Improving Crypto Short-Horizon Price Estimates with Order Book and Trade Flow

Summary

This study explores whether order book imbalance and recent trade activity can improve short-horizon estimates of cryptocurrency prices. Using exchange trade and depth records sampled at 100-millisecond intervals, it compares the squared-error estimates from the ordinary midpoint, a size-weighted midpoint, and several adjusted midpoint formulas. The tests find modest improvements from weighting top-of-book imbalance, with smaller contributions from the next depth levels.

The analysis then derives trade-flow measures from the frequency and average size of buys and sells, and combines these with book imbalance and recent price deviation. The reported squared error falls across the tested adjustments, with the combined estimate performing best in this sample. The author suggests trade flow may be more informative than posted size and notes that flow can relate differently to price in ordinary versus extreme conditions. These are exploratory, hand-tuned results from a limited dataset; the author cautions that parameters are not rigorous and may have little practical value. The estimates are not a validated trading strategy, and the evaluation does not establish out-of-sample performance or account for execution costs.

Key ideas

• Top-of-book size imbalance can shift a midpoint estimate toward the side with greater displayed liquidity.
• The next few depth levels add some predictive information, while deeper levels contribute less in this sample.
• Trade arrival rates and quantities can provide useful information about near-term price changes.
• Combining book imbalance, trade flow, and recent price deviation reduced the reported squared error in the sample.
• The formulas are exploratory and hand-tuned, so their reported accuracy does not establish live trading profitability.

Tags

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