johncoxme
12 posts
msg #121859
- Ignore johncoxme |
10/30/2014 10:13:22 PM
Wondering if there is a filter that finds stocks that go up 2% (success) BEFORE going down 2% (failure) on a weekly basis (from Monday’s opening bell for backtesting simplicity) for the past 52 weeks.
You’d never take a loss of more than 2% in the week.
Say a stock has an abnormally high success:failure ratio. Does that imply the stock intrinsically trades that way, or are stocks like thousands of people each tossing 52 coins and getting a few people (stocks) with a ridiculous number of heads?
Let’s say 7000 people (roughly the number of stocks) each toss 52 coins. The probability of getting 39 or more heads is about .000205 (binomial distribution, n=52, p=.5). Multiply that by 7000 and you’d get about 1 person (stock) which such a large number of heads (successes>39).
However, a stock’s performance is unlike a coin since trials (weekly performance) are not independent.
I would say if you run this filter and you get stocks with a success count in the high 40s (or > 75% success rate) then you get something cool to work with. I’d take those odds any day.
|