Dec

8

Some Testing:

I calculated from 20090101 until today the maximum 30 minute rolling standard deviation in the SPX at 1600, divided by the standard deviation in minute iterations for the day. My goal was to determine if there is a difference between a day where the move was made through the entire day, or a day where the move was made in only 30 minutes or less.

I am not writing this as a topic of research that has been completed, but rather summarizing the first 5 minutes of research that I have done so far, merely defining a direction. If I were to invest from close to close the day after a reading greater than 1 (the greatest 30 minute stdv spike was larger than the stdv of the day), the equity curve grinds from 1000 to 1019… not too great.

Below 1 it moves to 1176. The SPX moved from 1000 to 1198 (indexed at 1000). There is no advantage from this alone, but I did not take into account only large range days or only small range days which could very well be an important factor, I did try smoothing the daily data using a very short term exponential average and that did not help. Lastly, I tried taking the range of the output over the prior 1 year. From 2010 until present readings above .25 grew to 1056, readings below .025 grew to 1035 and the benchmark fell to 971. This shows more promise. The readings (as might have been guessed) are clustered in the bottom of the range given a few upward outliers (and if I am correct, I should have taken the log of the function to more evenly distribute the data as the lower side is bounded by 0).

Next steps will be to isolate large range days and small range days and then apply the output. A second test I intend to run is using the standard deviation of the minute data around a non horizontal regression (around the least squares regression of the day)… finally I will examine the last half hour, hour and 90 minutes. Another way to look at this might be to divide the range in a certain period of the day by the range of the entire day.

Just starting (now 15 minutes in!), but this is how I create a tree for further testing.

Note: just realized "present" was set to 9/30/2011.


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search