Forecasting stuff
suggested by Chris Y

It seems that forecasting stock prices and, in particular, crossovers with moving averages ... this seems a common practice.

>And you didn't know that, right?
Well, not the part about crossovers. I thought if you forecast the price then ...

>Then what else would you want?
Yes. However, it's an interesting problem so I surfed the Net and ...
>As your morning exercise?
Pay attention!
The thing I didn't like is that one looks at the historical data, fits some curve to it, then uses this curve to extrapolate.
Having done this, the forecast is fixed. There's no randomness. It's uniquely determined by the data over the past umpteen days.
For example, you might fit a straight line, like y = ax + b or maybe exponential growth like y = a bx or some other curve.
Excel even has functions to perform this forecasting: FORECAST(x,known_y's,known_x's)   and   GROWTH(known_y's,known_x's,new_x's,const)

>Don't they say the past is no indication of the future?
Something like that. Anyway, here's what I propose:

  1. We look at daily stock returns over the past 40 days.
  2. We compute the Mean and Standard Deviation of these 40 returns.
  3. We assume a Normal distribution of returns with this Mean and Standard Deviation.
  4. We select, at random, daily returns from this distribution for each of the next 20 days.
Of course, having generated a possible 20-day future, we can generate moving averages and Bollinger Bands.

>And I assume you have a spreadsheet to do this?
I'm working on it. You type in some stock symbol (like GE) and click a button and it downloads Yahoo data.
There's another button. Each time you click, you get some 20-day future ... something like this.

>What are all those coloured lines?
The predicted stock price, upper and lower Bolli-bands and 20-day and 100-day moving averages.

>But one of them ain't hardly moving.
The faint grey? That's the 100-day moving average

>What if I want a 30-day moving average or maybe ...
Okay, you can play with the spreadsheet.

Just RIGHT-click here and Save Target (or just try a regular click on the picture above).

suggested by Elliott P.

While it's interesting to see possible 20-day futures with each press of a button, there are those who tire of pressing a button and would like to see some Expected future price. So (as suggested by Elliott) the spreadsheet has been modified so that you can pick a number of "predictions" (like 1000) and click a button and the spreadsheet will run through 1000 20-day predictions and indicate what you might expect, 20 days into the future.

After completing 1000 predictions, the spreadsheet will say that (for example), the Expected change in price (from the last closing price) is $1.23 meaning a $1.23 increase, in 20 days.

>Or it might say -$1.23?
Uh ... yes.

>It'll be 100% accurate, I presume.
Would I do it any other way?