# Double Top or Breakout?

March 29, 2010

Caution: I am not prophesizing.My job is not to predict market. Its to trade it, and hence I am not reflecting any bias on any side.

Which one, do you think, it will be: Double Top or Breakout?

Advertisements

6 Comments
leave one →

Double top, double bottom….what difference does it make? Do they actually provide a statistical edge? π

a. My system doesnt trade classical chart patterns. π

b. I do.

c. Double Top, if you approach it via the classical long winded definition, then I have never tested it. But if you approach it via a support and resistance concept, then yes, it has a good testable statistical edge

“But if you approach it via a support and resistance concept”

How do you mean?

IMHO, the problem with estimating the efficiency of a pattern lies in quantifying the movement and time parameters, which is naturally dependent on the time frame considered. Unless one finds a way to standardize the computational methodology (freeing it from the movement-time dependency), or has an apriori expectation for each timeframe (which will most likely and eventually be doomed to failure in the back test), the parameters will require optimization for each time frame, making it vulnerable to data mining bias and type I error.

IMHO, the problem with estimating the efficiency of a pattern lies in quantifying the movement and time parametersNot necessarily. I mentioned, the presence of a”good testable statistical edge”. The edge means, if you take a trade on this idea,you have certain advantages.

which is naturally dependent on the time frame considered.Of course, but the time frames are self-similar, the relative proportions will just change moving into a finer or a grosser time frame.

Unless one finds a way to standardize the computational methodology (freeing it from the movement-time dependency), or has an apriori expectation for each timeframe (which will most likely and eventually be doomed to failure in the back test),Apriori Expectation of a time frame? What does that mean?

Apriori Expectation for each time frame? What does that mean?Eg:

In 1min data, movement-time rule for double bottom defined as:

–> The distance between the neckline and bottom (measured as a %) should not be less than 1%

–> The time interval between the two bottoms should be more than 60 bars and less than 200 bars.

–> The time interval between the top and each bottom should be more than 40% of the total time period and less than 60% of the total time period.

The apriori expectation for this 1min timeframe are the parameter values assumed – 1%, 60-200 bars and 40-60%.

Unless we introduce an assumption that market behavior is fractal in nature and consistent in every time frame, we cannot scale up and down the distance for different time frames as such, particularly not the distance figure.

Thats exactly what I am talking. If you approach double top or bottom by the long winded definition as espoused by investopedia.com or etc, then I havent tested it.

But, I dont approach double top as a classical chart pattern with some predefined params like 1% dip and 60-200 bars duration etc. I approach it as just, a manifestation of a resistance being present.

For example, the probability that a prior resistance is not broken is around 54%.Guess what! Its a good enough, entry edge for me. Coupled with that some of my own stats which represent the “goodness” of entry are better for such a decision method.

And about self similarity- I would like to go in favour of this.