Maple Leaf Sports & Entertainment gains an edge with an assist from AWS

This blog is co-authored by William Reyes, Data Engineer, MLSE, Sport Performance Lab.

How MLSE developed a cloud-based rapid prototyping solution to reduce time-to-insight

In the dynamic world of professional sports, analytics and data-driven decision-making offer competitive advantage to sports teams and leagues. New avenues for competitive advantage have emerged as the varieties of performance data available has expanded. Sports organizations no longer simply reach for better performance modeling to help with strategic decision-making—they also seek to hyper-optimize the speed at which data insights can be delivered and applied. As a result, teams and leagues around the world increasingly look to the cloud to scale data science and analytics solutions.

The Toronto Raptors, for example, moved their entire model training ensemble from on-premises hardware to Amazon Web Services (AWS). This allowed the team to address practical data science issues, such as model drift, more quickly. It also ensures player projections reflect up-to-date performance while maintaining alignment with strategic or stylistic shifts across the National Basketball Association (NBA).

The Toronto Raptors are one of many sporting franchises in the Maple Leaf Sports & Entertainment (MLSE) portfolio. AWS provides cloud services across the entire MLSE organization, including the Toronto Maple Leafs, Toronto FC, Toronto Argonauts, and other local franchises. While these clubs operate in different sports and league structures with unique rules and regulations, the various front offices all seek answers to similar questions. Therefore, common use cases emerge from sport science, opposition analysis, player recruitment, and more.

Traditionally, individual teams work in silos to build tools to address these common use cases. To put reliable performance applications into production, each organization requires deep sport-specific knowledge and an understanding of best practices in the modern software development zeitgeist. MLSE sees opportunities to solve for problems once and apply solutions across various teams and centralized services.

MLSE Digital Labs weaves together business and digital strategy to support innovative solutions across its ecosystem. Digital Labs seeks to use both innovation and collaboration to disrupt the status quo in sports. The Sport Performance Lab (SPL) is a Digital Labs program that drives research and development with the goal of optimizing team and player performance for the MLSE family of teams.

After building several solutions for MLSE Team Operations, Digital Labs recognized an opportunity to construct a shared infrastructure and development process to allow for rapid prototyping of new performance tools. Shared infrastructure and process allow staff to focus on analysis of their specific sport rather than building generic software that plug into the wider MLSE technology stack. From this realization and necessity, Protostar was born. Protostar is MLSE’s rapid prototyping framework, empowering teams to quickly deploy applications and reduce time-to-insight.

New team applications

The NBA is the first US-based professional sports organization to employ full skeletal tracking technology across the league with Hawk-Eye. For the 2023-2024 basketball season, the NBA installed a comprehensive camera array in each team arena to track the movements of players at 60 frames per second. This provided an opportunity to create new tools to manage, curate, and communicate data insights.

SPL’s biomechanics team built one of the first tools on Protostar. When conducting sports science research, it is critical to confirm the outputs of your models against what’s happening on the court. To accomplish this, an internal application aligns match footage with various incidents of biomechanical interest detected and extracted from raw data. The underlying Protostar system allowed for the development, production, and deployment of a tool to perform this function in days rather than weeks, despite dependences on various underlying data sources of different shapes and sizes.

Toronto FC (TFC) used another tool with the same underlying framework during the 2023 season. On a Major League Soccer (MLS) matchday, starting lineup sheets are exchanged between competing teams about an hour before kickoff. Only a few minutes are available to tweak team tactics before departing the locker room for on-field warmups. SPL built an application that allows TFC performance staff to quickly query historical lineups and identify previous games where their opponent deployed a similar lineup. This greatly reduces the amount of manual searching previously required in an environment where every second counts. While this use case was entirely different from the basketball example, the underlying infrastructure that stood up this application was identical.

Protostar dramatically reduces time-to-insight for various MLSE teams and departments by facilitating system integration, generating deployment templates, and accelerating the iteration process for all digital products in the wider MLSE system. And it’s helpful that it’s effectively a self-service solution.

While this rapid prototyping platform services multiple professional sports teams, Digital Labs is exploring opportunities to apply it outside of the immediate Team Operations theater. There are opportunities to stand up tools quickly for game day operations around verticals like ticketing, concessions, security, and other areas of interest.

Designed by the MLSE Infrastructure team, the core of Protostar is a robust platform built using the Amazon Elastic Kubernetes Service (Amazon EKS). Amazon EKS is in wide use by with many existing MLSE developer groups. This allows MLSE to leverage existing skills and expertise while incorporating foundational elements from previous projects. These span areas including data governance, engineering guardrails, and reusable deployment strategies. As a result, the team can focus more on unique aspects of the design for the Sport Performance Lab, such as data orchestration and speeding up deployments and updates.

MLSE identified three key benefits to using Amazon EKS as the core of its platform:

• The MLSE team can leverage the extensive Kubernetes ecosystem of operators and open-source technologies to automate and orchestrate many components of the environment, such as DNS, networking, certificate provisioning, secret management, on-demand compute, and application deployments.
• Amazon EKS uses Kubernetes namespaces as a mechanism to secure multiple workloads on the platform. Each workload or project has its own namespace and namespace service account to identity and interact securely with AWS services like Amazon Relational Database Service (Amazon RDS), AWS Secrets Manager, Amazon Elastic Container Registry (Amazon ECR), and Amazon DynamoDB. This also allows isolation of workloads, which is essential for protecting sensitive information.
• The service builds on concepts and patterns already developed, which reduces time to delivery, increases operational efficiency, and frees up more time to focus on high-value innovation and problem solving.

Applications built on this framework are granted specific advantages that are useful for SPL. In particular, the speed at which Protostar deploys changes to existing applications through its continuous integration pipeline allows for in-game adjustments to various tools. This is important in supporting applications that straddle the line between prototype and production application, which is common in the sports analytics theater, where last-minute changes are often requested by team staff.

Historically, this procedure operated without automation, dependent on manual processes where each application was customized and largely non-reusable. The absence of templates and containerization contributed to a lack of interoperability. Moreover, deployment timelines varied, ranging from one to two weeks. The following diagram illustrates the Protostar deployment process. It makes use of GitHub Actions to create a workflow to build, containerize, and push container images to Amazon ECR for storage. Finally, the images stored in Amazon ECR are deployed to Amazon EKS.

Protostar provides an internal closed-loop process through which MLSE can build, deploy, and iterate on tools effectively and efficiently. It can serve all SPL projects, but can also extend across MLSE teams and business units.

Today, when data is more accessible to people than ever, generating data-driven insights and getting them into the hands of team and business decision makers as quickly as possible is a high priority. AWS enables collaboration across MLSE Digital Labs teams to not just develop the infrastructure and deployment processes needed for one solution, but to reproduce and scale a constellation of solutions that responds to the needs of team and business operations across the MLSE ecosystem.