If the model is being fit on the historical data + new data, doesn't it mean it'll perform less accurately when tested? https://github.com/tradytics/surpriver/blob/52417660e44c9533ba4dfc49f0dcb45c885d05dc/detection_engine.py#L203 https://github.com/tradytics/surpriver/blob/52417660e44c9533ba4dfc49f0dcb45c885d05dc/detection_engine.py#L204