Advanced HMM - 3 States Complete

Hidden Markov Model
Aconsistent challenge for quantitative traders is the frequent behaviour modification of financial
markets, often abruptly, due to changing periods of government policy, regulatory environment
and other macroeconomic effects. Such periods are known as market regimes. Detecting such
changes is a common, albeit difficult, process undertaken by quantitative market participants.
These various regimes lead to adjustments of asset returns via shifts in their means, variances,
autocorrelation and covariances. This impacts the effectiveness of time series methods that rely
on stationarity. In particular it can lead to dynamically-varying correlation, excess kurtosis ("fat
tails"), heteroskedasticity (volatility clustering) and skewed returns.
There is a clear need to effectively detect these regimes. This aids optimal deployment of
quantitative trading strategies and tuning the parameters within them. The modeling task then
becomes an attempt to identify when a new regime has occurred adjusting strategy deployment,
risk management and position sizing criteria accordingly.
A principal method for carrying out regime detection is to use a statistical time series tech
nique known as a Hidden Markov Model[5]. These models are well-suited to the task since they
involve inference on "hidden" generative processes via "noisy" indirect observations correlated
to these processes. In this instance the hidden, or latent, process is the underlying regime state,
while the asset returns are the indirect noisy observations that are influenced by these states.