Wednesday, September 5, 2007

Timing the Markets With COTs

Who says you can’t time trades using the Commitments of Traders Reports? This COTs-based mechanical system beat the NASDAQ by 728 percentage points—with only one trade needed per year.

By Alex Roslin
Technical Analysis of STOCKS & COMMODITIES
May 2007

Ever wonder what the smart money is doing in the markets? How do the folks with the best information and deepest pockets invest their wealth? A small handful of analysts and traders has found an interesting way to tell. It’s called the Commitments of Traders Report, and its devoted fans say the report is the closest thing in the public domain to a holy grail of market forecasting they’ve found.

The COTs, as they’re known in the business, don’t make headlines like those celebrities of the world of economic indicators, the Consumer Price Index (CPI) or the unemployment numbers. But a growing legion of “commitments analysts” is glued to computers each Friday at 3:30 p.m. (Eastern time) when the Commodity Futures Trading Commission releases the latest weekly COTs numbers.

The data, which used to be available only to paying subscribers but is now free, shows how many of the world’s largest commodity-producing firms, index funds, and hedge funds are positioning trillions of dollars of futures and options bets in more than 90 markets—everything from frozen pork bellies to the Standard & Poor’s 500, orange juice, the Canadian dollar, and unleaded gas.

But this treasure trove of insider information is hard to interpret and not obviously usable in its raw form. The data doesn’t appear to correlate neatly with subsequent prices in the cash markets. What’s more, analysts don’t agree on how to act on the data.

Should traders position themselves in the same direction as the commercial hedgers? Commitments analysts often call these commercial traders the “smart money” because they are presumed to have the best market information. Or should we fade the noncommercial traders—usually known as the “large speculators” or simply the “dumb money”—who are often said to be positioned the wrong way at major market turns?

Some analysts suggest the commercials and large specs are most useful to watch when they take extreme net long or short positions, which often suggests a market turn is imminent. But when is an extreme really an extreme? Varying systems of measuring the extremes produce widely varying buy and sell signals, many of which fail to make any money at all.

One popular approach is to trade with the commercials in the physical commodities like gold and crude oil while fading the large specs in the financial markets like S&P futures and Treasuries. But it’s not clear what data this approach is based on. Analysts who have devised systems to analyze the COTs won’t reveal their methodologies publicly because they are proprietary.

And even these analysts generally say their systems don’t give trading signals per se. Instead, they suggest the COTs are best used merely as an early-warning system, which must be coupled with technical analysis to time trades.

But a closer look at the data shows a trading system can indeed be devised for the COTs. The results also suggest that following the smart money isn’t always the smartest approach. We actually need to know what the dumb money is doing—the really dumb money.

Follow the Dumb Money

One group of traders often gets overlooked in discussions on the COTs—the small traders. These guys are the hedgers and speculators whose positions are too small to be included with the commercials or large specs. They are listed under the “nonreportable” category in the COTs reports.

The small traders are the truly dumb money no one seems to care much about. But they may just be too dumb to ignore. Fading the small traders when they take extreme futures and options positions appears to be the best strategy for trading the S&P 500.

A switching strategy of fading the small traders when their net percentage-of-open-interest position was two standard deviations or more from its 22-week moving average would have resulted in a 316% profit since 1995. (This is the year the CFTC first released combined futures and options data in electronic form for free.)

That result was over two times the S&P 500’s gain of 139% in the same period. In addition, fading the small traders would have required only a dozen trades—six long and six short—or about one per year, leaving lots of time to spend your profits!

The system was based on trading on the next weekly open after the signal was given. The maximum drawdown for this system was 16%. (The figures are valid as of the week of December 4, 2006, and don’t include slippage, dividends, or commissions.)

Meanwhile, fading the large speculators would have also beat the market, but not by nearly as much. The best combination appears to have been using a 208-week (four-year) moving average and fading the large specs when their net percentage position was 1.5 standard deviations or more away. That would have led to a 59% profit since June 1999, compared to an 8% gain in the S&P. (The best result came with delaying the trade until the weekly open two weeks after the signal was given.)

The large specs were clearly dumb enough to fade, but nowhere near as dumb as the small traders!

The commercial hedgers did better. They made a 230% profit since April 1998, compared to 24% for the S&P. That amounted to a 0.5% profit per week in the market—less than the 0.54% profit of the small traders. The best result for this group came from trading with the commercials when their net percentage position is one or more standard deviations from its 156-week moving average. It was most profitable to delay the trade until the weekly open three weeks after the signal was given.

NASDAQ Timing System

On the NASDAQ, the small traders didn’t do as well. Here, the large specs truly were the dumbest money around. The best result from fading them yielded a 725% profit since May 1999. During the same period, the NASDAQ lost 4%.
The large-spec gain worked out to 1.83% per week in the market, while requiring only eight trades (seven of them profitable), or slightly less than one per year. The maximum drawdown for this trade was 9%.

This trade was taken whenever the large spec net percentage position was one standard deviation or more from its 156-week moving average. The best result came from delaying the trade until the weekly open one week after a signal was given.

The next best result for the NASDAQ was a tossup between the small specs and commercials. The best return for the small specs was a 592% return since Jan. 1997 (or 1.15% per week), compared to 79% for the NASDAQ. This result came from fading the small traders when their net percentage position was 1.5 standard deviations or more from its 20-week moving average. The trade was made on the next open after a signal was given.

As for the commercials, their best result was 469% since May 1999 (or 1.18% per week), compared to a 5% NASDAQ loss. This was from trading with the commercials when their net percentage position was one or more standard deviations from its 156-week moving average and buying at the next open after a signal was given.

Conundrum Solved

The results shed fresh light on the COTs data. Clearly, each commodity has its own personality and requires a specially tailored system of optimized trading signals. Unfortunately, analysts often cite the ups and downs of the data without a good understanding of what it means.

For example, in late 2006, the COTs data for the S&P 500 was a mess of contradictory signals. Here was an important market juncture during which many analysts fretted about a possible correction to coincide with the often-turbulent fall season of the midterm year of the U.S. presidency. What was the COTs data saying at the time?

In August, the large specs seemed to be giving a colossal sell signal on the S&P 500 as they built a record net long futures and options position. The dumb money got more and more net long through the fall of 2006. But fading the large specs would have been disastrous as the S&P rallied to multi-year highs.
In fact, the most profitable large spec timing system for the S&P 500 has been on a sell signal ever since January 2005.

As for the commercial traders, the most profitable system on the S&P 500 has been on a sell since January 2006.

What to do? The answer lies with the small traders. The really dumb money gave a timely buy signal last August when it swung to a huge net short position. (As of early March, the small traders were still on a buy.)

The data has been clearer of late for the NASDAQ composite index. In late November 2006, the large specs gave a sell signal. The small traders followed suit with a sell in December, followed by a sell from the commercials in January (after the data in the accompanying chart was compiled).

But fading the large specs has given such clearly superior returns on the NASDAQ that they’re the only traders we need to follow for this index. Case closed!

Alex Roslin is an investigative journalist and active trader in Montreal. He can be reached at roslin@videotron.ca. Watch for his COTs analysis at http://www.cotstimer.blogspot.com/. Thanks to Mike Gordon for his invaluable advice on Excel.

SIDEBAR
How to Build Your Own COTs Database

Want to crunch your own COTs data? It’s simple—and free. The hardest part is learning a bit about Microsoft Excel.

The COTs data can be imported into Microsoft Excel from the website of the Commodity Futures Trading Commission and easily updated each week. (The site has easy-to-use downloading instructions.)

Break out all the historical data for each commodity into separate worksheets. You can calculate the net position of a group of traders by subtracting its short position from its long position.

I found the highest returns consistently come from using the net percentage-of-open-interest position for each group of traders, rather than the absolute number of contracts. The percentage positions are found in columns AW through BE.

Calculate the moving average and standard deviation by using Excel’s AVERAGE and STDEV functions. In my systems, a buy or sell is signaled when the net position equals or exceeds the moving average plus or minus the standard deviation.

One time-saving trick for extending a calculation in one cell to an entire column: Click the first cell and, while holding down shift, click the last cell in your column, then press control-D.

To turn your data into a chart, click “Insert” in the upper toolbar and go to “Charts.”

SIDEBAR
Save the COT!

The Commitments of Traders reports are safe. Thanks to an overwhelming public outburst of support, the Commodity Futures Trading Commission says the 44-year-old reports won’t be discontinued. The commission announced in December 2006 it would keep publishing the data unchanged and start a new weekly report for some agricultural commodities to reflect the rise of index trading.

“The reports will continue,” says John Fenton, the commission’s director of market surveillance.

The commission sparked an outcry last June when it issued a call for comments on the reports, suggesting the data may be skewed by changing patterns in trading activity in some commodities. The notice asked for public comment on whether the reports provide a “public benefit” and should be discontinued. It said it was under no legal or regulatory obligation to keep publishing the data.

An unprecedented 4,659 responses came from 23 countries—by far the largest number the commission has received in response to such a notice in its 31-year history. (The previous record was 1,062 comments.)

The response was unanimous. Not one respondent said the reports should go. “Please, please do not discontinue this very valuable report,” one said. “Don’t you dare,” said another. “Leave it alone you knuckleheads,” the CFTC was advised. “Save the COT.”

Instead of dropping the reports, the commission decided to break out a new “index trader” category for 12 commodities, drawn from the current commercial and noncommercial groups. The new report will be tried as a pilot project for two years. The CFTC will then decide whether to expand it to other physical commodities.

Fenton acknowledges the wording of the call for comments created an unintentional controversy and promises that the COTs are here to stay.

“I would say it’s a permanent decision—as permanent as things are,” he says. “We didn’t mean to imply that we were really planning that. We didn’t consider discontinuing the reports was likely. We were trying to consider all options.”

SIDEBAR
Best Combos

The most profitable systems I found for trading the COTs on the S&P 500 and NASDAQ are listed in sidebar Figures 1 and 2 for each category of traders—commercials, large specs and small traders.

The three variables for each system are: (1) the number of weeks used to calculate the moving average and standard deviation; (2) the number of standard deviations that the net percentage-of-open-interest position must equal or exceed above or below the moving average in order to get a buy or sell signal; and (3) the timing of the trade following a signal (whether on the first open the week after the release of the COTs report, or up to three Mondays later).

We trade on the same side as the commercials and fade the small traders and large specs.

I calculate the size of my position based on the maximum drawdown for the trade and my risk threshold of 2% of total assets.

The data in sidebar Figures 1 and 2 is accurate as of early December 2006. And it’s important to remember that past results are no guarantee of future earnings.

[Accompanying this story were two tables showing the results for the best setups for the S&P 500 and NASDAQ for each of the three groups of traders.]

(c) Copyright Technical Analysis of Stocks & Commodities

POST-SCRIPT TO THIS STORY: Since this story was published, I've made some major refinements that you'll see on my blog: www.cotstimer.blogspot.com. (Click the "Latest Signals & Results" link in the navigation bar for specifics on my new S&P 500 and NASDAQ signals.)

No comments: