What does correlation mean to you?

I'm going live with a new EA and I'd like to apply it to as many currencies as possible. Right now the topic of correlation is in my mind. I don't understand why we focus on correlation between currency pairs. Should we not focus more on the correlation of a system producing signals for two currencies? Who cares if the currencies are highly correlated if the results of the trades are not similar whatsoever?

I use different settings for different currencies based on that currency's volatility. There are many instances where two correlated pairs will not place the same signals in that day. OTOH, yes, there are days when the currencies trigger similar trades. Their results might end up differently though. One might straight out win, one might win a little, one might lose a little, one might straight out lose. Is that what we should be focused on?

In that instance, would I be able to apply the EA to almost every currency? The EA is a trending breakout model that loves volatility.

Right now I'm trading it on EUR/USD, GBP/USD, GBP/JPY, NZD/USD, AUD/USD, and EUR/JPY. What do you guys suggest for currencies I should be trading it on? Am I hurting myself?

there is only one meaning to correlation .. scientific definition

correlation is a fundamental topic in analyzing financial time series (charts) for statisticians. auto-correlation (correlation of a series with itself) can be used to compute the weights of a weighted moving average indicator.

You're correct, it's correlation of equity curves that one should worry about, not correlation of the markets being traded.

Ok. Well then answer this does question. Does correlation mean anything to you? Do you care about it between two currencies or between the signals youre receiving from your systems?

I'm wondering how concerned I should be about the correlation between the currencies I listed.

if you can find cross-correlation between two currencies that does not change with time, then you can confirm direction of one currency with another

Yes, I care, I only trade non-correlated strategies live.

I'm not sure what you mean, common sense and a little math should tell you the worst possible case, and then you can decide on your level of concern.

If trading any two markets that are highly correlated (either +1 or -1), effectively the risk is being doubled.

It has been advised to trade non-correlated instruments. All currencies vs the USD should rise if there is a fall in USD.

That could be said for a million uncorrelated markets too though. It is possible that uncorrelated systems will still produce losing results at the same time. I think my point is to question whether it is necessarily true that "correlated" currencies would produce losing results at the same time so much more often that it would be unwise.

For example, I was trading cable and eur/usd last week. Both have fairly similar charts, but they produced different trade sequences throughout the week. The most common difference would be one currency triggering a trade while another simply sat out of the market.

Im getting a feeling that this may just be another instance of it varying by system. But still, I don't know how to answer my own question. I'm only trading with 1% per trade, but I'd like to go to 2% for typical purposes. I just need some sort of confirmation that perhaps we are taking correlation too seriously.

For markets that have 0 correlation, it is possible that they may both/all produce losing trades, but highly improbable.

There are many factors, including the time frame, but most importantly, the system when talking about correlation.

Based on the charts above, on a day time frame, going long *any* of the majors (vs USD) on the last day would have produced losing trades. However, on a 60m chart, on 26/11/09 @ 04:00 GMT for EUR, O=C (a doji), but on GBP. O-C = 42 ticks.

The best I can say is that you will have to do the backtesting with *your* system and time frame to make a determination. Calculate the correlation coefficient (CC) for the time frame you're using, back test, and determine the CC between the returns of each.

Only then you will be able to answer your own question.

Anything else anyone says will be general or a rule of thumb.

Will also add:

1) For highly correlated instruments, what will you do if the system gives an entry signal for both instruments? This is where it is possible to multiply the risk if both signale are taken.

2) Again for highly correlated instruments (eg SP, ES, SPY), normally one would give a signal before the others - the 'leading' instrument. Knowing that all 3 are correlated, a person would normally only take a position in 1 instrument.

Over time, instruments will not maintain the same level of correlation. It is something to be aware of when a system gives similar signals for multiple instruments.

...the picture is again more complex for multiple strategies over multiple time-frames and instruments. Goggle 'epps effect'.

Hello,

I'm new to ForexFactory but not that new to forex. I've been trading demo for 2 years now and sick and tired of my 2$/h job here in Romania (at a bank) i've decidet to go live.

Anyway i have a question for you guys. I made a spreadsheet to calculate correlations and plot the correlation between two pairs for the last 50 hours on a graph. All i'm asking is why the correlation between EUR/USD and USD/JPYis so chaotic? I expected it to be negative all the time but it jumps from -1 to 1 all the time. I made the two pairs to start from the same point to better see the link between them.



Thank you!

I like correlation.