--lookback

🦉 AthenaAI is an AI-powered ensemble machine-learning trading bot for PocketOption, built on BinaryOptionsToolsV2 → Get it on GitLab
💬 Need help? Join the Chipa Discord · ⚡ Go no-code with ChipaEditor

The maximum candle history (in candles) handed to the feature engine for each training sample.

What it does

In _build_range() (src/training/dataset.py), the feature vector for candle i is computed from the slice:

chunk = candles[max(0, i - lookback):i + 1]

so each sample sees at most lookback + 1 candles of history. In the parallel path of build_samples(), each worker’s candle slice is also extended backwards by lookback candles so chunk boundaries don’t truncate anyone’s history.

Choosing a value

• The default 200 matches the bot’s live BotConfig.lookback default (reference); keep them equal so live feature vectors see the same amount of history as training did.
• Larger lookbacks slow sample generation roughly linearly (each feature_engine.compute() call processes more candles) — the training guide’s troubleshooting suggests reducing --lookback if sample generation is slow.
• It must comfortably exceed --window and the feature engine’s longest indicator (26 candles), otherwise features come back None and samples are skipped.

Examples

python train.py --lookback 200        # default
python train.py --lookback 100       # faster sample generation, less context

Interactions

• --window — the indicator window inside this history.
• --max-candles — caps the whole dataset; --lookback caps per-sample history.
• BotConfig.lookback (reference) — the live counterpart.

See also

• Feature engine concept
• Training guide

💬 Unsure how this interacts with the rest of the configuration? Ask in the Chipa Discord — or prototype strategies no-code with ChipaEditor.