About a year ago I read about an interesting trading strategy. I found it interesting and quite profitable because it did not use graphical indicators and other junk. The strategy concept is based on the stock market and correlation with the futures market. Speaking in more detail, there is a stock market in which there are certain indices (NYSE, DOW and NASDAQ) and with certain parameters of two or more indices with a probability of about 70%, a predictable correlation occurs in the futures market (ES and NQ). The screenshot below roughly reflects the correlation of one market with another. It was decided to automate the trading strategy and test it. Predictably, the strategy showed very good results in the backtest. Taking into account all factors (commission, slippage, P/L 3-1), the result was close to reality. I tested it on a demo account, the results became even more similar to the harsh truth, but there is still a profit. If you support this topic and this idea, I am ready to provide a limited number of strategies written by me with source code for conducting tests and receiving feedback from you. Feedback will be very important for the final writing of a full-fledged strategy. NOTE: In order to generate the sequence you need a data feed that gives you: ^TICK, ^TICKDJ and ^TICKQ, plus data for the E-mini markets you intend to trade.
You can't profit from correlation* ** you need cointegration or one index to lead the other. GAT * pendantically there are derivatives whose prices are a function of correlation levels ** though better correlation forecasts will improve risk targeting and management
The CME gives a 70% margin credit (performance bond offset) on a position thats long/short pairs of indices. Long/short NQ against ES will give you a margin discount of approximately 70%, so the CME knows the indexes are correlated and provides a capital efficient way for you to lever the spread (differential of contracts). The market internals are useful, but you need real technology and experience to use them.
Generally you're right but it might be possible because I read a paper where someone actually derived the trading strategy based on the time varying roughness index which is a similar measure to correlation but not quite the same but basically they would buy rough sell smooth or it was the universe one of the other anyway I thought it was a big waste of time and not very efficient and not well motivated by underline principles. Even if you find a set of co-integrated variables they need to be such that they're co-integrating factors are stable over time . Also I've never liked time series or discrete time methods
Even if you find a set of co-integrated variables they need to be such that they're co-integrating factors are stable over time . Also I've never liked time series or discrete time methods[/QUOTE] Amen. Good for science. Bad for finance. The only people getting rich using ts methods in finance are selling the textbooks.
This is a 3-Yr correlation of the major indexes: But if you tried to use a simple Martingale to trade the Russell/Nasdaq in 2023 straight up, you would've gone bust : You are going to want to create your strategy based on a Gaussian Copula distribution model:
You need a lagging variable for correlation to work. Otherwise, you would already know the direction of the asset and correlation would be irrelevant.