Friday, 24 August 2012

Bearish Downturn in COT Data Accelerates

Now it's the turn of the metals to confirm the broad downturn in trader positioning in the weekly Commitments of Traders data from the CFTC. Friday afternoon's COT report had more bad news for bulls.

It pushed my setup for copper out of cash into the bearish column, as you'll see on my latest signals table. It also ended my silver setup's long-time bullish signal and put it in cash. That setup would have gone short if the commercial hedgers had gotten just a little more bearish in their net position as a percentage of the total open interest.

In gold, the wrong-way small traders got heavily bullish - another bearish development. But their signal has a two-week trade delay, so gold remains bullish for the time being. 

And natural gas is also seeing heavily bearish positioning, though it remains in cash next week.

My S&P 500 and Nasdaq-100 setups remain bearish as before, while my 30-year Treasury bond signal is bullish again next week (meaning the yield would fall - which is generally bearish for equities). Have a good weekend, and good luck next week.


abe martin said...

Another correct call Alex. The ten year was oversold on the 14 day RSI and the COT positioning data confirmed that buying was occuring.

If you don't mind me asking Alex, from the research I have done on the COT positioning data, I noticed that all the positions lagged prices by a week (due to the lag in reporting the positioning by the CFTC). I am truly surprised how you were able to use a lagged variable into providing more information than many would have supposed. Not all the markets within the COT data have mean reverting positioning data, ie we do get trending in net shorts and net longs for extended periods of time, eg look at gold. The next logical chain is for me to presume, you might be using the data for short term mean reversion signals, as you would with any price oscillators. However I would surmise that short term reversion to mean is different for each market and the parameters have been optimized to each asset class.

Still puzzled.

Alex Roslin said...

Hi Abe,

Thanks for your message and thoughts. Some signals correct, others not - it's a long-term thing, so I can't boast or feel badly about any short-term run of good or bad signals.

But yes, the system I'm using is mean-reverting, and I think that is generally a failing of oscillators in technical price-based systems. In discretionary trading, I only act when I see a trend change in the price data, so I end up ignoring oscillators.

I think the alternative system for the COT data could be based on identifying not when the extreme position is hit - but rather when the trader positioning unwinds from an extreme position - i.e. when it, first, crosses a signal line, then second, re-crosses this same signal line in the other direction.

But that's hard to program.

Yes, short-term reversion is different for each market, according to my experience, and that is why I optimized the parameter values for each market. (Note that the parameters themselves are the same for each market; rather, the parameter values are what changes.)

Please feel free to make any other observations or to ask about anything else that's unclear. I am curious about your research.

Best regards,