provided data relating to approximately 7200 corners of past Premier League seasons, which were used to train the model, which proved to be effective in 90% of cases. What distinguishes TacticAI from previous experiments is that for the first time the model is capable of provide advice to the coaches who will use them. For this reason, TacticAI could prove to be a useful tool for improving tactics and game plans, especially during corners.

How TacticAI’s algorithm works

Football fans will be curious to know how TacticAI’s algorithm works. First of all, it must be stated that, of all the types of set pieces available in the game of football, corners are probably in the best position to provide an adequate amount of data with which to train an artificial intelligence model like TacticAI and for several reasons.

First, corners are definitely more frequent compared to other types of set pieces, such as penalties; corners are kicked by one always fixed position and, moreover, they offer aopportunity to score on goal greater than goal kicks and lateral throws.

These characteristics make the corner kick a relatively simple situation to schematize and, therefore, predict. We say “relatively” because predicting the random movement of individual players involved in a corner is actually not that simple. This is why TacticAI uses a geometric approach in addressing the problem, converting the situation present on the pitch into one graph representation (the graph would be a set of nodes or points).

In the scheme designed by TacticAI each player is treated as a node in the graph and, thanks to these representations, its algorithm is able to make a whole series of calculations that allow it to understand which player is most likely to touch the ball and make a shot and how the opponents and teammates should position themselves to increase or decrease the probability of this happening. All invaluable information for coaches, whether their team is playing on offense or not.

TacticAI Chart |  Geopop

The potential of the AI ​​assistant for football

According to what was declared by Google DeepMind researchers, TacticAI already offers a good level of accuracy in predicting the outcome of a corner kick and in offering coaches truly useful suggestions in these particular phases of the game. In fact, on the page dedicated to TacticAI published on the Google blog, the researchers declared that the Liverpool football experts involved in the study they preferred TacticAI’s suggestions 90% of the timeindicating how effective it is already in reading game situations and predicting the outcome.

The use of similar solutions will allow coaches who use them save a significant amount of timeas they will be able to develop tactics and counter-tactics more quickly than in the pre-AI era given that they will no longer have to review a large amount of footage of previous matches with which to “study” their opponents.