Train a FreqAI model and backtest rolling windows

Goal

Run a reproducible FreqAI backtest and verify the models trained in each window against their subsequent predictions.

Steps

1. Start from FreqaiExampleStrategy and config_freqai.example.json; set LightGBMRegressor, freqai.identifier, and the training window. Ensure features and labels use only information available at each point. [freqai]
2. Download data covering train_period_days, startup_candle_count, and the test interval so the first test candle has sufficient training history. [running]
3. Run freqtrade backtesting --strategy FreqaiExampleStrategy --strategy-path freqtrade/templates --config config_examples/config_freqai.example.json --freqaimodel LightGBMRegressor --timerange 20210501-20210701; save the config and result. [running]

How to verify

Confirm logs show a model trained for each backtest window; check that the first test candle follows the available training history and that results match the chosen identifier.

Troubleshooting

For missing-history errors, extend the downloaded range or pre-test training data. If a rerun retrains unexpectedly, inspect identifier and model cache. Do not deploy the example strategy live.