Friday, 2 May 2008

Sells for Bullion, Banks and Russell 2000

Okay, that's just nuts. I think this afternoon's Commitments of Traders data on positioning in the major markets has set a record for most new signals and wildly swinging holdings. The data is saying the market is steaming into some kind of Drake Passage of turbulence. Check out the table linked on my Latest Signals page for all the gory details. Some highlights:

- My trading setups for gold, the HUI Gold Bugs Index/XGD Canadian Gold iShares and silver have all flipped to bearish. The gold setup uses a two-week trade delay, meaning execution is for the open of trading May 19, while the other setups have no trade delay, which means execution is for Monday's open.

- My NASDAQ 100 trading setup has flipped to bullish after the commercial traders increased their net long futures and options position.

- As well, my trading setup for the Dow Jones Industrial Average, which has been bearish since late January, has flipped to bullish. The DJIA setup fades the "dumb money" large speculators, who have suddenly gotten quite nervous and reduced their net long position to a specific point that in the past has signaled a good trading opportunity. From our contrarian point of view, their bearishness actually tends to be a bullish development.

- However, my Russell 2000 setup feels quite differently about things these days. This setup is based on trading on the same side as the commercial traders, and it just flipped to bearish. As well, my setup for the BKX Bank Index, based on the three-month Eurodollars contract, has flipped to bearish for execution on Monday's open of trading.

- What to do about all these contradictory signals? I've thought about this and come up with some new rules I'm going to try out for this kind of situation. It often happens that data from two closely related markets contradicts each other, and I have a hard time stomaching putting on long and short trades at the same time in such cases. What I did was study the correlations of various markets (I'll post the table soon) and out of that emerged this rule: trade signals only when a majority of the setups in highly correlated markets are positioned in the same direction, whether long or short.

So in this case, I found the NASDAQ 100 isn't highly correlated with any of the other equity markets. This means I'm going to act on that signal regardless of what the other markets do. However, DJIA, RUT and BKX are all highly correlated (which I defined as over 85 percent correlation). Of those three markets, two will be bearish as of Monday's open of trading (BKX and RUT), while one will be bullish (DJIA). So this means I short BKX and RUT and ignore the new DJIA signal. Incidentally, the S&P 500 is only highly correlated with the DJIA under my definition. Under my new rule, I shouldn't have a trade on for that setup or DJIA because the opposing signals simply cancel each other out.

The other major group of highly correlated markets consists of gold, silver, platinum, copper and crude oil. As of Monday, two of these five will be bearish, so the majority will still be bullish. That means I'll maintain my long copper and crude positions and avoid going short gold or silver.

I'll test out this new system for a while and see if it needs refinements. Your input of course is welcome. I also recognize I still need to do some actual number-crunching to see the impacts of this rule on the statistical robustness of the setups. Have a good weekend.


Anonymous said...

Is it just me or is your setups getting more arcane?

Anonymous said...

"I've thought about this and come up with some new rules..."

It difficult to see how your performance numbers could be meaningful considering that you seem to be constantly changing the system.

Anonymous said...


Interesting results from your timer.

To me, if you look at the shift in aggregate it smells of "rotation". Going from safe haven metals into riskier asset classes that do well in early stage recoveries. But banks typically do well in recoveries so I am surprised by that result.

This does not mean there will actually be a recovery, but at least the traders are anticipating it.

The key question I have for you is: how would the timer performed in May-July of 2000 when we had the Nasdaq recovery headfake. I imagine you would see a litany of conflicting signals in a very volatile time period.

Quite interesting to see that the Nikkei (given the time lag) appears to be anticipating this?


Alex Roslin said...

Hi to the first two anonymous comment writers: Are the setups getting more arcane? The setups themselves haven't changed. I'm thinking about new rules to manage them better on a portfolio basis. Other such rules are my existing risk-management rule and portfolio allocation rule.

As to the second comment, the system has remained pretty much the same since I started it more than a year ago. The setups work in exactly the same way. The specific parameter values have been improved a lot. I'd think it would be a good thing to have setups that performed more reliably and profitably in testing. I've learned a tremendous amount in the past year and still have a lot of plans to continue refining what I'm doing. In fact, I think the trader who doesn't continually learn more from their experience is the one who won't survive long in this game.

As for the past performance numbers, I update those on the Latest Signals page table any time I change a setup.


SilverDragon said...

Hi, Alex.

I know you've improved the robustness of several of your signals, but frankly your system looked more appealing when it had fewer trades. I'm a _very_ small trader, and trade friction is significant for me. Commissions would consume much of the gains for me on your setups that trade roughly monthly now.

That said, I really don't have a problem with shorting banks and the Russell 2000 while holding the DJIA and the SP500. That looks to me like overall, large caps look good to traders, but financials and small caps don't. And as the commenter above pointed out, the rotation factor explains much regarding the metals.

Thanks for your work in publishing your results!

Alex Roslin said...

Hi Silverdragon,

Thanks for your comment. I understand your dilemma. I don't have a huge account either, and position sizes are of course even smaller because of my risk-management controls. I should have noted that I consider my new rule on trades in highly correlated markets as just a temporary layer of risk control as I work on refining my setups.

As I've said in previous posts, it's important to follow each trade that a setup gives if you're trying to get anywhere near its past return from testing. As well, differing setups work on varying timeframes, so following several setups in a broader market simply improves overall portfolio diversity.

Once I'm done my revising process, I'm going to revisit the rule on highly correlated market trades and see if it's still necessary or could be replaced by some other portfolio-wide rule.

Best regards,