Originally Posted by unicorn Why did you divide the components of the DMI oscillator by volume? |
I'm not sure if I am answering the right question here, lack of sleep perhaps. Why did I leave out the third stage of the adx?
The first two stages process a mix of momentum and price thru ema smoothing, those two stages are relatively compatible with each other.
The third stage uses a different form of historical access, it samples n=adx_bars on a bar by bar basis. That puts it out of step with the method ema's use, which sample only the current bar and add it in to a single weighted average, they only ever deal with just two numbers, not n=adx_bars of numbers.
Ema's essentially follow the recent price more closely than most forms of averaging and the third stage was further behind reality than the first two stages. This could have the output going down slightly (say) at a time when the price is tending up (say). On an adx, the main signal line only cares about the strength of the movement, it does not care which direction the movement is in.
But the oscillator did need to distinquish up from down and thanks to the third stage it was having trouble getting this right. It was apparent when the oscillator was near the zero line, that it was getting + and - mixed up, there would be jerky movements, zigzagging across the zero line.
The quick answer was to remove the third stage which was causing the error.
I have since got to a stage were it might be possible to put the adx_bars thing back in again without causing a problem but I have not had time to look at it and I am not sure if there is a need to go further with it.
A desensitizing setting has since been added to partly undo the increase in sensitivity that happened when the third stage was removed. Progress comes in pieces, one step at a time, so things are always kind of interim.
If I have answered the wrong question then please try asking again.
Cheers
Bruce