One Sentence Summary:

AI Trading Bot uses state-of-the-art Deep Machine Learning method to train a trading bot to Buy/Sell Crypto Currencies and US Stocks


Layman Version:

Trading is complex and it requires human to digest information both from individual stock price change and market trend. But this task is better handed to machine that is able to do it effectively and efficiently. 

Our bot is designed to catch uptrend and at the same time resist loss due to a large down trend. Our bot is able to short stock or crypto currencies at the right trending period. 

It takes a lot of computing power to continue to work and calculate the optimal configuration to maintain our bot for trading. That is why we need your support.

Join My Discord !


Our goal and to-do milestones

so far around 2,500 hours has been invested to build this project, including an App.

  • (done) Methodology research

  • (done) Math required for construct a trading model (in-depth)

  • (Alpha) Trading bot

  • (Alpha) backtesting bot

  • (On-going) Training/testing data preparation

  • (On-going) StockTrading-v2 is near completion.

  • (On-going) sand-box trade

  • live trade, all previous items integrated !


Features marked with their tiers. 
Higher tiers get ALL features from lower ones.

👑 Supporter 💎Diamond 🥇 Gold 🥈 Silver 🥉 Bronze



Technical Version
(Skip this if you are not technical person):

We uses Algorithms such as DDPG, reinforcement learning, feature extraction, numerous filters and market data to train our bot. A model and a backtesting bot and a live trading bot is developed to connect to real market.



fig1. Profit return based on two input parameters. 


fig2-1 We force the robot to only 'Buy' (no shorting of cryptocurrency), it tries to avoid making major loss.

When market is trending upward, it tries to catch it, but when market is trending downward, we can avoid trading it with Selection Techniques.


Fig2-2 and Fig2-3 shows our trained bot is used with another cryptocurrency without any modification. It proves that the model is generic and is not restricted to the original trained cryptocurrency. Both 'Buy' and 'Shorting' of cryptocurrency are enabled so it can gain profit even during downtrend.

fig2-2: Transfer learning for BCH

fig2-3: Transfer learning for LTC

Fig3 shows a Machine Learning chart that we had. Green arrow shows that we pick the model when it reaches highest result.


Note: We only disclose limited content due to the fact that the project is not totally completely. But we will publish detail test result in our monthly report.





video01: video for trading: https://streamable.com/xsvoro


fig4: final gain for the video above


Progress Update

Please visit our AI Hedge Project blog to get latest update.

We update our work progress and keep you updated.

https://medium.com/ai-hedge-project


Thank you for supporting us from the very beginning !



Common Q & A

  • Q: Since the bot seems to be earning money, why do you still need to raise fund?
    A: One man vs. team work vs. enterprise scaling affect how long a business entity can last. My aim is to make it last. To be honest, this trading model requires a lot of computation power and all financial model require constant monitoring and maintaining. Besides, I would like to develop more of this as trading model should adapt to market, which requires more funding to keep this ball rolling smoothly.

  • Q: Do you trade US stock?
    A: Yes, in fact, our coming StockTrading-v2 (current version) will include both crypto currencies trading and US. stock trading.