Moving Average Crossings are useless for timing, at least in this case

I just spent $1000 for a tout that uses moving average crossings to time weekly option trading. It's not working so well!

So I thought I'd check it out.

Here's a test for SPY using daily closes for ~10 years (2517 days), 50 & 20 day moving averages, and looking at the price change 5 days after the crossing.

Bull Crossings Up = 20 , Bull Crossings Down = 6 , Fraction Correct = 0.7692

Bear Crossings Up = 15 , Bear Crossings Down = 10 , Fraction Correct = 0.4

Overall Fraction Correct = 0.5882, Fraction Up = 0.6863

Bet of $1 traded on crossovers becomes $1.45

Bet of $1 traded on Buy & Hold becomes $2.1043

So it was an overall Bull market. What would it have looked like if I just randomly picked 5 day intervals

Here are 5 runs:

Random date Up = 32 , Random date Down = 19 , Fraction Up = 0.6275

Random date Up = 30 , Random date Down = 21 , Fraction Up = 0.5882

Random date Up = 33 , Random date Down = 18 , Fraction Up = 0.6471

Random date Up = 32 , Random date Down = 19 , Fraction Up = 0.6275

Random date Up = 26 , Random date Down = 25 , Fraction Up = 0.5098

At the very least, we can say that this commonly used method does NOT improve trading results!

Okay skeptics of my skepticism - What is wrong with this test?

I should have run it on a different equity, commodity, forex, etc.?

I should have used different moving average lengths?

I should have added some additional selection criterion?

By what criteria would you make those decisions?

Submitted June 21, 2014 at 01:24PM by hsfrey http://ift.tt/1iX0Dl7