thanks cooper
I understand the difference between the 2 trades - but this forum is about stat arbs - and that's what I'm concentrating on here.
I guess I'm essentially questioning the relevance of the correlation data as it's looking back over an extended period - in view of the fact that the latest earnings report and even a single price sensitive recent announcement can throw an instant spanner in the works.
So there may be pairs that are given a strong 80 or 90% correlation by the software program that may essentially be no longer correlated... and should no longer even be considered a pair.
How do we determine which pairs indeed remain correlated and to what extent?
And even if you are not trading during this reporting season as you say, how do you determine the
best correlated pairs and the relevance of the data when you do start to trade again - as you would still be largely looking at past correlation data - that's inescapable.
Hey mate,
I have been trying to learn a lot about stat arb since the commencement of this thread (what can I say - I'm an engineer), so I'll have a stab at answering this.
The only way that correlation data can be generated is by looking back through historical data. Only by analysing the way that prices have moved in relation to each other over an extended period of time can an accurate assessment be made of whether or not there is a statistical relationship between the two equities.
Indeed, it has been previously mentioned that the backtesting ability of the software will be enhanced to greater than 12 months - which will then allow for correlation to factor in previous reporting periods.
You are indeed quite right that price sensitive announcements may have a significant impact on short term price movements - but by this logic you could argue that all technical analysis is inherently questionable. After all, whether you use trends, breakouts, support and resistance, whatever - always you are relying on analysis of prior price movements and "hoping" that any identified patterns are repeated.
As PT has mentioned a number of times - this is a numbers game. By determining a high previous correlation, the numbers suggest that the probabilities are high that prices will move together in the future. However, nothing is guaranteed - and all this method does is provide an mathematic edge - as with any other analytical method - which can then be exploited through appropriate application of money management techniques.
You are right - it is inescapable that you are looking at past data - but as with any mathematical technique all you are trying to do is put the odds in your favour by identfying as strong relationship and betting / speculating / gambling that that relationship will continue into the future.