New A.I. technologies might potentially improve the game strategy by using the power of big data.

Published: 2023-03-11

This week, the Philadelphia Eagles and Kansas City Chiefs players and coaches will put in countless hours in film rooms to get ready for the Super Bowl. In an effort to identify opponent trends they may exploit and to shore up any vulnerabilities, they will analyze positions, plays, and formations as said in an article from Brigham Young University, published on ScienceDaily.com, 2023.

Although reducing the time and expense associated with film study for Super Bowl-bound teams (and all Professional and college football teams), new artificial intelligence technologies being developed by engineers at Brigham Young University might potentially improve the game strategy by using the power of big data.

Researchers at BYU, including professor D.J. Lee, master's student Jacob Newman, and Ph.D. candidates Andrew Sumsion and Shad Torrie, are automating the laborious task of manually evaluating and annotating game video. The researchers have developed an algorithm that can reliably identify individuals from game footage, name them, and establish the composition of the attacking squad. This operation typically requires the assistance of several video assistants.

While Lee, a professor of electrical and computer engineering were discussing this with his colleagues, they realized that, whoa, we could definitely teach an algorithm to accomplish this. They scheduled a meeting with BYU Football to learn about their procedure, and saw right away that could complete this task much more quickly.

Even though the study is still in its early stages, the team has already achieved greater than 90% accuracy with its algorithm for player recognition and tagging, along with 85% accuracy for identifying formations. They think that in the future, the technology will be able to do away with the time-consuming and ineffective practice of manual annotation and analysis of recorded video utilized by NFL and collegiate teams.

Lee and Newman started by watching actual game film that BYU's football team had given. They began to evaluate it and quickly discovered they required some extra perspectives in order to train their algorithm correctly. They then manually annotated 1,000 photos and videos from Madden 2020, a game that depicts the field from above and behind the offensive.

Using the usage of those photographs, they trained a deep-learning algorithm to identify the players and then fed that information into a Residual Network framework to calculate the players' positions. In order to detect which formation (out of more than 25 formations) the offense is utilizing, their neural network employs location and position information. This may be anything from the Pistol Bunch TE to the I Form H Slot Open.

When the player position and labeling information is precise, according to Lee, the computer can correctly recognize formations 99.5% of the time. One of the hardest formations to recognize was the I Formation, which has the center, the quarterback, the fullback, and the running back lined up one in front of the other.

The AI technology, according to Lee and Newman, may potentially be used in other sports. For instance, in baseball, it may map out player locations on the field and pinpoint recurring trends to help teams improve their defense against particular batters. Instead, it might be used to find soccer players in order to develop more successful formations.

There will be a lot more you can do with this data once you get it; you can advance it, Lee added. "Big data can assist us in understanding this team's tactics or that coach's habits. You might find it useful to know if they'll punt or go for it on 4th Down and 2. It's a tremendously interesting concept to use AI in sports, and it will be worthwhile if we can give them a 1% edge."