High-frequency markets whisper. Introducing a research paper
Instantaneous Volatility Seasonality of High-Frequency Markets in Directional-Change Intrinsic Time is a recently published article by Vladimir Petrov (PhD research fellow of Marie Sklodowska-Curie initiative BigDataFinance at the University of Zurich), Anton Golub (Flov Technologies AG) and Richard Olsen (Lykke Corp). We invited Vladimir Petrov, who is also a member of Lykke Research Hub team, to introduce the research in our blog.
Why did you get interested in the topic?
For decades price activity has been of interest for not only academic researchers but also industry practitioners whose income is tightly connected to the price fluctuations. That activity, also called volatility, serves as a proxy of the market’s stability and reflects the price tendency to vary over time. Stability is essential for high-frequency markets as much as it is essential for the overall financial system (just think about the fact that even ten years after the global financial crisis we are still experiencing its aftermaths). Traders should always carefully weigh their decisions if they do not want to erode the fragile market’s balance and thus to get financially hurt after performing a set of particularly important transactions.
In our recently published research work, we made efforts to disentangle the very core volatility properties of several high-frequency markets by employing the directional-change intrinsic time concept.
How would you, in short, describe your research to a person who is not a specialist in your area?
We estimated high-frequency market volatility by using event-based time. Ticks are dictated solely by price curve moves according to the selected concept. Revealed by the novel approach volatility is conventionally called instantaneous volatility. As you could guess from above, the volatility of financial time series is traditionally measured in physical time. One always has to have data collected over a non-zero period of time to perform the “classical” volatility estimation: historical prices changes over a day, a month, a year, etc. But irregular economical events, highly unexpected political decisions, or nature surprises, shaking the financial system, cannot be efficiently captured by the regularly spaced price change measures. This is where the directional-change concept manifests itself in all its glory. This concept is fully independent of the price changes speed and frequency defined in physical time. Only the sequence of alternating price curve reversals defines the evolution of the given exchange rate. As a result, the concept never misses the most important information pronounced in the substantial price variations.
Were there any results that puzzled or amazed or disappointed you?
First, the shape of the instantaneous volatility seasonality substantially depends on the scale of the observation. It appears to be crucial to think about the size of the directional-change threshold selected for analysis. To understand the importance of that observation, think about the classical volatility estimator. The classical estimation technique does not restrict the frequency of observations used to compute the volatility size. The period can be a minute, an hour, a day or even larger but the eventual result will always be the same. One just needs to make a tiny adjustment, to “annualize” the measured volatility by multiplying it by the square root of the number of observations fitting a calendar year. It is conventional to say that classical volatility is able to scale with time. But the instantaneous volatility, as we show it in the paper, is very sensitive to the size of the selected threshold. Depending on whether you observe a fraction of a percent, one percent, or ten percent price moves, the volatility and its seasonality change the shape. It is difficult to underestimate the importance of that scaling impact. When someone does comprehensive research by observing a price evolution, or when one employs tricky risk management or trading algorithms, it is typically convenient to select some parameters, some thresholds, some time intervals. The performed analysis shows that “some” is too general when we are talking about the intrinsic time concept. You do need to be specific about the trend size selected for the observations.
No more spoilers here, you are welcome to go forward and download the paper which is freely accessible online. All experiments and expected as well as unexpected discoveries are thoroughly described there.
For whom can your research be useful?
The volatility size is the omnipresent parameter in the majority of risk management, hedging, and trading tools. The novelty of the proposed estimator and the revealed seasonality patterns can be directly used for building more comprehensive and universal algorithms. We are confident that the provided research is essential for high-frequency traders, market-makers or those who price derivatives for any underlying and of the arbitrary time to maturity. Although there are no particular trading or option pricing algorithms provided in the current research paper, the results of our work will hopefully be used for new efficient liquidity provision algorithms, cheaper, and safer priced derivatives. We uploaded all developed programs used in the work to a public GitHub repository so people could save some time learning the intrinsic time world. The link to the repository is also provided in the original manuscript.
With the high level of confidence we can state that as finance is an essential component of everyone’s lives, society in general can benefit from the conducted research. Whether you want it or not, price changes and trillions of the trading volume indirectly affect the financial stability of your country, the profitability of each household, and, as a result, your very life. That is, in one way or another everyone experiences instant shocks and long lasting markets’ stability. So, you’d better be aware of the most advanced tools to readily deal with the next surprise the market is going to test us by.
