You need to objectively define what constitutes a trendline. After that all you have to do is find an algo that does this sufficiently efficient in real time. For an objective definition, see DeMarks New Market Timing Techniques (they are called TD Lines in his lingo). Or if you prefer to have someone decipher that into plain english, check out the same definition in Schwagers On Futures: Technical Analysis, chapter 3, Trends. For an implementation (algo) similar to what Ensign and others do, check out Stocks & Commodities V. 24:11 (page 36-44): Building Automatic Trendlines (by Giorgos E. Siligardos, Ph.D), article can be bought online. Hope this helps. EDIT: You could also check out the source code for some charting platform that lets you see it, like the NinjaTrader indicator or some stuff that is available for Sierra Chart. Personally I find their implementations lacking in more than one way. EDIT2: You did ask for a paper or a book, but also see ProfLogics post, he already outlined the general algo quite eloquently.
Gann isn't accurate either. I guess you just don't know till you either program it or have someone program it for you.
I do too for the most part but for someone just getting a handle on reading a chart they can be a nice guide. Programming a computerized version is really good with forcing someone to objectively define the term "trend" though.
@SeventhCereal Here's the math: http://www.educalc.net/2104083.page If you only need to see it on the chart and don't care to calculate it, both NinjaTrader 6.5 & 7 have it - it's called Regression Channel and it's great. As you move it with your mouse or resize it, it recomputes itself automatically. Hope this helps.
I should have used the term, consistent. Consistent and how it relates to computing the real time support and resistance levels as they are created in real time.
I think you can compute a trend ling using a time based chart. In NinjaTrader, my time based chart does automatically compute and draws trend lines, and also draws absolute HL, HH, LL, and LH My issue with the version of charting using ProfLogic indicator is that it will for example create a HL, but if the HL is violated, it will then erase the HL, and recreate another one. So if you were to view the chart for the whole day, it looks like it actually predicts a HL that predicts the future, but does not show failures of price action since it covers them up. However, my coding of HL does not erase an older HL, it creates a perfect HL based on time that actually predicts future price changes accurately. So for example, lets assume in a rule of ProfLogic is that if you were going to take a long, price 1st needs to make a HH, and then go back and make a HL. Yes, I do have all of his rules, but this is to make it simple for someone to understand and this is not to be super critical since I actually believe ProfLogic is a profitable trader. Would you rather use a chart indicator that creates a HL with the possibility that it will go back and erase it, or one that creates an absolute HL with no possibility of the chart indicator to go back and erase it?
You can compute a trend line on a time based chart but the results will not be consistent. Sorry but the only way a higher oscillation price point will be replaced with another higher price point is if it is a continuation of the same longer term oscillation. Then the label depicting that oscillation high migrates to the higher price point. Each resistance and support area is objectively, logically and specifically defined. My charts do not "predict" anything and never have. It's obvious you don't understand how I read a chart or where my decision points exist. I never take a trade based on price solely making a HH. Never have and never will. It is a physical impossibility for my charts to "erase" a Prime Resistance oscillation once a Prime Support level is attained and it is a physical impossibility for my charts to "erase" a Prime Support oscillation once a Prime Resistance level is attained. Minor support and resistance levels are replaced on a continual and ongoing basis because those are the noise oscillations in between the extreme, more important oscillation levels.