Unless I’m mistaken, it is well established that high period moving averages can act as a representation of support and resistance levels.
If this is true, then the trick would be to determine what period (how many previous days) to use and where the MA is applied to (High, Low, Close, etc).
From what I’ve read, most people just use a large number, perhaps one that even encompasses the previous year, but ultimately choosing a number is a crap-shoot and would be logically different for each pair.
Additionally, the accuracy of these moving averages-as-support/resistance is limited to a specific window of time. That is to say if you some how found a moving average setting that strongly correlates with perceived support/resistance levels, then as soon as you switched to a different time window, the MA would lose accuracy with regard to the s/r levels.
This leaves me to wonder…
What would happen if someone used a brute-force computer program to apply a large number of varying Moving Averages and tallied how well each one correlated with previous s/r levels? I would guess that each Moving Average would be trivially ineffective with the exception of a few narrow ranges of MAs.
Once you compile a list of MAs that show a strong correlation to the s/r levels, then not only would they be fairly useful for technical analysis, but you could also look back and figure out why that specific MA is effective.
I imagine it would be interesting
Uploaded with ImageShack.us
Did this manually; the colored MA’s are the ones with noticeably higher conformity.
I used a range of 100 to 720 days in increments of 20 days; Some of the MAs are applied to highs (dark grey), some to lows (light grey), and for a few of them I chose linear weighted as the method (white) instead of exponential (grey).
The conformity of the colored MAs was apparent no matter how far back I went; the window I used was simply the most recent.
The five best Moving Averages were:
340 High, Exponential
340 Low, Exponential
140 High, Exponential
140 Low, Exponential
100 Low, Linear Weighted
I don’t know about you guys, but I find that interesting as hell.
Thoughts?