Brian J. Haag wrote that it's more than 12% more likely that on a given day the market is up rather than down.

For every 100 down days (SP500 since 1980) there were 112.2 up days, so the uncertainty component in predicting daily returns strongly favored the long side.

However another possibly more important aspect of prediction is whether it is easier to forecast bullish or bearish periods. If bearish patterns were easier to find than bullish, this might out-weigh the 112/100 odds for longs. But in the real market panic sell-offs (of various incarnations) test consistently bullish subsequently, whereas there do not seem to be predictive antecedents to big declines. If this remains true, then there will still be profits as long as one is brave enough to buy earlier than everyone else (who also know that panic is bullish) and endure the pain of intermediate-term losses, and risk that they continue deeper.

In light of recent market behaviour these arguments are easier to make (and take) than (say) 2002. It would be a lot more fun to invest if up and down days were homogeneously dispersed. But the problem is that upward drift comes in streaks (like recently), is taken away in lumps, and human nature is hard to ignore. How many traders, in the face of increasing losses or even going under, got more and more bullish in 00-03?

"Gee, hun, sorry we lost the house and owe all that money to your dad's homebuilding company. But the good news is that there is high probability it will go up from here!"

Vincent Andres writes: 

I would posit (to avoid one-line questioning) that:

  1. There are fewer bearish situations than bullish ones and hence, mechanically, fewer predictive patterns to find, so the task is harder.
  2. Bearish situations are sharper than bullish ones and hence perhaps the predicting patterns are also sharper, (kind of homeopathic reasoning), i.e., easier to detect (perhaps more contrast). I try a parallel: I see a snake, and I quickly get my hand away. The move is sharp, and so is the pattern/snake.
  3. Bearish patterns maybe come with short or very short notice (think of the snake threat), while bullish patterns may be seen with enough notice. So bullish patterning might be bit easier.

Kim Zussman responds:

My (not well stated) hypothesis is that, due to the immutable nature of human fear/panic reaction, people are generally over-reactionary and bullish for the future. However there do not seem to be similarly predictive preconditions of bearish futures; prior neither to sharp declines nor gradual ones. Or alternatively, there are such preconditions but they are harder to find

2nd hypothesis: Even though the probability of up and down days is close to the same in historic totals, the way days aggregate in actual ordered series is non-random to an extent not explained by actual 0.52/0.48.

I am thinking about other ways to test this (the pairs study was a first step) such as comparing runs in actual series to randomly shuffled days from actual series. As always suggestions and amplifications are encouraged. 





Speak your mind


Resources & Links