Predictive analytics for football play outcomes

30 April 2012 » Football, IBM, Writing

Inspired by the NFL playoffs in January, I wrote an invention disclosure on how to merge player statistics with current field conditions to yield a visual probability of how a play might turn out.

The end result I envisioned was that you could watch the game on television as a spectator (or in real time as a competing coach) to see what the outcome the system predicted at the start of a field goal or 3rd down pass, for example.

As a sports fan the technology adds value to what broadcasters currently provide with digital first down markers and 3D play analysis.

As a coach, you could confidently plan what your next play would be. If the likelihood of scoring a touchdown was high, you could more quickly decide whether to kick the extra point or go for a two-point conversion ahead of time.

Beyond American football, the technology could be applied to many other situations, such as ice hockey, as well as non athletic events.

  • Example Embodiment #1: Field Goal
    In the NFL, a place kicker lines up to attempt a field goal. Using statistical data about the player (his history of successful field goals from this distance in this stadium) as well as sensor or other real time data about conditions on the field, the system overlays a heat map onto the image on the television screen, showing solid orange where the kick is likely to go (and fading opacity farther away from the center based on the probability). This provides a good indication where the kick will end up, and whether it will be successful. (Figure 1)

    Figure 1: Likely outcome of field goal attempt in this context
    Example Embodiment #1: Field Goal

  • Example Embodiment #2: Pass Play
    In the NFL, a quarterback has his team lined up in a pass formation. Using statistical data about the player (his history of passes from this field position in this stadium) as well as sensor or other real time data about conditions on the field, the system overlays a heat map onto the image on the television screen, showing orange where the ball is likely to go (and fading opacity farther away from the center based on the probability) and yellow highlights the probable receiver. This provides a good indication where the pass will end up, or to which player he will pass. (Figure 2)

    Figure 2: Likely receiver of the pass and field position of the catch in this context
    Example Embodiment #2: Pass Play

  • Example Embodiment #3: Hockey shootout
    In the National Hockey League (NHL), a hockey game has goes into overtime and comes to a shootout to determine the winner. Using statistical data about the player (his history of one on one shots against this goalie in this venue) as well as sensor or other real time data about conditions on the ice, the system overlays a heat map onto the image on the television screen, showing orange where the puck is likely to go (and fading opacity farther away from the center based on the probability). This provides a good indication where the player will shoot.

IBM decided not to pursue a patent, but published the idea to protect the intellectual property. The full article is available behind a paywall at IP.com.

Would be interesting if this serves as prior art for any later invention that gets implemented.