Strategy #0000000117 · Machine learning and sequential decisions
Reinforcement-learning trading
Enter three position actions, corresponding return changes and turnover cost to calculate cumulative reward.
Reinforcement learning frames trading as a sequence of states, actions and rewards. An agent can choose positions, then receive rewards from subsequent price changes less turnover costs. Deep Q networks and policy gradients can learn policies but may overfit simulated environments and omit market impact or non-stationarity. This small scenario scores supplied actions; it does not train an agent.
No agent training, state features or market data are supplied; entered actions are not a learned policy.
Three ways in
Understand the story. Then inspect the risk.
Start with a familiar situation, check the payoff, then test how the numbers change.
The idea · in two minutes
What reward follows position actions through price changes after transaction costs?
Reinforcement learning frames trading as a sequence of states, actions and rewards. An agent can choose positions, then receive rewards from subsequent price changes less turnover costs. Deep Q networks and policy gradients can learn policies but may overfit simulated environments and omit market impact or non-stationarity. This small scenario scores supplied actions; it does not train an agent.
Enter three position actions, corresponding return changes and turnover cost to calculate cumulative reward.
No agent training, state features or market data are supplied; entered actions are not a learned policy.
- Input assumptionsEnter three position actions, corresponding return changes and turnover cost to calculate cumulative reward.
- Model outputNo agent training, state features or market data are supplied; entered actions are not a learned policy.
Illustrative research examples only. They are not live signals or a recommendation.
The mechanics · payoff and exposure
r_t=a_{t-1}\Delta p_t-c|a_t-a_{t-1}|
a is position exposure, Δp the next-period asset return and c cost per unit of position change.
Greeks: the directional intuition
- 模型
- Measures sensitivity to implied volatility.
- 資料
- Measures sensitivity to implied volatility.
- 執行
- Measures sensitivity to implied volatility.
r_t=a_{t-1}\Delta p_t-c|a_t-a_{t-1}|
Practical context · published observations
Reinforcement-learning trading: read the record with its caveats.
The chart uses the published figures or explicitly labelled study inputs below. A historical index or model replay is not the same as an investor’s realised account return.
No agent training, state features or market data are supplied; entered actions are not a learned policy. Source ↗ · Methodology ↗
Change assumptions and inspect the scenario
Inputs illustrate the strategy mechanics; they are not live quotes, forecasts or investment advice.
Variants · same family, different risk levers
Data quality, constraints and implementation costs affect results.
Compare modelling and execution assumptions.
| Related strategy | Relationship | What changes | New cost or risk |
|---|---|---|---|
| Base scenario | Manual inputs | Reinforcement learning frames trading as a sequence of states, actions and rewards. An agent can choose positions, then receive rewards from subsequent price changes less turnover costs. Deep Q networks and policy gradients can learn policies but may overfit simulated environments and omit market impact or non-stationarity. This small scenario scores supplied actions; it does not train an agent. | No agent training, state features or market data are supplied; entered actions are not a learned policy. |
| Conservative version | Lower exposure | Raise cost/risk assumptions and constrain positions. | May reduce participation. |
| Validation version | Out-of-sample test | Freeze the method before evaluation on untouched data. | Needs more clean data. |
No agent training, state features or market data are supplied; entered actions are not a learned policy.
Read beyond the summary
References
Primary sources and research used to describe the strategy and its limits.