Detecting Non-Random Market Content with Multivariate Analysis

Summary

The document argues that financial returns can appear close to random under traditional time-series analysis while still containing patterns detectable with modern multivariate statistics. It describes an analysis inspired by methods used in nuclear physics, applied to high-frequency movements in Euro futures. The identified component is trend-following behavior that varies with volatility ranges, and the work uses a trading system to encode the relevant conditions.

To examine whether the pattern extends beyond one contract, the same procedure is applied to DAX and cacao futures, described as largely uncorrelated with the Euro future. Similar results across these examples are presented as evidence that the system captures recurring features of market behavior over a ten-year period. The authors explicitly caution that these examples do not establish generality. The excerpt gives no detailed statistical tests, trading costs, or risk-adjusted performance figures, so the claimed non-random content should be treated as a sample-specific finding rather than a universal or proven trading edge.

Key ideas

• Traditional time-series methods may fail to detect small departures from random-walk behavior.
• The analysis uses multivariate statistical methods to examine high-frequency futures data.
• A trend-following component associated with volatility ranges is identified in Euro futures.
• Similar findings are reported for DAX and cacao futures using the same trading-system ingredients.
• The examples cover a ten-year period, but the document cautions that they do not establish generality.

Tags

# About the non-random Content of Financial Markets


# About the non-random Content of Financial Markets









For the pedestrian observer, financial markets look completely random with erratic and uncontrollable behavior. To a large extend, this is correct. At first approximation the difference between real price changes and the random walk model is too small to be detected using traditional time series analysis. However, we show in the following that this difference between real financial time series and random walks, as small as it is, is detectable using modern statistical multivariate analysis, with several triggers encoded in trading systems. This kind of analysis are based on methods widely used in nuclear physics, with large samples of data and advanced statistical inference. Considering the movements of the Euro future contract at high frequency, we show that a part of the non-random content of this series can be inferred, namely the trend-following content depending on volatility ranges. Of course, this is not a general proof of statistical inference, as we focus on one particular example and the generality of the process can not be claimed. Therefore, we produce other examples on a completely different markets, largely uncorrelated to the Euro future, namely the DAX and Cacao future contracts. The same procedure is followed using a trading system, based on the same ingredients. We show that similar results can be obtained and we conclude that this is an evidence that some invariants, as encoded in our system, have been identified. They provide a kind of quantification of the non-random content of the financial markets explored over a 10 years period of time.

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.