Principles

Algorithmic trading, also known as black-box trading, is the use of high-speed computing infrastructure to automatically generate market or limit orders based on incoming information. RSJ uses algorithmic trading for both market making and directional trading.

The trading algorithms used by RSJ are based on sophisticated mathematical models. The models are an outcome of applying the results of deep theoretical mathematics on financial markets. A team of RSJ mathematicians continuously improves the models and creates new ones. The input of these models are the tick-by-tick data from the traded market as well as from other sources. The parameters of the models are calibrated by using advanced statistical techniques. Every day we log tens of gigabytes of tick-by-tick market data.

As far as the mathematical models is concerned, it is important for them to be robust. Since the financial markets are not stationary it is necessary to update and calibrate the models frequently by using the most recent data. The optimal modeling approach is a compromise between statistical estimation error and stationarity of the financial market. In order to minimize the estimation error one would like to use as much data as possible, while avoiding the effects of nonstationarity of the price series calls for using as short historical data as possible. Note also that a proper modeling approach requires considering jumps in financial time series (so-called jump-diffusion models) since prices change by jumps frequently. It is very dangerous to make the assumptions of continuous price dynamics. This is especially true in todays environment.