Visualizing Episode History

All offline environments support render_history() to visualize episode performance after training or evaluation. The method automatically detects the environment type and renders appropriate plots.

Overview

The render_history() method is inherited from the TorchTradeOfflineEnv base class and provides consistent visualization across all 3 offline environments:

• SequentialTradingEnv
• SequentialTradingEnvSLTP
• OneStepTradingEnv

The visualization automatically adapts based on the trading mode (spot vs futures).

Usage

# After running an episode
env.reset()
for _ in range(episode_length):
    action = policy(obs)
    obs, reward, done, truncated, info = env.step(action)
    if done or truncated:
        break

# Visualize the episode
env.render_history()  # Display plots with buy-and-hold baseline

# Without buy-and-hold baseline
env.render_history(plot_bh_baseline=False)

# Or save to a variable for later
fig = env.render_history(return_fig=True, plot_bh_baseline=True)
fig.savefig("episode_history.png")

Visualization Types

All environments render 3 subplots:

1. Price History with Actions: Shows the asset price over time with long/short actions marked as green upward triangles (long/buy) and red downward triangles (short/sell). Position closes are shown as orange (long close) and cyan (short close) markers.
2. Portfolio Value vs Buy-and-Hold: Compares your agent's portfolio value against a simple buy-and-hold baseline strategy.
3. Exposure History: Visualizes exposure percentage over time with green areas for long exposure, red areas for short exposure, and flat sections for no position.

Parameters

Parameter	Type	Default	Description
return_fig	bool	False	Return matplotlib figure instead of displaying
plot_bh_baseline	bool	True	Show buy-and-hold baseline comparison

Implemented in torchtrade/envs/offline/base.py. Requires matplotlib.