Reinforcement learning provides a solution
to this problem by modeling portfolio
selection as a decision-making problem in
an environment with unknown, dynamic,
and stochastic return distributions
Solution
The problem in portfolio trading is to
optimize the selection of assets and positions
to maximize returns while taking into
account uncertainties, transaction costs, and
market impact
Problem
• Portfolio allocation refers to the process of dividing an investment portfolio among instruments
• Reinforcement learning is a type of machine learning in which an agent learns to interact with its
environment in order to maximize a reward
• The agent learns by taking actions and receiving feedback in the form of rewards or penalties. The goal is
to find an optimal policy (actions given states) that maximizes the expected cumulative reward
• Applying reinforcement learning to the problem of portfolio allocation can increase the likelihood of
achieving long term alpha, higher returns and lower
risks
• Alpha Machine is using state of the art reinforcement learning algorithms and is built on top of
QuantConnect, the leading quant solution
• Alpha Machine is flexible to trade any asset categories, such as stocks, etf, futures, forex etc.
• Alpha Machine can be used in full auto mode (trading automatically) or as an expert system/decision
support system to analysts
10
Procedure
• Download fresh data
Update
Train
• Train new models on fresh
data
Trade
• Trade top models ranked
by specific kpi (net profit,
risk adjusted returns etc)
Test
• Back test new models on
new data producing
various kpi’s
10
§ The system had been tested in the forex market with brokers such as
oanda and fxcm
§ With conservative risk, leverage of 10:1, the system produced monthly
returns of 5-10%
§ With higher leverage of 20:1, returns are 10-20% with higher
drawdowns and standard deviation
Leverage 20:1. In Dec 2022 100 pips stop loss was
tested
