Scribd
Upload
40 views20 pages

Mehrsan Javan - SPORTLOGiQ

The document discusses how AI can revolutionize sports team performance and fan engagement by providing reliable data and insights. It highlights the capabilities of an AI engine that generates actionable insights from game footage and player data, enhancing fan interaction and team strategies. The company, founded in 2015, aims to transform the sports industry by leveraging advanced computer vision and machine learning techniques.

Uploaded by

Yungdrung Dorje
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
40 views20 pages

Mehrsan Javan - SPORTLOGiQ

The document discusses how AI can revolutionize sports team performance and fan engagement by providing reliable data and insights. It highlights the capabilities of an AI engine that generates actionable insights from game footage and player data, enhancing fan interaction and team strategies. The company, founded in 2015, aims to transform the sports industry by leveraging advanced computer vision and machine learning techniques.

Uploaded by

Yungdrung Dorje
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 20

Revolutionizing Sports Team Performance &

Fan Engagement
with AI
THE PROBLEM

The sports industry relies on


incomplete, unreliable, and manually
collected data…

We believe AI can change that.


Industry Landscape

+$800B Industry
Stakeholders need relevant information at the right time
Deep Insights Real-time Insights
Who We Are

Founded in 2015
Two innovation centres in Canada

61 Teams

100% Coverage

Long Standing Partnership

Story Automation
The Operating System for Sport

● Unearthing never before seen insights into


Verticals
performance
● Enhancing the conversation amongst fans
● Supplying a data hungry fan base with new, Game Models & Insights

shareable social currency


Computer 158 million
● Giving teams new ways to find an edge
DATA POINTS PER GAME

Vision Data 576TYPES OF “ON-BALL” EVENTS

Collection
756,000 XY COORDINATES OF ALL PLAYERS

and helping them find ways to win

Game Footage
Generating Insights from X

Can we automatically AI Engine


generate robust, useful team
insights from partial MACHINE

observations? LEARNING

COMPUTER
VISION
What are the Insights?
AI Engine: from Video to Insights

Location Data
Broadcast
Footage Game Models
Body Skeleton

Actions/Activities
Step 1: Computer Vision

Self Camera Calibration & Map Building


Tracking
● Player/Object Detection
● Player/Team Identification
● Player Re-Identification
● State Estimation & Data Association
Step 1: Computer Vision

Body Pose Estimation in 2D & 3D


Activity Recognition
● Individual Actions
● Group Activities and Interactions
What Does Computer Vision Tell Us?
Can we automatically generate
robust, useful team insights from
partial observations?
Step 2: Game Models

●Player Performance Challenges:


Evaluation ●Partial Observations
●Story-Point Identification ●No Ground Truth Data
●Team Strategy/Tactic
Evaluation
●Recommendation
●Prediction
Multi-Agent Reinforcement Learning as a Solution

Inverse Multi-Agent Reinforcement Learning


●What is a team/player trying to optimize?

Behavioural Reinforcement Learning


●What is the actual playing style of teams/players?

MARL abstractions to measure similarity


The Same Play, Again!
Translating Insights into Sport Language

Passing Options
The number of unobstructed passing lanes to teams/players
Translating Insights into Sport Language

Line Breaking Passes and Options


A clear passing option (an unobstructed passing lane) that would break one of these defensive lines
Real-time Processing is Difficult

Computer Vision Game Models


● Take the output of the vision
● Processing 1080p @ 30fps for
○ Generate predictive analytics
○ Self camera calibration
in real-time
○ Detection, PlayerID, and Re- ○ Generate actionable insights
ID (combined together) in a timely manner

○ Tracking – Classical
optimization process
○ 2D/3D Body pose
○ Activity recognition
The real challenge is…

The sports community must be


comfortable with the stories we tell,
and the tools we use to tell them.
Thank you!

sportlogiq.com

You might also like