Inferring Order Flow from OHLCV Data with Confidence Gating

Summary

This article develops bar-level order-flow measures for M1 futures data, where OHLCV bars do not reveal each trade’s aggressor side. It describes an OHLCV approximation of VPIN based on where a bar closes within its range, signed volume imbalance, relative session volume, a smart-money index comparing early and late session returns, and short-term flow momentum. The author relates these estimates to informed trading and price impact, while emphasizing that tick data would provide more direct measurements.

The measures are assembled into an analysis structure whose confidence is reduced by elevated microstructure noise and set to zero for sessions with sufficiently high jump intensity. The article reports that its VPIN proxy stays near the middle of its range across the studied NQ sessions, with modest elevations in stressed periods; it also reports an extreme early sample volume ratio due to the rolling baseline warming up and winsorizes reported statistics. These are research illustrations, not calibrated probabilities or proof of predictive profitability. Fixed-time OHLCV bars lose the fixed-volume bucketing property of canonical VPIN, and the article warns against treating the proxy as a statistically valid probability of informed trading.

Key ideas

• Bar closes within their high-low range are used to infer approximate buying or selling pressure.
• The OHLCV VPIN proxy is based on fixed-time bars and does not reproduce canonical fixed-volume VPIN.
• Signed flow, relative volume, intraday return shape, and flow momentum capture different aspects of activity.
• Noise and jump measures from earlier analyses reduce or eliminate confidence in the flow estimates.
• The reported NQ M1 results show modest proxy variation and do not establish calibrated probabilities or trading profitability.

Tags

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