Analytical Wizard Required, page-19

It's reasonably easy to get historical Dow (or whatever) prices from Yahoo:

http://table.finance.yahoo.com/k?s=^dji&g=d

I computed that since 1990, ignoring break-even days, the Dow had 1604 up days, and 1444 drops.

By my computation Up run distribution was:

1 248
2 141
3 76
4 38
5 15
6 11
7 5
8 0
9 1
10 0
11 1

Downs were:

1 259
2 174
3 73
4 23
5 12
6 5

Eyeballing is enough to conclude this is not random. e.g. after 6 down days, the next one was always up. And after 2 down days the next day has significantly more than its fair share of ups.

A different story for up days which managed far more longer runs.

Hence Dow closes are demonstrably reasonably predictable - i.e. not random.

Can those of you claiming otherwise provide evidence to support your assertions?


A more useful exercise I've been conducting privately is to smooth out share movements, approximating them by swings of )say) 25% ups (balanced by 20% downs). e.g. only registering an upswing when a 25% gain is reached or surpassed. I found that resultant movements of individual shares were far from random and thus debunked the lame hypotheses that helped Black and Scholes win a Nobel Prize for their now totally discredited method.