Risk Management on High-Frequency Trading (HFT)

The tools to measure risk, VaR and Stress for example, can also be used for higher frequencies but should be considered the fact that they have a non-linear behavior more pronounced on intraday returns, requiring a more refined analysis.

The big issue I see is to understand the strategy used to then understand what the most appropriate methodology to present a risk to the decision maker.

The HFT universe expands the universe of data exploration, with respect to risk some phenomena must be considered, for example, effects of liquidity throughout the day.

Financial Risk Management

On Volatility versus Risk I commented on Risk and Volatility, without however mentioning practical activities related to the direct use of metrics for risk. 

 

Market risk, which is the possibility of financial loss due to the typical ups and downs of the market, should, as the intuition suggests, reflect the degree of oscillation that the asset provides.

The simplest metric, known and used for this purpose is the standard deviation. The standard deviation has inaccuracies in quantifying the risk, because strictly financial returns do not follow a normal distribution, which is the necessary premise to calculate the standard deviation. The metric of the standard deviation necessarily produce a bias, always deviated from the true value.

The biggest problem with the standard deviation, however, is not the premise that it performs when considering the normal data. The standard deviation by itself does not present the risk management of an efficient way, because it still lacks the statistical scenario was considered.

Value-at-Risk

The methodology Value-at-Risk (VaR) fills this gap. It was developed in order to produce management reports more objective and intuitive. His metric is well established in statistical terms and reflects exactly what one would expect from a market risk measure. The VaR results in a net asset value, or a possibility of financial loss. This possibility of financial loss is tied to a time interval and statistical significance. So we say: The portfolio has the possibility of losing $X on Y days (or hours) with a statistical significance of Z%. 

 

Financial processes, however, do not follow a behavior said homoscedastic or constant over time but heteroscedastic. The volatility can undergo dramatic changes as in times of crisis. These phenomena can cause a sharp drop in equity of a particular institution leading to bankruptcy in many cases. 

Stress 

An analysis complementary to VaR, called Stress analysis is used to simulate situations of panic and financial crises. These simulations analyze the tails of the distributions of returns and evaluate the portfolio in more extreme scenarios. It is possible to simulate historical crises, specific situations and scenarios to highly pessimistic and uncorrelated. 

It is necessary to understand the main feature of the market is the strong stochastic process which asset prices are submitted. Unexpected situations are possible and even expected, and it is impossible to predict turbulence and crisis situations with sufficient precision. The Financial Risk Management, however, has enough tools to evaluate and quantify the risks and fluctuations to which the portfolio is submitted.