Posted in:

How Computer Vision and AI Change the Data Game in Sports Betting

Sports betting is a good example of an industry solely driven by data. Considering an action-packed live environment, sports betting presents data challenges like no other industry. The standards are quite high. Data has to be accurate and real-time. And data capture needs to be scalable and stable. Most of all, data in sports betting tolerates no errors. Even the smallest of mistakes can lead to millions of dollars in losses. 

Data capture is currently dominated by human scouts, meaning data is limited by human factors. Speed of capture and data accuracy vary significantly depending on a scout. The major complexity of data capture is managing a team of data scouts, who can get tired, sick or distracted.

This creates a great opportunity for innovators to flex their development muscles. It makes a lot of sense to automate data capture in sports betting. One of the market leaders in the space is OSAI – AI-powered software that captures sports data from live streaming.

OSAI uses AI and Computer Vision to analyze low-latency live video and extract data after each frame. This creates a real-time feed of accurate data on all game events, including those outside of human scouts’ reach. AI data scout by OSAI is 52% faster and 70% cheaper than human scouts. 

The software not just solves the major challenge of scout management – AI is consistent regardless of the time of use, number of matches, etc. It also captures a much wider range of data points, including those outside of human eye reach. OSAI data feed includes all game events, player positions, zones occupied, ball speed, and more. 

Based on 4K+ matches a month, OSAI is 52% faster than human scouts with a consistent minimum data accuracy of 97.5%. 

OSAI’s AI data scout is available for table tennis and eSports (FIFA, NBA, NHL, Rocket League). And the company has several other sports in the pipeline. 

We decided to take a closer look at the evaluation of OSAI performance. Computer Vision is a tricky technology to master, especially in a fast-paced environment with small object (ping-pong ball) tracking. 

Accuracy findings: 

Data accuracy is absolutely paramount. OSAI constantly monitors the performance of AI data scout and continues to improve on the accuracy. The minimum accuracy of AI data scout is 97.5% across all games. To illustrate, please see below a sample of accuracy analysis from one of the halls of LigaPro: 

As you can see, the median accuracy is above 98.5%, with a minimum of 97.5% and a maximum of 100%. 

Data speed findings: 

With the development of in-play and super live betting formats, the speed of data collection becomes critical. The comparison of game event data input reveals that AI data scout performance is not just comparable, but better than the human scouts. 

We’ve selected two data points to demonstrate. AI data scout by OSAI is 24.1% faster in identifying the start of the game and 51.2% in keeping the score. 

OSAI slower than a scout Equal

(diff. <1s)

OSAI faster than a scout OSAI slower, % Equal, % OSAI faster, %
Event >5s 5s 4s 3s 2s 1s 1s 2s 3s 4s 5s >5s
POINT 1 0 1 0 9 28 412 210 205 45 5 1 7 4.2 44.6 51.2
BALL_IN_GAME 0 3 7 13 57 69 534 61 68 43 18 7 20 16.6 59.3 24.1

The sample of 430 LigaPro matches in April reveals the raw data latency of 14.7ms (8ms + network latency). 

AI data scout by OSAI delivers similar results in eSports data capture. Due to the nature of eSports, OSAI only needs a stream of 5 frames per second to collect data on all game events. On average, OSAI is 52% faster, delivers data 1.6 seconds before a human scout.

AI data scout set up

OSAI setup varies depending on the sports discipline. In a nutshell, OSAI needs a low-latency video feed to capture data. For eSports data capture, OSAI is deployed in the cloud and the minimum requirement is 5fps video. For table tennis, due to its fast-paced nature, OSAI needs 120 fps video. This often means the installation of a special camera that can film at this speed and a computer for local processing (to minimize data latency by performing all calculations locally). 

OSAI uses AI and Computer Vision to capture accurate data for sports betting in real time. The company also developed product offerings that apply this data to enhance the viewing experience, boost sports performance and provide new digital assets for sponsorship. For instance, OSAI also has a real-time Mixed Reality solution, that adds immersive and contextual animations to the sports live streaming. 

Relevant links: 

AI Data scout for esports, promo video:

AI data scout website: 

Fonbet case study (esports):

LigaPro case study (table tennis):