Correlation and Leading/Lagging?

Correlation via Wyckoff:

http://i22.photobucket.com/albums/b313/Hcour/FXDaily1.gif

The top 3 charts are daily ending 11/12, AUD-EUR-GBP vs USD, the bottom 3 are USD vs CAD-CHF-JPY; so naturally the top 3 are somewhat positively correlated to each other and the bottom 3 are somewhat positively correlated, and the top 3 are somewhat inversely correlated w/the bottom 3, w/the the JPY being the least correlated of all.

On 11/08, the top tier is in an uptrend, the bottom tier on a downtrend, w/the AUD briefly consolidating in the uptrend while the JPY consolidates to a much greater degree in its downtrend. So note, the 2 least correlated are consolidating as the others continue to trend.

On 11/09 on the top tier the AUD and GBP have relatively wide-spreads down closing near the lows, the EUR has a weak possible upthrust. On the bottom tier CHF has a down bar and the JPY has a wide-spread down bar closing on the low, breaking previous significant support in its consolidation (following an upthrust move). So there are 5 of the 6 charts positively correlated on this day, signifying weakness.

http://i22.photobucket.com/albums/b313/Hcour/AUD-USD1.gif

Now looking at the AUD/USD's current consolidation on the 4-hr chart. There is a Buying Climax of the petering-out variety, followed by a break of the up trendline to the Automatic Reaction, thus establishing the boundaries of the trading-range. At 1 a wide-spread breakout closing on the high that does not follow-thru and is wiped out at 2, that one bar in fact wiping out most of the previous struggling rally, back to the bottom of the range and a Shakeout at 3. It then takes 22 bars and some more to get to the next high at 4, another test that fails as the next two down bars to 5 wipe out pretty much everything that went before in a fell swoop.

Here is significant weakness. Price then goes sideways at resistance w/lots of upthrusts before breaking down.

Harold
 
http://i22.photobucket.com/albums/b313/Hcour/EURUSD3.gif

EUR/USD, daily in upper-left, hourly in upper-right, 4-hr (trading timeframe) on bottom.

EUR/USD has held during the recent weakness w/good relative strength, but I'm now looking for it to plunge and catch up (or down) a bit, as weakness is still evident on the other currencies. On the daily there is an apex forming, even more evident on the 4-hr.

On that 4-hr chart, note the real climax at A on the upthrusting bar, followed by the quick Automatic Reaction at B forming the boundaries of the trading-range. Then several upthrusts in a row at C on a Secondary Test. Here there is a sharp break of the uptrending support channel line at D on a wide-spread closing on the low, which becomes a shakeout of the range at E at former resistance become support. The rally to F comes up short and an apex now forms in the bottom-half of the trading-range.

Harold
 
Don't think about correlation.

It's a subject much debated amongst us FX traders.

And there is a definite reason that should put this on hold for all time:

When you consider the EURUSD and USDCHF for example...

When the EUR/USD goes up, then the USDCHF comes down. Statistically, yes, they are inversed.

BUT, there is a third pair called the EURCHF. So there aren't two variables, there are three. It's not a bi-dimensional system, it's now 3D. :)

It's really about the strength the three currencies have in relation to one another.

When EURUSD is rising, the Euro is stronger than the Dollar and if the USDCHF is dropping, it means that the Swiss Franc too is stronger than the US Dollar.

So the last is the Dollar and the first two are the Euro and Swiss Franc.

However, if the EURCHF would be rising very hard due to high demand, therefore making the CHF extremely weak and the Euro very strong, we have a deviation from correlation that would make the EURUSD rise, USDCHF rise and EURCHF rise.

I think the lack of correlation should be played always on the synthetic pair that automatically has a much bigger range. (the case with GBPJPY and GBPCHF).
 
I did a few studies on my own a while back. I want to present to you a way in which you can trade correlated pairs in FX.

I call a triangle, a group of three currencies that, taken separately, have all positive yield spreads.

Take the GBPUSD, GBPJPY and GBPUSD pair. All of these three pairs have positive yield spreads. Out of these three, the most lucrative is the GBPJPY which has the highest yield spread, meaning that when there is yield "value" to buy, this one will be the favorite.

Now, although we know the yield spread is positive and everyone should be buying, there are situations in which ALL three pairs are oversold, and that is an anomaly from a fundamental point of view.

Because of this, the most valuable (from a yield perspective) currency is the GBP and the most overvalued is the least attractive, meaning the yen.

Take a look at what happens to the GBPJPY when this anomally occurs.

I won't tell you the parameters of the 3 stochastics, but they are fairly large and rational.

6q0sl7c.png
 
I don't know if rcanfiel is watching this, but this is a fundamental situation that wouldn't have been revealed without the aid of TA.

Also when you enter a trade after this situation occurs, you enter and exit relying on nothing but TA.
 
Quote from hcour:
...As for detrending, I'm not familiar w/it. After reading your post I read some explanations of detrending but I still don't really understand what it means when it says "eliminating the trend". What exactly is being eliminated? I still see trends in the charts using detrended data. Forgive my ignorance, as I said at the beginning of this thread, I ain't real smart, 'specially when it comes to math stuff. I suppose this is a method of more accurately comparing 2 data sets?

I get my daily data from CSI/UA which I import into MetaStock. CSI has a detrending option and I do see a difference between it and the regular charts. The detrended chart does show a clearer divergence in the recent Feb/Mar/Apr lows of the CHF and EUR, w/the EUR clearly showing more strength/demand in the range. I shall investigate further. ... As for cycles, I've tried them but could not incorporate them into my strategy. ...
More...
This thread may be 5 years quiet, but it contains interesting ideas. Cycle analysis often benefits from detrending, but not intermarket analysis. Two sine ways lagged 90 degrees have zero correlation, but yet one exactly leads the other.
I am not a big fan of correlation between Forex pairs. It is too easy to confuse oneself with four, six, eight, ... economies in one play. If alex.samant, can keep it straight, more power to him.
What I wish to add to the discussion is that there are different correlations. Inputs can be Price, Log Price, Return (diff Log Price), and the calculation can be Linear Correlation or Rank Correlation.
I haven't spent much time looking at the short-term correlations described in this thread, but it hasn't already been covered elsewhere, it is probably worth exploring.
 
Steven.Davis said:
This thread may be 5 years quiet, but it contains interesting ideas. Cycle analysis often benefits from detrending, but not intermarket analysis. Two sine ways lagged 90 degrees have zero correlation, but yet one exactly leads the other.
I am not a big fan of correlation between Forex pairs. It is too easy to confuse oneself with four, six, eight, ... economies in one play. If alex.samant, can keep it straight, more power to him.
What I wish to add to the discussion is that there are different correlations. Inputs can be Price, Log Price, Return (diff Log Price), and the calculation can be Linear Correlation or Rank Correlation.
I haven't spent much time looking at the short-term correlations described in this thread, but it hasn't already been covered elsewhere, it is probably worth exploring.
More...

Given 2 identical sine waves with a 90 degree offset, correlation returns 0 because you haven't included a 90 degree offset in the time stamps or index values of the variables being compared. IMO sine waves are so different from price, they aren't a good model for exploring the principles behind successfully rewriting a correlation function to return coefficients for non linear, event based correlations.

Tho, just so this doesn't degrade to a discussion of the merits of studying sine waves, I agree sine waves are an excellent model for understanding the basis of other methods, for instance such as pitchforks, ect. I strongly encourange anyone who doesn't understand all the intricaices of sine waves to get busy studying them.
 
Last edited:
You must log in or register to reply here.
Back
Top