Important note: The code here is not optimised to be used in market conditions, it is instead should be used as an educational tool to teach key prinicples.
- data:
- Contains all the csvs used for backtesting.
- intro_agents:
- Contains 3 simple intro agents demonstrating key pedlar functionality.
- agents:
- Contains the agents produced during the sessions. Frequently redesigned with improvements and optimisations.
- Decision tree agent
- Linear agent
- NN agent
- RNN agent
- RL agent
- Explicit search
- Random search
- Grid search
- Gradient-based search
- Evolution-based search
- Sharpe Optimisation
- Ensemble of agents
- K-Armed bandits
- Time dependent behaviour (e.g. varying behaviour around market open/close)
- Real time statistics
- Add post-analysis of trades made.
Imperial pedlar server → (requires access to Imperial Wifi Network)
You can contact us at: algo.trade@imperial.ac.uk
