AI Made Friendly HERE

Sevilla FC partners with IBM to create gen AI tool for scouting

Tech

The new product, called Scout Advisor, is built on IBM’s Watsonx and uses natural language processing to recognize jargon scouts use to describe players.

Sevilla FC, the perennial LaLiga contender known for its analytical prowess, collaborated with IBM to apply generative AI as a way to streamline the player recruitment process and utilize comprehensive information in the player recruitment process.

This new product, called Scout Advisor, uses natural language processing and semantic analysis to pore through textual scouting reports to identify player attributes, even when described in different ways. This enables the football operations department to enter a few search words and receive subjective and objective information that meet the criteria. It’s a way to leverage technology to better emphasize human evaluators.

“Now, we can fully use the expert opinion of the scouts in a more systematic process of finding good players,” said Elías Zamora Sillero, Sevilla’s Chief Data Officer, “and this also opens the possibility of using this NLP and generative AI technology in order to solve another problem with natural language in other areas of the club.”

Sevilla, the 2023 Europa League champion and a Champions League qualifier this season, has previously developed a series of AI-powered tools to aid its operations. AI Football helps with match analysis, AI Ticketing delivers personalized marketing, AI Sponsoring aids the commercial team and AI Radar was the original scoring tool.

More recently, AI Tracking more systematically details player movement so that clubs can be sure to receive transfer fees on homegrown academy alumni. That last tool was rebranded as Transfer Tracker and, in partnership with LaLiga Tech, is available for licensing by other clubs.

With Scout Advisor, Sevilla FC Sporting Director Victor Orta can enter a query such as “extremo con desparpajo,” which roughly translates to “winger with confidence,” and receive results that capture the essence of the request.

“I don’t need to review 45 reports for a player to know the opinion of my scouting department for the player. In perhaps two minutes, I can get all the information that is good for me for to make the decision,” Orta said. “This is a revolutionary tool for a director of football that gives time.”

Zamora’s team first began working with the University of Sevilla before contracting IBM to deliver generative AI product built on Watsonx. This type of tool is similar to what MLB’s Texas Rangers R&D department built last season to sift through scouting reports, even when written with dense sport-specific jargon, to capture playing characteristics that are hard to quantify even with advanced tracking systems.

“In many cases, similar themes are expressed by the scouts in a different way, so the style is not unified,” Zamora said. “And because of that, to make such a dictionary is very complex. So the task that we have to face as a data science department, is to develop NLP algorithms oriented toward the automatic extraction of key words from the scouting reports.

“It is as easy as describing with your own words — that is the key — which is the kind of player that you want to look for,” he added.

Sevilla’s prior embrace of AI used more traditional machine learning models that required intensive training and human supervision to create. Scout Advisor uses a pre-trained, foundational large language model (LLM) from Watsonx that could be automatically refined under computer-powered, self-supervised learning.

IBM Client Engineering Manager Arturo Guerrero noted that it had to be a “solution that fits the style of Sevilla,” recognizes the jargon scouts use and enable searches to find players similar to the style and ability of another player. All of this helps blend the subjective and quantitative realms of the scouting process.

“The results are based on semantic analysis, which is not just looking for a specific word,” he said. “It will understand the meaning of the prompt, and it will find similar reports to that. And then, obviously, we get all the stats and quantitative information of the player.”

Originally Appeared Here

You May Also Like

About the Author:

Early Bird