In algorithmic trading, you need solid quantitative models.
State space is a approach that has been very successful in control theory.
In the last few decades state space models have also become popular in finance.
Kalman Filter is a popular state space recursive algorithm.
Kalman Filter can be used to predict price multi-step ahead.
I have written this blog post in which I explain Local Linear Trend AR State Space Trading System.
The main aim is to predict price 25 hours from now using hourly data.
First I build a local linear trend model then I add an autoregressive model on to it.
Using structural time series state space model, we can break price into different components.
Like the trend, autoregressive, cyclical, seasonal, exogenous etc.
We can also model structural breaks and outliers using state space models.
Read the post I explain how to build a state space model for currency pair price.
