Click here to Skip to main content
15,867,308 members
Articles / Artificial Intelligence / Machine Learning

Soccer Field Strategy - Using SAP HANA and Amazon Sagemaker

Rate me:
Please Sign up or sign in to vote.
2.78/5 (5 votes)
2 Nov 2019CPOL4 min read 7.8K   2  
This article describes how AI can be utilized to make a better team strategy from Manager (or Team Coach) in a live soccer game by utilizing SAP HANA and Amazon Sagemaker capabilities together.

This article is part of the Cloud AI Challenge with SAP HANA and Amazon SageMaker. This entry is not meant to be a full article - it's purely just an outline of an idea - and will be removed once the contest has concluded.

Introduction

This article describes how AI can be utilized to make a better team strategy from Manager (or Team Coach) in a live soccer game by utilizing SAP HANA and Amazon Sagemaker capabilities together.

Background

Let's discuss the Use Case first and then check how SAP HANA and Amazon Sagemaker can fit here.

As we know, soccer is the most entertaining sport in the world involving billions of dollars in the market and lots of people watch it. For example, Paris Saint Germain (PSG) spent 222 million euros for Neymar transfer from Barcelona, Spain in 2017 - 18 transfer window. National and Club soccer Teams are investing more in the Technology area compared to the last decade.

There is a lot of data available either in a relational and not relational format in Soccer but not utilized effectively in Predictive or Prescriptive Analysis area compared to other sports. I agree some clubs are started ML for their Transfer or Marketing strategy. But here, I am explaining how we can effectively use the data to make instant strategic decisions when the team started to play or after the Half time of the match or crucial time like the last 10-20 minutes.

Let's explain a little bit in detail. In Live soccer matches, most of the Team Coaches are failing to substitute proper players from the bench even though they have good players. Or what would be the strategy for the last 15 minutes like setting up higher pressure or trying to keep more possession or trying to play with a long pass or through the pass or introduce another attacker or winger, etc.

Consider a single or multiple AI model that predicts the picture of what will happen in the next 20 minutes and how it would change the probability of winning more goals if we modify the model variable and present this instruction to the coach. Not bad, really!

So what is challenging here! Definitely we have trained a model before match start and the coach can predict a better strategy like Defensive or Offensive, better playing team, and better formation, etc. Though this prediction and Prescriptive Analysis are very helpful to the coach, can you imagine what is the impact of a good prediction from the coach while the team is playing.

We can expect data in a different format including relational and nonrelational structure while the team is playing. The relational structure includes - Players details like how many fouls committed, yellow/red card possession, corner, etc.

The non-relational structure mainly includes - video of the play so far, different images like pitch map and ball possession, etc. Consider a model based on inputs from both relational and nonrelational data and the ability to Retrain Model based on different parameters and predict the output at each interval of time while the match continues.

Approaches

What is the technical challenge here - first, we need a database, that can be used to store Relational and Non-Relational data like videos, images, etc., no compromise on data read/write access speed, ability to fetch data using SQL Query, Spatial Query, Live Data Streaming, etc.

Second, we need a platform that can be used to Build, Learn, Train Model and Predict the input.

SAP HANA is a better candidate for database because it is In-Memory DB and Data retrieval and update is super fast. Also, it supports both Relational and Non-Relational Data with inbuilt Analytical capability. In our use case, we are expecting high volume of different types of data that should process and analyze immediately.

Amazon Sagemaker provides a highly scalable machine learning environment and framework to Analyse, Build, Train and Deploy the model. We don't need to worry about data volume, infrastructure, logging, monitoring, deep learning framework. Amazon Sagemaker expects input data as either its own environments like EC2 or streaming from external sources. Here, we can utilize HANA high-speed streaming capability.

As I mentioned, Amazon Sagemaker deploys the model in its infrastructure, but with the integration of Tensorflow, we can deploy model containers in Amazon Fargate, and finally consume them from SAP HANA, express edition using SAP HANA External Machine Learning Library (EML).

Points of Interest

There are different tools and frameworks available in ML But SAP HANA and Amazon Sagemaker together perform better predictive and prescriptive analyses for time-critical scenarios having high volumes of data.

Please note that as of now, FIFA is not permitting to fit any Sensor with the player and if it allows it, then this would be an entirely different use case.

History

  • 2nd November, 2019: Initial version

License

This article, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)


Written By
Architect
United States United States
Architect, Full Stack Developer, and Exploring Cloud and AI.

Comments and Discussions

 
-- There are no messages in this forum --