How to maximise your esports marketing campaigns with machine learning

Once a niche hobby reserved for basement-dwelling teenagers, eSports is now a global revenue heavyweight. The global eSports market is currently valued at over 1.38 billion U.S. dollars and is set to reach an almost-2 bi...

How to maximise your esports marketing campaigns with machine learning

Once a niche hobby reserved for basement-dwelling teenagers, eSports is now a global revenue heavyweight. The global eSports market is currently valued at over 1.38 billion U.S. dollars and is set to reach an almost-2 billion-dollar market value by 2025. As a result, eSports marketing campaigns are becoming increasingly complex. Marketing managers and eSports organisers are tasked with creating hyper-personalised campaigns at rapid speed. To meet this demand, advertisers are innovating their marketing efforts with AI. 

Machine learning is transformational for eSports marketing. It enables real-time understanding of player needs and preferences, faster and more accurate player matching, and dynamic in-game creative optimisation – all at a fraction of the cost of traditional marketing methods.

Despite these advantages, many marketers remain intimidated by machine learning. In this article, we’ll demystify machine learning for eSports marketing and demonstrate just how valuable it is. 

Why is fan engagement crucial to esports?

eSports did not evolve from nature or commercial endeavours. Rather, it is the product of evolving millennial interests and behaviours. In other words, eSports is a community creation. Without dedicated fans, eSports would not exist, which is why their engagement is so crucial. Outside of live eSports competitions, live streams and virtual tournaments maximise reach and open up sponsorship opportunities. As eSports continues to grow both online and in the metaverse, marketers need to keep fans consistently engaged. 


Machine learning can empower streamers with engagement tools to overcome competition. One way AI can engage fans is through automated content distribution. We live in an on-demand age, where the needs and desires of audiences are constantly evolving. Machine learning attribution models can adapt to new data in real-time. This makes it ideal for the dynamic nature of online advertising. ML can also help to identify unique selling points of your brand and craft content that will capture the attention of your target audience. Once you’ve directed audiences to your stream or eSports competition, the onus is on your broadcast to keep fans engaged. Putting on the perfect stream takes years of practice and trial and error. Fortunately, AI can sift through years of data to determine the viewer’s needs and patterns. ML could hold the key to the perfect fan experience. 

The challenges to implementing computer vision into esports competitions livestream

While AI for esports is an almost-perfect marriage for fan engagement, it does not come without challenges. Implementing computer vision into esports competition live streams is not infallible. We’ve had first-hand experience facing challenges to integrating our technology into live streams:


  1. Latency: Speed plays a crucial role in seamless gameplay. Whether it be on Twitch, Youtube, or Discord, viewers and gamers alike have all suffered the dreaded ‘lag’. Unfortunately, latency (the delay between your camera capturing an event and the event being displayed to viewers) is a common annoyance for streamers. Latency means that the algorithm is not able to track streams in actual real-time.
  2. Concurrency: Since esports is an always-on activity, multiple concurrent live feeds might need to be tracked. Computer vision may struggle to sufficiently track every key event across several games that are being played simultaneously.
  3. Timing– The length of a game is unpredictable. CV solutions may not be able to handle live feeds, lasting less than the duration specified in input. 


Our machine learning solutions for esports competitions live streams

At Sparkd, we understand the power of AI in esports. Optimisation via AI can be the key to new eSports sponsors, and bigger and better eSports competitions. Our machine learning API is underpinned by an understanding of the very core of eSports – how it’s marketed and how to get fans engaged. By integrating this knowledge into our technology, we developed a machine learning pipeline capable of extracting information from games streamed on Twitch automatically and in real-time.

To develop the API, we created a dataset of games and tested them against multiple different scenarios, with multiple streamers. We then annotated the videos by logging the time of key actions like goals, and the frames where the actions took place. We were thorough in making sure that the timestamps of the games were correctly identified to build a highly accurate solution.

By using template-matching and OCR (Optical Character Recognition), we finalised an algorithm that can automatically extract detailed live scoreboards and statistics. This data can then be repurposed for game efficiency, marketing efforts, and all-around better production. Content creation can be not only automated but also optimised. Highlight reels can be developed automatically, without the need for any manual editing. Our trained algorithm can detect key events so that you don’t have to go searching for the ‘money shot.’ This way, you can source quality images and videos that are literally programmed for high-engagement levels. 

Final Thoughts

As the eSports world continues to grow, standing out is going to become more and more difficult. Once you land fans, it’s equally important to keep them engaged through eSports marketing campaigns. Marketing in the metaverse cannot be directionless. It needs to be jam-packed with value for viewers. That’s exactly what our API offers. 

Work with Gemmo today for your AI needs.

Check out our other articles: 

Start your AI Journey Today

Author: Johanna Walsh 


Related Articles

Gemmo's noise classification case study with Sonitus