Next-Tick Prediction and Order Placement in Chinese Futures

Summary

The document frames a high-frequency trading problem in Chinese futures: low liquidity and large price fluctuations make order execution difficult, while market data arrives in half-second updates. The author asks how to predict the next tick or improve order placement when activity between updates is unobserved. This delayed, discrete view of market information may limit methods designed for continuous streams.

The author reports trying a staged logistic-regression approach to estimate whether a price changes, its direction, and then its size, with a geometric distribution for the final step. They also mention arithmetic and geometric Brownian motion and a mean-reversion approach involving a later-dated futures option. These are experiments rather than supported recommendations: the document gives no features, sample design, performance results, execution tests, or comparison against simpler order-placement rules. Its main value is defining the data-timing constraint and candidate modeling directions, not demonstrating a reliable predictor.

Key ideas

• Chinese futures data in the described setting updates every half second, leaving price activity between updates unobserved.
• The trading problem combines next-tick prediction with the practical challenge of getting orders filled.
• The author decomposes logistic regression into movement occurrence, direction, and move size.
• Brownian-motion and mean-reversion approaches are also mentioned as experiments.
• No validation results are given to establish that any proposed approach improves fills or trading performance.

Tags

# Estimating the next tick movement in Chinese markets


# Estimating the next tick movement in Chinese markets












I'm working on high frequency trading in the Chinese Futures market and I've been having a bit of trouble with getting orders to go through due to the lack of liquidity and large fluctuations. To tackle this problem, I was thinking about working out a model to somewhat predict the next price tick so that I can send my orders accordingly and achieve a higher percentage of successful trades.

The problem is that Chinese futures markets are different from other markets because data is only released every half a second. Rather than a continuous stream of information, quotes and other information is updated every half second so the time in between is a sort of black box which makes some models hard to apply. I was wondering if anyone has some suggestions for either how to predict the next tick (which models or variables to try) or a better way to place my orders to increase the number of successful trades.

Things I've tried so far are logistic regression (decomposed to guess 1. if there will be a price change 2. given there is, is it up or down 3. if its up or down, how much will it move using geometric distribution), arithmetic brownian motion, geometric brownian and mean reversion with another future option for a later date (most hopeful).

Thanks in advance

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

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