Here is a summary of every projects I’ve worked on in my free time. Some of it is forked based on existing repo and others are original ideas. I am currently in the process of pushing everything to github. In terms of technologies, it includes python, linux, cloud and sql experience.

Poker AI

Repository: https://github.com/Dinjambes/poker_ai

Used an existing repository base on [https://www.science.org/doi/full/10.1126/science.aay2400](this paper) and deep counterfactual regret minimization to create an AI for a specific variant of poker called All-in or Fold. With this project I’ve gained valuable experience in machine learning and game theory.

  • Used Google Cloud (thank you Google Free Credit) to train multiple versions of the AI overnight
  • Built range chart as seen below to see the regret/estimated value of each starting hands depending on the position (UTG, UTG+1 following a call, UTG+1 following a fold, etc…)
  • Parsed real world poker logs to analyze more in depth
  • Implement All-in or Fold variant and add estimated value of promotions running on the site
  • Calculate All-in equity as a short-cut when training the AI
  • Modify the AI to be able to run on multiple cores faster

The project was started in 2020 and was implementing the jackpot promotion available on GGPoker at the time. The promotions have since changed and I would like to implement some changes to train the AI and see what the results are.

UTG This is the chart for the UTG regret of a trained AI model. It is intersting to note that many suite connectors are +EV. This is because of the jackpot that could be won on the website if you made a straight flush with your hand

DFS Backtesting

Repository: coming soon™

The goal of the project was to implement a backtesting application to test the theory written in [https://spiral.imperial.ac.uk/bitstream/10044/1/77220/2/DFS_Final_15Oct2019_Part1.pdf](this paper) (guess you can see a pattern here). This was to complement my machine learning knowledge with more statistics knowledge and play with sports statistics and sports odds which I both love. This project was focus on NBA DFS contest.

This project gave me experience with data scrapping, Dirichlet distribution, linear regression and monte carlo simulation

  • Used a statistical model to predict a player result in a specific game based on closing prop line for that player
  • Scrap past DFS contest data and player prop closing line data to backtest models
  • Display the result of the backtesting

Ethereum blockchain alert

Repository: coming soon™

This project was made to learn more about the the blockchain, more specifically the Ethereum one. With this project, I did a big deep dive into what Ethereum is how it works. I then setup a local node on my computer and connected a python app to it so that it could scan for spike in gas price.

  • Learn about Ethereum node and setup a local node
  • Connect to my local node and generate alert for gas price spike
  • Using my local node again, scan for the historical price of different assets on the blockchain and create graph showing the data
  • Implement Ethereum node on the cloud