Intraday Project
I created a package which provides gym compatible environment to simulate intraday trading based on stream of trades, either historical or real-time.
This project was inspired by TensorTrade, but it was written from scratch, and it is a completely original source code.
The main idea was to go deeper from candles to the actual stream of trades. Because candles lose a lot of information, for example:
- How many trades during a period were initiated by buyers or by sellers?
- At what price did most of the trades happen during a period?
- At what price there were almost no activity?
- Were there many trades with small amounts or less trades with big amounts?
- etc.
A lot of trained models seem to perform good during training, but often fail to show any positive result in real trading. I believe one of the reasons is that they do not take into account some important details, like:
- orders are executed with delays
- there is always a spread between bid and ask prices
- limit orders may not execute even if price touches them
- accounts are liquidated if they go bankrupt
- trades commissions
This project tries to simulate trading environment as realistically as possible. This allows to train trading bots using reinforcement learning algorithms.