Sport Analytics Platform

Sports Info Solutions turned to Devox with the following requests:

1. Help develop a range of new products that would pave the way for a company into betting;
2. Modernize and enhance existing products and services with new features and upgrades that have not been possible in the past due to their architecture;
3. Centralize all products to one easy-to-access location, making it available for an average consumer to find, view and analyze relevant sports statistics.

Sports Leagues Draft

The Drafting event for any given sports league is a vital annual event in the realm of sports: this is when sports teams and sports college alumni get together to draw freshman players into the teams. Respectively, the Draft application provides all the necessary tools to conduct such a draft based on data-driven information.

Its main features include:

• Big Board of players with relevant analytics (overall score, ranking, body parameters, information and general comment);
• NFL and NBA Teams board divided by teams;
• Scouts board that provides scout’s bio, school and grade;
• Leaderboards by players’ positions;
• Games statistics (mostly for betting purposes);
• Glossary that explains every metric.

Platform for Comprehensive Sports Data Management

We’ve developed a platform that gathers, transforms, and stores sports data, whose primary goal is to serve the relevant sports data and organize it in a usable format. The platform serves the historical data via sport-agnostic REST API​, generates markets and betting lines​, and delivers upcoming and live data flows to customers and internal software in real-time​.

Architecture

We used R programming language, which is employed for machine learning projects. Within the solution, it is used for betting lines generation, as well as helps leverage the laws of math for stakes generation. Our team also engaged .NET, which was used to design data collectors. 

Data flow process

Data collectors (downloaders), written on .NET, download data from public API registers (MMA, football, basketball or other leagues) and transfer data to our AWS pipeline. At this stage, the system additionally engages Lambda data processing services – it allows the developers to program and push the function that would process the given data. It then saves the data to the data storage (RDS instance SQL server), transferring it further into a Lambda pipeline responsible for market generation. It’s the R-based Lambda, covered into docker and pushed into AWS. This data is saved to the databank and transferred to the Diffusion (3-party software), which enables the user to acquire this data in real-time. Alternatively, it is possible to get data from the database with REST API’s help. It’s also possible to schedule data processing to obtain relevant data as for the following moment.

Data collection process

The data-collecting modules (Intake Services) are responsible for:​

• Being aware of current and relevant data in real-time and updates acquiring;
• Pulling in data from third-party API (MMA Registry’s API, Baseball Registry’s API, etc.)​;
• Normalizing data​;
• Publishing updates to appropriate SNS topics.

Data Collection and Processing

• Intake Services: These services were responsible for maintaining current data awareness, pulling in data from third-party APIs, normalizing it, and publishing updates.
• Data Processors: Processor Lambdas read updates from SQS queues, write updates to the database, and publish these to the Diffusion platform.
• Market Generation: Containerized R functions read data, generate markets using machine learning or mathematical rules, and write these to the database. They also publish generated markets to Diffusion​.

Diffusion

Diffusion is a platform for real-time data delivery working as a data gateway, meaning it’s the versatile point of access to real-time data. With Diffusion, there’s no need to program individual REST APIs as it provides the ready-made SDKs for every popular programming language. The platform also allows to make minimal data transformations in no-code mode, which allows to easily and quickly organize data by topic or other criteria. It includes:

• Data Gateway. Makes it easy to integrate with any data source using WebSocket-based SDKs in all popular languages​
• Data Transformation. Provides a flexible, hierarchical data model to organize in-coming event-data in a multi-level topic tree structure​
• Data Distribution. Includes a real-time Event Broker that can scale to millions of concurrent connections​

Data serving

Data serving handles the real-time delivery of Competition, Team, Player, Statistic, and Market JSON objects​ a user can obtain right away. Diffusion is used both internally and externally for upcoming and live data. It solves the issue of being able to seed a client with initial data while also providing real-time updates of that data​

We created the REST API Service for customers to call endpoints to pull in data on their own schedule, getting data directly from the DB.