Running the Numbers is Easy

[william the wie]

The 100 year trendline in stock prices is known. And daily data goes back 30 years or more giving 30,000+ data points.(EODDATA offered that much last time I checked.)

92% of all price moves end up within two SD of the trendline instead of the 95% predicted by statisticians. This leads me to a question.

Why is this data treated as a 97% correlation: 92/95; instead of a 62.5% correlation 5/8? The 5% predicted outliers have 1500 data points and the 3% unpredicted outliers have 900 data points. Predictions based on 200 data points are considered pretty solid so 1500, 900, much less 2400 data points are extremely solid.

Looked at one way Random Walk appears highly rational looked at from the opposite direction it looks like a bad acid trip. Comments?

 

[william the wie]

Toddsterpatriot, Zander, Toro,

 

[Toro]

I honestly don't know.

 

[william the wie]

I honestly don't know.

I definitely sympathize. Customers, lenders and employees definitely complicate your strategies much more than mine. Some of my rules which I doubt you would be able to follow with other people's money include:

Always have a positive carry hedge, I decided on that when I realized I didn't have a hedge against my Berkshire position.

Have a balanced portfolio of low beta positions and write covered options against them to fund my hedges against the S&P.

Unless there is a profitable reason to do something in my account I ignore it. Unless there is something in the news that affects me I take a look about once or twice a week at my account to see if anything has slid under the radar, if not I don't worry about it.

I doubt seriously that such an approach would work with your job but it is more profitable for me than racking up transactions costs. If the market makes a huge move every 6-8 years I should do well.

 

[jwoodie]

Confidence level?

 

[william the wie]

Confidence level?

Believe it or don't something like 99.9% that the results will be somewhat better than owning a single zero roulette wheel even without the ability to refuse bets and setting up sequential bets so that one outlier pays off multiple times.

But first let me correct math errors on my part and that of others:

200x30= 6,000 not 30,000 my bad.

2 SD+ at either end of the curve is treated as 2%. When combined they are treated as 5%.

The frequency of 2 SD+ drops below trendline is @ 6.25%. Even rounding down to 6% that equals more than 360 data points which is more than the 200 data points used in multiple regression formulas to test a proposition.

The lowest cost straddle in a set of options is supposed to identify the expected price at expiration +/- 2SD and usually overshoots because of the fixed strike prices but does not try to account for shifts prior to the strike price. That creates a much higher than expected rate of return.

 

[william the wie]

The number I use for no actionable move per time period is 0.93. Given that structural and financial leverage amplify moves and @1600 market days in eight years that is a very nice house PC being offered to me even though there is greater amplification downward than upwards. Consensus option prices indicate an expected 2% +/-0.5% per qtr market rise for the S&P and a probable 3.5 outliers per qtr to reduce losses it looks safe to me.