Saturday, August 11, 2007

How to Lose at Trading the Stock Market

"I need to see the market making new highs to confirm an uptrend," a trader recently told me with a knowing air. "I'm not going to sell until we get confirmation of a downtrend."

I haven't heard from that trader since the market began its high volatility decline. I shudder to think what would have happened had he consistently bought highs and sold lows.

The reality, however, is that the stock market has never rewarded the obvious. A belt-and-suspenders approach to trading has never produced superior returns. Give me a short-to-intermediate term moving average and I'll show you the returns that will refute our trader's strategy.

Today's little exercise draws upon data from Barchart.com, which has a nice "performance" feature that tracks the P/L of various stocks using various technical indicators for entries and exits. Here we'll be looking at the S&P 500 Index (SPY) and the Commodity Channel Index (CCI), an oscillator that captures market strength and weakness.

We'll follow our trader's strategy by buying SPY when the 40-day CCI hits +100 (shows strength) and exiting the position when it crosses below +100. Similarly, we'll sell SPY when the CCI hits -100 and exit the position when it crosses back above -100.

This strategy may produce a fair number of whipsaws, but should capture the big trends.

Over the past two years of trading, this strategy has produced 49 trades, with an average holding time of 8 days. We've had 11 winning trades (including one that is presently open) and 38 losers, about a 22% win rate. The total number of SPY points lost over the two years was a bit over 23 (or 230 S&P futures points), which roughly translates to a 16% loss of capital if we assume no leverage and equal position sizing for each trade.

In other words, during a distinct market uptrend, trading a strategy that buys strength and sells weakness has lost significant money.

But let's look further under the hood to understand *why* such a strategy fails.

If we break out the trades, we find 35 long trades and only 14 shorts. This reflects the upside bias in the market: we haven't often dipped below -100 on an intermediate-term oscillator.

Of the 35 long trades, we would have had 9 winners and 26 losers. Cumulatively, those would have lost us 6.45 SPY points. The winning trades would have brought in 14.40 points, or 1.60 SPY points per trade. The losing trades would have cost us 20.85 points, or .80 point per trade.

When we examine the short trades, we would have had only 2 winners and 12 losers. The winning trades would have made us 1.37 SPY points or about .69 point per trade. The losers would have cost us 18.22 points, or about 1.52 SPY points per trade.

In a limited way, the strategy worked; it did catch some large moves on the long side. As a result, the average size of the winning long trade was twice as large as the average size of the loser. Even so, however, the long trades would have cost us money. Why? Because losing trades outnumbered winners by about 3:1. Stated otherwise, even in a bull market, strong moves only followed through with further strength 25% of the time.

It's when we look at the short side that we see the trading system completely break down. The win rate on trades was about 15%. On top of that, the average size of losing trades was more than double the average size of the winners! Selling weakness over the last two years has been a complete and utter disaster.

Now for some disclaimers: Yes, there are other ways to trade the CCI; my purpose is not to dis this particular indicator. And, yes, results would be different if you added a variety of stops and money management elements to the mix. And, of course, results would have been different had we examined the strategy at other points in market history or if we had pre-selected a wonderfully trending market over these past two years.

My point is this: If we use SPY as a proxy for the stock market (as the most actively traded index ETF) and adopt a strategy of "buy confirmed strength, sell confirmed weakness", we would have lost money at an alarming pace. Even if we had employed it as a long-only strategy, it would have lost us money in a solid bull market! Trading the short side in this environment led to severe losses.

In fact, I would go so far as to say that the "buy strength, sell weakness" strategy is so bad that it's promising. Simply reversing (fading) the strategy in SPY and adding a time stop (to avoid the few long uptrends) would have made a trader a chunk of change.

Of course, nothing is 100%. The "sell weakness" part of the strategy had us going short SPY on 8/9/07 and, as of Friday's close, the trade was profitable. The system sold at 145.39; it remains an open trade. The previous short trade was on 7/26/07 and covered on 8/8/07 for a loss of 1.81 SPY points (about 18 ES futures points). That was a bearish period, but still the strategy managed to lose a decent sum. You have to marvel at the consistency of the market's ability to punish the obvious.

RELATED POSTS:

Why Short-Term Traders Lose Money

How to Lose Money Buying in an Uptrend

Reversal Effects and Fading the Herd

The Market is Rigged Against Human Nature
.

7 comments:

Bri said...

Buy low, sell high. All of the systems that I have backtested successfully seem to work on this principle, at least for indexes.

opw said...

I have no doubt that 'buying strength' does not work in the example above. But it should be also clear that buying strength is a very loose term and the example above is your definition of buying strength.

Besides, the indices have a bias of snapping back from extreme values.

Buying strenght in stocks is not neccesairily a potential disaster as my own account shows me.

Below is a link to a relatively well know study that shows buying strength can work.

http://turtletrader.com/blackstar-funds.html

Yes, there are some people that find this study flawed, but then, all backtesting is flawed...

Brett Steenbarger, Ph.D. said...

Hi Bri,

I think you're right: fading dominant sentiment does have an edge--

Brett

Brett Steenbarger, Ph.D. said...

Hi OPW,

Great point; thanks. I tried to clarify the issue of trading and fading strength in my subsequent post. I appreciate the thoughtful comments--

Brett

techfarmer said...

Anatomy of developing a trading system based on past data:

I'll use this example as a case study so anyone can comment and find any holes.

I've done some research regarding performance based on the day of the week:

http://techfarm.blogspot.com/2007/08/s-500-returns-by-day-of-week-1962-2007.html

I discovered that Monday is the only losing day, and Tuesday through Friday are positive days.

So based on this, one could come up with a trading system:
Buy the SPY on Tuesday morning and sell on Friday afternoon.

Aside from backtesting these results, can you see any problems or strengths or weaknesses in this sort of system?

Related question: How do we research/backtest results using free tools?

Thank you.

Brett Steenbarger, Ph.D. said...

Hi TechFarmer,

Thanks for the link and interesting idea re: a system using day of week returns. You'd want to include considerations of drawdowns into any effort to make this a trading system. I don't know of any free tools to really test a system well, though TradeStation is free for account holders.

Brett

info said...

Hi Doc:

Does your study then imply that the way to make money in trends is to enter on pullbacks? By definition this would afford a better risk/ return?


RONIN