Tuesday, December 12, 2006

Perceptual Distortions In The Stock Market

One of the signs that a field of study has become a science is the recognition of the need for controlled observation with tools that transcend the limits of the human senses. From the electron microscope to the Hubble telescope, advances in science have followed the acceptance that the unaided human eye cannot perceive all of nature's patterns.

Behavioral finance researchers teach us that unaided human senses are particularly ill-equipped to accurately assess the market's patterns. Our tendencies to overweight recent events and highly salient events; our overreliance on simple heuristics; our emotional weightings of losses vs. gains: these create perceptual biases that skew our responses to risk and reward under conditions of uncertainty.

Still, surprisingly, we find traders evaluating markets based upon nothing more than the unaided eye. Inevitably we look at a chart, notice a slope, and declare a "trend" in place. Can such visually-based analysis succeed?

I decided to try a little experiment. I went back to 2004 (N = 734 trading days) in the S&P 500 Index (SPY) and identified all occasions in which the market was up on a one-day basis, a four-day basis, and a nine-day basis (N = 234). We might call that an "uptrending" market. Four days later, SPY was down by an average -.07% (115 up, 119 down). That is weaker than the average four-day gain for the remainder of the sample of .22% (299 up, 201 down). In other words, the market that--to the unaided observer--has been consistantly up over the last two weeks has had no bullish edge whatsoever. Indeed, it has tended to underperform its recent norms. This fits with my recent research showing underperformance when we trade above a moving-average benchmark.

But let's take the experiment a step further. Suppose we isolate those occasions in which the market is up more on a nine-day basis than on a four-day basis and up more on a four-day basis than on a one-day basis. This would represent, to the unaided eye, a nice steady uptrend. We have 115 such occurrences in our sample. Four days later, SPY is down by an average -.11% (53 up, 62 down). When the market has been down on a one, four, and nine-day basis, but not in a visual uptrend, the next four days in SPY have averaged a loss of -.02% (62 up, 57 down). In other words, we've seen the worst short-term market outcomes when markets have looked their best. The more consistent the visual uptrend, the worse the near-term returns.

In my next post, I will investigate our perception of markets moving downward.

The perceptual distortion in the market is that the unaided human eye tends to extrapolate straight lines into the future. Like a ball tossed into the air, we expect markets to continue moving in their most recent direction. Statistical analysis is a tool for transcending the unaided eye. It, among other things, is what separates astrologers from astronomers, alchemists from chemists. Market psychology begins with the recognition of the limits--and limitations--of human perception.


Mike_Trader said...

Dr. Brett,
Excellent observation! I think more work needs to be done in this area. I conducted an experiment on my own to figure out why my breakout setups were not making money. I could easily see on the chart all the good breakouts from sideways consolidation patterns. All I had to do was put in a Stop order under the support (or above resistance) and let it ride...right? For my experiment, I went back in time picking days at random and started at the opening bar. I then advanced my charts one bar at a time to simulate the trading day and recorded all buy and sell decisions based on the close of each bar. What I found was that while the good breakouts always catch your eye, I was missing all the failed breakouts! These were the 1 and 2 tick breaks of support (or so I thought) that never followed thru and melted into the 'sideways consolidation'. I did capture some real breakouts but I invariably covered too early with all the failures still in my head. As a result, most days showed losses which mirrored what was happening in my realtime trading! Moral of the story is: Seeing is believing but are you seeing everything?


yinTrader said...

Hi Brett

I believe there is a tendency for novice traders to look at the trading Time Frame only, without checking back on the higher time frames to determine if the trend is still good.

It is generally true that if the higher Time Frame is up or down, the lower Time Frame that one trades in will likely follow the same direction of the higher Time Frame, in good time.

Hope this makes sense of what I am trying to say.

Brett Steenbarger, Ph.D. said...

Hi Mike,

Great example with the breakout setups; thanks for the note. Your point about seeing and believing is very well put!


Brett Steenbarger, Ph.D. said...

Hi Yin,

I think the point my article is making is that when the higher time frame also is "trending", near term outcomes are even worse. Many commonly held assumptions about moving averages, trends, and the like simply do not pan out when put to the test, especially during the recent bull market.


Brandon Wilhite said...

Dr. Brett,

Great post that points out a common flaw/problem...a comment and a question. Comment---obviously there is information that is not on the chart which should be taken into account when trading (i.e. fundamentals, volume, 'tone' of intraday action, and I'm sure more). This kind of testing is eye-opening to do, and I've done it many times myself. I can't tell you how many times I've thought I found an edge, but then the hard data didn't pan out. Question---do you think the results also have something to do with the mean-reverting behavior of the markets in the last few years? In other words, the regime has changed?


Brett Steenbarger, Ph.D. said...

Hi Brandon,

I think you're right: this is a regime that is different from what we saw during the late 1990s, when momentum trading made money. The key is to identify the regime we're in--and to identify when it is shifting!