Hi JK,
<< When the trend changes, I'll be interested to see how you all adjust your settings and handle the shift. >>
As you figured out, that part is tricky and subject to error. Not so easy to define... my approach is to use other criteria to define trends, then apply regression channels. And, be prepared to disregard the channels entirely, since they may be misleading when there are trend shifts. On the other hand, they can also be useful to detect such trend shifts.
But anyway, regarding the question at hand. Let's say a stock is in an uptrend, but then heads south, fails at the 200 sma, confirms that failure by making a lower low and a lower peak below the 200, then there is a bear cross. At that point you have full confirmation of a trend reversal (by my definition at least). So, the regression channel must take that into consideration and be adjusted accordingly. It is likely it will not be very useful in the early parts of a new trend, mostly because the underlying distribution is so small that each data point has too much error/uncertainty associated with it, and hence also the calculation of line equations that define the channel.
The Schaeffersresearch site you linked is a very good one. There is a ton of information there you will be hard pressed to find elsewhere, particularly on sentiment analysis. That's one of their areas of expertise, and IMHO is an area that is widely underappreciated.
Very nice work, thanks.
T |
