It turns out that bullish signal - which is to be executed on the open of Monday, July 20 - is far from being an all-clear for the market. The latest COT report shows the wrong-way small trader crowd putting on a huge net long derivatives position. So much so, in fact, that their signal in my S&P 500 trading setup has just flipped to bearish.
The small traders - that's the unfortunate little guys, who tend to get it wrong at key market turns and give all their money away to the commercial folks - bumped up their net long position as a percentage of the total open interest to 1.8 standard deviations above the moving average I use to gauge their sentiment. That's higher than any time since mid-Dec. 2008 - right before the big post-New Year bust.
Mind you, things aren't all that bad right now. The commercial traders are still big believers in the market - enough so to keep my S&P 500 setup from going all the way over to bearish. The setup will go back to cash after a week in the bullish column, as you can see from my newly updated latest signals table.
Other highlights:
- Financials: My trading setup for the BKX U.S. Bank Index isn't quite bullish, but the data has gotten nearly so. This last week, the large speculators - whom my setup for BKX trades alongside - significantly increased their total open interest, as you'll see on that latest signals table. Their signal has now gone to the bullish side. But this setup works with three different groups of traders, who all have to be aligned to get a trading call. And aligned in the same timeframe.
As it happens, all three signals are presently bullish. But the setup hasn't gone bullish because of a trade delay of two weeks for the signal based on the small trader total open interest. Two weeks ago, these guys were giving a bearish signal. If the other two signals can hold on to their bullishness for just a week longer, the overall setup will go long. But in these times, that's a big if.
- Natural Gas: After four weeks in cash, my setup for natural gas will go bullish on next Monday's open, July 13. The large specs have ratcheted up their total open interest big-time. My other signal in this setup - based on trading alongside the small trader total open interest - has been bullish four weeks. But thankfully, I don't just use a single signal for this or my other setups. My testing has found that often just hasn't been too reliable, statistically speaking. Natgas has gotten destroyed in the past couple of weeks, so I'm glad I booked a little profit and went to cash.
Good luck the rest of this week. See you back here Friday for more. And check out my portfolio page with an update on my single position's results.
7 comments:
Monsieur Roslin, i had hoped to find some appropriate YouTube, but je regrette.
http://ourgeorgiahistory.com/ogh/Roy_Riegels_earns_his_'Wrong_Way'_nickname
Although the SPX hasn't broken its H&S neckline OIH & XLE may be better leading indicators.
A USDollar bounce will put pressure on commodities & therefore the SPX will be weaker than the Nasdaq. OIH & XLE have already broken necklines on H&S patterns. I think the XLE is particularly vulnerable because its neckline is downward sloping.
Regards,
dave
Wow - good one. Poor Roy! Thanks, dave.
Finally! Small traders jumping on in.
Have you ever backtested NOT taking a delayed signal when you know it has to be reversed the following week?
Hi Adog,
Thanks for your question. It's an interesting idea, but I haven't tested that scenario.
One issue is that would add several new rules to the strategy. That generally reduces robustness and the maximum number of weeks you can use in the period over which the moving averages and standard deviations are calculated.
Regards,
Alex
Monsieur Roslin,
There was also http://www.centennialofflight.gov/essay/Explorers_Record_Setters_and_Daredevils/corrigan/EX16.htm
Btw, i suspect, & suspect only, that your 3-week delay avoids some noise, some chatter & gets you into the sweet spot of a move ?
Regards,
dave
Alex,
I don't see how this would affect the moving average and standard deviation calculations. If you get a BUY signal to be executed in 3 weeks, then you'd need to look "forward" (1-3 weeks; <= lag period) to confirm that it's not a sell. If it's in excel you could use offset(signalcell,lag,0). If the point is to avoid the noise, then wouldn't you want 3 weeks of confirming signals, otherwise i really don't get the logic behind lagging the signals.
Hi Agog,
The issue isn't related to programming this in Excel. That's easy to do. What I was saying is adding more rules to the trading strategy of the kind you're talking about affects the maximum periods that can be used for the moving average and standard deviation is because of a standard rule for trading system development: Testing should be limited to setups that do not use more than 10 percent of the available degrees of freedom in the dataset (i.e., the weeks of historic data available for backtesting).
Increasing the number of trading rules uses more of the available degrees of freedom. With my current system combining two groups of traders, I can't use a moving average/standard deviation greater than about 40 weeks for most of the markets in the COT reports, which usually go back to 1995.
More trading rules also increase risks of curve-fitting and reduce the confidence intervals for profitability and beating the market. It's very easy to find setups that produce great theoretical results in backtesting. It's much harder to find ones that even remote reliability for future trading.
If you want to read more on the issues around degrees of freedom, check out Robert Pardo's classic books, The Evaluation and Optimization of Trading Strategies and Design, Testing and Optimization of Trading Systems.
As for why the trade delays work, I can only surmise that it's because of the time it takes for extreme market positioning to work its way into prices. It's like looking at market sentiment; prices get overbought for a while, then they tend to revert back to more normal ranges, but there's a lag.
In my testing, I look at delays from zero out to eight weeks for every group of traders, and the ones I settle on just happen to score highest by all my measures of robustness.
Regards,
Alex
Post a Comment