Monday, 18 February 2008

Loonie Traders At Bullish Extremes

The Canadian dollar is bullish. That's at least according to my newly revised trading setup for the loonie based on trading positions as reported in weekly data by the U.S. Commodity Futures Trading Commission. My setup works by trading on the same side as the small traders in Canadian dollar futures and options when they hit specific points of positioning that has led to profitable, statistically robust returns. See my Latest Signals page table for all the details.

The small traders have steadily built back up their net long position as a percentage of the total interest since it hit an interim bottom in early November. In fact, the smart money in this market hasn't been this bullish in absolute terms since Aug. 2006. The setup works with a seven-week trade delay, so that previous bullish episode would have translated into a buy around Oct. 2006. That would have been a little premature, as the Canadian dollar's latest incredible ascent started in Feb. 2007. That's an example of why it's usually not that helpful to look at net trader positioning without the help of some kind of filter. We want to compare the current position to recent norms that have proved helpful in identifying good historic trading opportunities.

As it happens, when compared to recent trader positioning, my loonie setup has hit a position more bullish than at any time since Feb. 2007. Accounting for the seven-week trade delay, that earlier positioning would have given a buy in early April 2007. And that's exactly when the loonie exploded upwards from 86 cents to $1.10 by last November. More recently, the setup got a two-week bearish signal starting with the Oct. 30 COTs report, which would have profited from part of the Canadian dollar's recent plunge. The current bullish signal, from the Nov. 13 COTs report, took effect as of the open Jan. 7 and is now down 75 basis points.


Anonymous said...

very interesting blog and signals.
however, your observations regarding the Can$ do not make much sense.
If you simply look at the continous chart of the C$/US$ future and look at what the commercials did, you can see within a blink that the Can$ has always in the past 7 years risen substantially after the commercials had accumulated significant long positions. as in most currency pairs, a high short position by them was not as significant and was not always a good short entry. That said, the commercials are net short, though not extreme so it doesn't make much sense to me to expect a substantial move upward.
a question regarding the time-lags for your setups: isn't it dangerous to work on these lags as there seems no rationale behind these varying lags? in other words, might it just be by accident and therefore rather a curve-fitting variable that you enter into the equation?

Anonymous said...

Warren Buffett agrees:

Alex Roslin said...

Thanks for your comment. Regarding the commercial traders, interestingly, my testing didn't find any setups with even a minimal profitability that traded both the long and short side for the Canadian dollar following the commercials. Strange, but true. And that's from looking at hundreds of thousands of possible combinations.

You may be right that some setup just following the long side would work with the commercials, but if I can find one that works on both the long and short sides, I prefer to use that one. This is especially important for me because, as a Canadian, I might need to hedge my U.S. dollar holdings from time to time.

Secondly, you raise an important issue - curve-fitting. The number of parameters I'm using in this system is constant from market to market - moving average, standard deviation and trade delay. More specifically, I buy and sell when the net position hits a certain number of standard deviations from the moving average, offset by zero to eight weeks for acting on the signal. That's not any more variables than lots of trading systems out there.

See, for example, Curtis Faith's excellent book Way of the Turtle for some examples of how other mechanical systems are designed.

Overfitting is when a system becomes too complex. Three variables isn't too complex. I'd agree with you, however, if I were to be adding and dropping variables for the sake of getting a better historic result. But that's not how my system works.

As well, apart from the various profitability measures I check for when looking for the best setups, I also test their Sharpe ratio, MAR score and confidence intervals for both profitability and market-beating performance. This is all to see if the past results were just a fluke. These tests all help me discard many highly profitable setups that probably wouldn't be robust going forward.

Regarding the rationale behind the trade delay, the idea is to reflect the way traders accumulate positions in various markets. I've noticed that traders often accumulate positions in a market steadily for some time before their sentiment changes and prices follow suit, changing direction. Prices also may not follow suit right away, as the trader positioning takes its time to impact the markets.