◦ QUANT·RL TERMINAL

DQN / PPO / A2C — reinforcement-learning backtester

engine online

Asset

AAPL works on any timeframe/dates EXCEPT its daily 2020–2022 training window.

Timeframe

4h uses the last ~2 years only (intraday data limit).

Starting capital ($)

Start

End

Model

DQN

PPO

A2C

Models pre-trained on AAPL (daily, 2020–2022, 1M steps).
Backtests run out-of-sample — instant, no training. AAPL is allowed except on its exact daily 2020–2022 training window.

ⓘ About the models

DQN — learns the expected payoff of buy vs sell and picks the better. Most active trader; strongest in training.

PPO — learns buy/sell probabilities, nudged in small safe steps. Moderate and steadier.

A2C — an “actor” picks moves and a “critic” grades them. Data-hungry; often very passive (few trades).

All three were trained once on AAPL daily, 2020–2022. Here they trade out-of-sample on whatever you choose — so the results show how well (or badly) they generalize to new assets and timeframes.

◢◤

Configure a run and hit RUN BACKTEST.

The agent will trade the asset and report its current call.

RUNNING…