The client is a betting prediction site. They examine the chances for the most popular sports, including horse racing, NFL, NRL, and AFL betting. They are one of the leading betting agencies that are trusted by hundreds. 

11 months of cooperation | Gaming/Entertainment

Tools Used:

SQL | Python Jupyter Notebook | Google collab | GCP Big Query | MS-Excel.

Challenges: 

  • Understanding the NFL’s specific realm of American football. This includes being familiar with the terminology used as well as the sport’s laws and regulations.
  • Knowing which API to use for the specific purpose because other APIs were providing nearly identical information with slight variations before calculating the teams’ ELO and Power Ratings.
  • Recognizing the variables that will most effectively affect the predictive models.
  • Before the games start, no API provides precise information on the live teams. Therefore, we also had to adjust the predictive models according.

Solution:

To comply with the client’s requirements our team started working efficiently on the ELO rating and Power Rating and understanding how it works. 

We extracted the data from the required APIs and created a function in python to convert the extracted data into a readable format. 

Following extraction, we cleaned and munged the data by the needs of the domain. 

We were able to extract historical data from the required APIs, which contained all previous match information. then cleaned it, and dumped this data into GCP Big Query.

We calculated the ELO and Power Ratings of the teams for specific parameters which got us higher performance. The bettings predictions were calculated using two parameters:-

  • Historic Data from previous matches
  • ELO and Power Rating parameters
  • Live betting was predicted an hour before the match.

After the particular calculations. We built the required predictive models according to the client’s requirements. The model was built by using more than 0.6 million data points.

Results:

The client has all the necessary data which was extracted from the APIs and then cleaned and munged accordingly and dumped into GCP Big Query. 

Now, they’re able to predict which team will win/lose 1 hr after starting the games. The client is now able to increase its user count which also saw an increase in its revenue.

The Architecture:

Related Posts