High Frequency Trading: Markets, Practice and Models.

On an introductory note, High Frequency Trading (HFT) can simply be defined as a trading technique which, through the use of sophisticated computer systems and algorithms allows us to elaborate buying and selling orders in a matter of milliseconds, taking advantage of the slightest imperfections in the markets and/or the difference in price of the same bond quoted on various markets. Obviously HFT cannot be satisfactorily defined in so few words, and the aim of this work is in fact to present, as precisely as possible, a concept that has been around for some time now but is still unknown to many. Since HFT concerns 75% of the exchange volume of the USA markets, and is contagiously spreading to European markets, one of the aims of this work is to ascertain whether HFT is an advantage or a disadvantage or both for the stability of the global financial system.

In this sense, the starting point can be what is known as “Flash Crash” day that is May 6th 2010, when a selling order arriving from a high frequency software provoked losses for over 60% in some ETF bonds quoted on the USA market, before they in their turn could just as quickly recover and consolidate (more or less) initial levels. In this sense, therefore, we will endeavour to understand if this new trading technique can be controlled by man of if there is a real risk of it slipping out of hand or being manipulated, provoking even more serious disasters than that in May 2010. In fact, some paragraphs in the second chapter have been dedicated to illustrating the associated regulations.
Therefore to sum up this thesis, without any research or simulation activity, it tries to explain the basic concepts of the modern trading technique which is slowly (but not too slowly) expanding into the world's major financial markets. This will be possible thanks to the understanding of the mathematical models dedicated to the volatility and latency, which have been illustrated in the main part (chapters 3-4-5). The former are particularly important since they were suggested with the aim of foreseeing and checking possible market turmoils like those 3 years ago. The latter will help to understand the reason why we are induced to looking for low values of the same with more and more insistence.
Concluding, and in extreme synthesis, the structure of this work is as follows :

1)The first chapter emphasises the concept of High Frequency Trading, and the two major concepts connected to it, that is “latency” and “co-location”. Furthermore, this chapter shows the most common strategies used, both legal and illegal, and the most important advantages and disadvantages connected to HFT. Particular consideration has been given to the MTA situation, that is the Italian one;

2)In the second chapter, the events of the first (6th May 2010) and the second (23rd April, 2013) Flash Crashes are explained in detail. During these events many bonds and ETFs underwent great losses in a matter of minutes, plunging losses (> 60%), and then returned to pre-crash level. In particular paragraph [2.5] shows the measures taken to avoid similar crises happening again . All this supported by the inquiry made by the CTF and the SEC;

3)The third chapter starts to tackle the concept of HFT from a more mathematical point of view. It introduces two of the most famous measures devised with the aim of preventing (or trying to) dangerous turmoils in the market: the PIN measure and it's later development of the VPIN measure (and its variants), were both developed by [Easley et al.];

4) The fourth chapter contains a structured criticism of the two mathematical measures illustrated in chapter 3, promoted by [Anderson and Bondarenko], in “VPIN and the flash crash“ (2013). The entire chapter shows the results of research carried out by [A&B] and the conclusions arrived at as regards reliability or unreliability of the systems;

5) The fifth chapter, together with the two preceding ones, is the heart of this work, since it shows on a theoretical level and with practical examples, the latency models developed by [Moallemi and Sağlam];

6) Finally, the sixth and last chapter, contains the final observations and a comment on the conceptual subject of this thesis.