Testing Recurrent Bitcoin Order-Flow Models Across a Market Bubble

Summary

This paper proposes a deep recurrent model that uses exchange order flow to model directional price movements at high frequency. Order flow is described as the stream of arriving orders that contributes to price formation. The central question is whether a model trained in one market regime remains useful when conditions change substantially.

To assess temporal stability, the authors train the model on data from before the 2017 Bitcoin bubble and test it during and after that period without retraining. They report that the model remains stable through the highly volatile bubble period and benchmark it against existing deep-learning approaches to price formation. The provided description does not specify the data source, precise prediction target, evaluation metrics, or the benchmark outcomes. It therefore supports a claim of stability in the reported experiment, but leaves open how broadly that finding generalizes to other assets, venues, or market regimes.

Key ideas

• The proposed deep recurrent model uses exchange order flow to model high-frequency directional price changes.
• The experiment trains on pre-bubble Bitcoin data and tests during and after the 2017 bubble.
• The model is evaluated without retraining through the more volatile period.
• The authors report temporal stability and compare the model with existing deep-learning approaches.
• The available description omits metrics and data details needed to judge broader generalization.

Tags

# Deep Recurrent Modelling of Stationary Bitcoin Price Formation Using the Order Flow


# Deep Recurrent Modelling of Stationary Bitcoin Price Formation Using the Order Flow









In this paper we propose a deep recurrent model based on the order flow for the stationary modelling of the high-frequency directional prices movements. The order flow is the microsecond stream of orders arriving at the exchange, driving the formation of prices seen on the price chart of a stock or currency. To test the stationarity of our proposed model we train our model on data before the 2017 Bitcoin bubble period and test our model during and after the bubble. We show that without any retraining, the proposed model is temporally stable even as Bitcoin trading shifts into an extremely volatile "bubble trouble" period. The significance of the result is shown by benchmarking against existing state-of-the-art models in the literature for modelling price formation using deep learning.

Shown in full with attribution under the source's licence. Licence: abstract CC0

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.