motivated by e-mail from Ron McEwan
Comparing Historical Prices: Part II ... a continuation of Part I
My brother-in-law and I had invested in a stock whose recent performance was pretty lousy, so I thought I'd try to predict when (if?) it'd recover and ...
>And you dragged out that comparison spreadsheet from Part I, right?
Yes ... except that it matched the past month with some historical month, but didn't give me a prediction.
Then I remembered a modified version that Ron sent me, so I modified the spreadsheet to include a price prediction for the next week or three.
My assumptions went like this:
The spreadsheet looks like this (and you click the picture to download):
- Whatever happened in the past was a result of investors reacting to the situation at that time.
- If the current gyrations in the stock price occurred in the past, then we'd expect investors would react in the immediate future as they did in the past.
- So we examine every 1-month period over the past umpteen years to identify that month which matches the current month.
- Then we stare in awe and amazement at the subsequent stock behaviour.
You can see that the month starting Jan 24, 2007 is the "best" match to the current month and, since investors are quite predictable, see how they reacted in the following month?
So we can now confidently predict that, although XOM closed today at $82.49, in ... let me count the days ... 18
days it'll be at $86.85 so we can get ready to sell.
Here are a few more predictions, comparing the month ending March 7, 2008 with some historical month ... then predicting the following month:
For another example, I looked at my lousy performing stock. It's now at $67.64
Based upon the "best" historical match I confidently predict that in ... let me count the dots.
Okay, here's my prediction: In 10 days it'll be at $73.13
... and I'll make a bundle.
>That's a lot of B.S. ... and you actually believe that stuff?
B.S.? Is that Black-Scholes?
We predicted a TSX of 13,818.84 by April 7, 2008, up 4.0%. It actually closed at 13,745, up 3.5%
We predicted a DOW of 12,033.31 by April 7, 2008, up 1.2%. It actually closed at 12,612, up 6.0%
We predicted MSFT at 27.50 by April 7, 2008, down 1.3%. It actually closed at 29.16, up 6.7%
Pretty lousy, I'd say.
Yeah, I guess. However, I was chatting on FWF about some other topic and
Shakes mentioned auto correlation.
That got me to thinking that, instead of minimizing the maximum error (between the last month and historical months),
maybe I should maximize the correlation between them.
That is, I start with a $1K portfolio and calculate the correlation between the subsequent portfolios and ...
>So you modified the spreadsheet above, right?
Yes, it now has that option.
For example, here's a prediction based upon that criterion:
>You're predicting a DOW at 14,009.96 by July 16, 2008?
Yes, but you must understand that ...
>And you have the temerity to give two decimal places?
I could give more if you'd like.
Comparing Historical Prices: Part III
The spreadsheets noted above compare daily movements, recent and historical.
If you'd like to see a longer time period, here's a spreadsheet that does months instead of days:
>Looks like the charts have "recent" DOW, as of Oct 20/08 ... right?
Right. It also says that the bottom will occur in 9 weeks.
>Huh? Nine weeks from Oct 20? Isn't that Dec 25?
It is, indeed ... and the bottom is a Christmas present.