Avoiding S-curve error
The future is bumpy. It comes in spurts, and then it pauses.
It's tempting to connect two dots and draw a line to figure out where the third dot is going to be.
In the long run, that's a smart way to go. For example, if we look at the cost per transistor in 1970 and again today, we can make a pretty smart guess about where it's going in the future.
But we won't get there in a straight line.
Consider this graph (from this must-read article):
If you connected the first two red dots (1885 and 1925), your prediction for dynamic range today would be have been way off, far too low.
If you connected the second two dots (1928 and 1933) again you'd be way off. Too high by far.
That's because science doesn't march, it leaps.
The S curve is flat, and then it's not. It's punctuated. A technical innovation changes the game, industry takes a development generation to incrementally pile on, then it happens again.
You can't multiply a one-year increase (in computers, your income, your height, the cost of a commodity) by a hundred and figure out what it's going to be in a hundred years, any more than a salesperson can multiply one day's commissions to figure out a year's pay.
Day trading is a risky business.