Case Study: FONBET - Collect data with AI data scout on eSports disciplines

Fonbet and OSAI worked together on the development of an AI data scout for eSports disciplines, that collects custom data points.

Case Study: FONBET - Collect data with AI data scout on eSports disciplines

>4 K matches per month

0.5s data latency

Fonbet - is a well-recognized global betting provider, covering most of the sports disciplines and operating in Europe and the Eurasian region.  Fonbet and OSAI worked together on the development of an AI data scout for eSports disciplines, that collects custom data points.

The project had two main objectives:

  1. Collect accurate data faster and more cost-efficiently
  2. Open up new betting markets through real-time custom data points

The eSports AI data scout project was split into stages, starting with eNBA first and scaling to other disciplines.

OSAI uses AI/ML and Computer Vision to collect and analyze data from the live video feed. The technology was originally developed for table tennis, therefore covering eSports required some adjustments.

OSAI team spent about 3 months adjusting and testing the system. AI data scout for NBA was first piloted in August 2021 on just over 1,000 matches. In less than a month, it was ready for the production release. Today OSAI collects real-time reliable data on over 4,000 NBA games a month.

AI data scout at eSports NBA match

OSAI AI data scout simultaneously collects more than 20 data points every second of the game. This includes data on team possessions, points scored, free throws, fouls, and more.

Please find a summary table of the data points collected by the AI data scout below:

Game state 

Start, Finish the game. Start, Finish period. Start, Finish overtime. Start, Finish game clock. Timeout.

Team possessions

The logic of changing possessions is based on updating the timer of 24 seconds per attack, so the recognition of possessions has an accuracy of 95%

Point 

1, 2, 3 

Free throw

1 or 2 free throws. Can’t detect attempts outside of the three-point, for this case we send 2 free throws.

Foul

Foul resulting to free throw

Betstop / Betstart

Depending on different game situations, could be adjusted by customers request

Timer

Timer value and start/stop timer events

Team names


Score 


Offensive/defensive rebounds


Foul type

For every foul event

With the automated data collection, Fonbet was able to expand its eSports portfolio and optimize the cost structure.

AI data scout  - 52% faster

AI data scout by OSAI uses live video for data collection. Due to the nature of eSports, OSAI only needs a stream of 5 frames per second to collect data after each frame. On average, OSAI is 52% faster, delivering data 1.6 seconds before a human scout.

Traditionally, data collection speed can significantly vary due to human operation. Depending on the scout's expertise, the efficiency of work, time of play, and other factors, the speed of data collection can have a discrepancy of 1-10 seconds.

AI data scout by OSAI ensures consistency in data collection with the average data delivery speed of <1 second (justified by the network latency).

For some game events, AI data scout collects data 5+ seconds faster than a human.

Breakdown of the speed of capture of data points, AI data scout vs Human

AI data scout - 97.5% accuracy

Data accuracy is one of the key metrics used by Fonbet to evaluate OSAI performance. It’s important to note that human scouts do have a high accuracy rate, but it is not 100%. Humans are prone to errors. Normally there is a process in place to correct most of them on the fly, leading to a second delay in data delivery.

Fonbet used human data scouts as a benchmark for comparison, assuming 100% accuracy and speed. AI data scout by OSAI shows an average of 97.5+% accuracy, with errors automatically corrected within 1 second. AI data scout collects data according to the official rules. Any discrepancies in the data collected are caused by the difference between human and computer decisions. To maximize the system accuracy, the OSAI AI data scout recognizes on-screen information (like score) to correlate the data streams.

AI data scout by OSAI - accurate data in a more cost-effective way.

The success of the NBA AI data scout led OSAI to further develop its solution to cover other eSports disciplines according to Fonbet’s priorities.

In less than 3 months from the first pilot of NBA AI data scout, OSAI launched automatical data collection for eNHL. NHL AI data scout was tested for several months on over 2,000 games before going into a production phase in March 2022.

While the NHL data scout was being tested, OSAI also commenced a close pilot of eFIFA AI data scout. As the neural network was already trained on the hockey eSports training the new model for football eSports was much faster. AI data scout for FIFA is now ready for the production phase as well.

AI data scout at eSports FIFA match

OSAI and Fonbet continue working together on further developing the technology and introducing more custom data points and new betting markets. Both in physical sports and eSports.


Interested to learn more about AI data scout and computer vision technology by OSAI? Get in touch today!