From Big Data to Artificial Intelligence. Descriptive Vs predictive

In football and its processes, uncertainty predominates in a very significant way given the intrinsic complexity of the game. Reducing this uncertainty is imperative for football clubs if they aim to optimize the performance of its most important assets and reduce risks in the decision-making process in each main departments.

Current sports analytics show gaps in the provided answers and higher precision is possible if we are able to obtain a deeper underlying knowledge. In football, this deficiencies can be solved through the correct use of new technologies, being artificial intelligence (AI) the discipline that best fits this scenario.

Currently, most institutions specialized in this sector analyze the data. AI provides resources that are beyond the reach of these types of analysis, given that it is capable of transforming the past data in new knowledge that can be applied in the predictive and prescriptive dimensions.

Olocip is committed to combining the most advanced modeling strategies, capable of transforming spatial-temporal data into transparent statistical models, with a thorough research on the differential factors of sport and understandable to human beings.

Olocip is an external AI department for sport clubs that combines experience and scientific knowledge to effectively address predictive and prescriptive analytics. This approach enables us to adapt our tools to the methodologies and idiosyncrasies specific to each club. This adaptation capacity and technology allows the development of solutions that allow us to analyze what is happening, make predictions and also indicate which parameters must be modified for what we really want to happen.

Through the use of predictive and prescriptive AI, we minimize uncertainty in decision making problems in different departments such as players’ performance predictions, real time match analysis, injury prevention, market value prediction, or fan engagment among others in order to achieve high-precision responses.

Such example, and focused on the performance of players, we understand that the real concern from the sports directors, lies in knowing their performance in a future context, and in a contextualized way, ie the player positioned in the new environment, new team, league or teammates…

At the beginning of the 2018/2019 season and during the presentation at the World football Summit in Madrid, it was decided to make and publish both the performance prediction of Karim Benzema in Real Madrid, analyzing how it would affect the departure of Cristiano Ronaldo, as well as the performance prediction of Cristiano Ronaldo, in his new Club, La Juventus, new league, teammates … that is to say in a contextualized way…

Regarding goals, Olocip redictive models established at the beginning of the season, that Ronaldo would achieve a goal every 118 minutes, finally he had scored one goal every 128 minutes for Juventus. Regarding shots, in Juventus was recorded 5.7 shots and Olocip’s predicted 4.92 in September 2018.

Regarding lose balls, the prediction was established at 1.3, registers aligned with the final statistics,1.21. Another example, assist, Olocip predictive models indicated that Cristiano would offer one goal pass every 333′. Finaly the Portuguese adds one every 336′

Regarding Karim Benzema in the new context, mainly without Cristiano. The model’s algorithm provided more significant detail, believing that Benzema would see his 2017/18 average goals-per-game ratio of 0.20 increase to 0.53 in 2018/19. at the end of the season was. 0.62.

If, at the beginning of the season, a decision had been made regarding Cristiano’s goals in Juventus assuming he was going to keep the figures of the previous season (descriptive analysis: 1.02) the mistake made against deciding from predictive values (AI analysis) would have amounted to more than 31% (1,02 vs 0,70).