Order Arrival Clustering and Real-Time Parameter Updates in Crypto Markets

Summary

This analysis tests whether cryptocurrency futures trade arrivals follow a Poisson process, using aggregated YGGUSDT trades from a volatile trading day. It examines buy-side trade intervals and the number of orders observed per second. The observed distributions differ substantially from Poisson expectations: intervals show local peaks, while rare high-activity periods occur more often than the model predicts. The author suggests that periodic algorithmic activity, changing arrival rates, and interactions between orders may explain the deviations.

The article then evaluates a rolling mean as a simple real-time forecast for order frequency and traded volume. In the reported sample, using recent observations—particularly a short two-second history—reduces forecast residual error compared with a fixed average. This supports adapting market parameters as activity changes. The evidence comes from one asset and one date, and the rolling-mean example is explicitly introductory; broader validation and more advanced time-series methods would be needed before using it in a trading system.

Key ideas

• Trade arrival intervals and per-second counts in the sample diverge from Poisson assumptions.
• Order activity clusters, with changing rates and possible event dependence contributing to extreme counts.
• A rolling mean can update estimates of order frequency and volume as conditions change.
• The reported short-history forecast has lower residual error than a fixed-average prediction in this sample.
• The analysis covers one asset and date, so its results require broader validation.

Tags

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