What works instead is using the weighted price or to a lesser extent the median price.
By cutting off the wicks you effectively get something like a median price because you filter out a big part of noise (the wicks) and get a trendline closer to the median (center of candle body).
What you say is true for one or two candles that you can change around at will. But what you need to draw a trend line is a larger number of candles. For 10 or 20 candles the effect of changing cut off times (changing time zone) evens out.
can't believe I just did this...
what do these 2 have in common?
exactly, the same underlying data:
(just with "timezone" offset by 50% of a candle width)
EDIT: and this was just randomly drawn without even trying to maximize difference.
you don't "filter out noise", you just get "different noise"
Nice example If you do the same thing with more candles you will see that the trend becomes more precise by the number of candles you add.
As mentioned above, you can change around a few candles at will but cant do the same thing with many candles more. Then you WILL actually filter out some noise by cutting off the wicks and shadows that would distort your trendline other wise. An extreme example to illustrate what I mean:
Where would you draw the trendline ?
Another view using the same sample (I could have choosen a better one but Im too lazy right now):
Which of the 3 trendlines provides the earliest trend reversal indication? (remember: the internal trendline is considered broken when 1 or 2 candles close above it, depending on how much reassurance you prefer)