What indicator or code do you use to determine the convexity/concavity of the price? I know MACD and Momentum can show this to a degree but they are not reliable when the price change slows after a sharp move.
Maybe comparing the current close with the previous n bars and using a formula to filter out the random jumps and dips? Or maybe there is an indicator for this?
You are trying to find highest high and lowest low? In this case, we always need confirmation before considering these price levels, to be sure that the new detected highest high is effectively a new top and vice-versa. There are many different indicators already to find them: zigzag, fractals, donchian channels, etc.
I’m also interested in a price convexity/concavity indicator. Mathematically speaking, it’s a regression of prices (or variations of price) on time square for n periods. In other words, we have y = ax + b with y = price or variations of price and x the square of time (1 ; 2 ; 3 ; … ; n). The sign of a is the indicator of concavity or convexity. If someone can explain how to program this…
@Alf639 Do you have a picture of how the curve look like for this formula? I can help, but be aware that at the end, we’ll find the same exact tops and bottoms than any over way to find them as explained in my last post.
W%R is a great tool for new highs/lows over ‘x’ period fwiw . I find it an under rated and extremely versatile indicator . The 50 line is an Ma cross as well amongst other things
Thanks for this valuable input! You are right to say that Williams% is under rated! Will look into it 😉
With a few days delay, some complementary elements: a function is convex if its second derivative is positive. To obtain a second derivative in the form y = b, we need an approximation function that is quadratic (y = ax² + bx + c with x = t in the case of a time serie analysis) and not y = ax + b as i indicated earlier. It is therefore necessary to estimate on n periods (the only parameter of this possible indicator) using error minimization a; b and c, the parameters of the quadratic function. If “a” is positive, it will also be positive in the second derivative and the price curve estimation will be convex, otherwise (a < 0), the curve will be concave. A change in sign of a can be seen as a trading signal to enter/exit a trading position.
Of course, using a regression function has nothing to do with finding high or low points. It’s a momentum indicator…
You can find a quadratic regression function here: Quadratic regression
Made a semaphore (tops/bottoms) indicator from it: Quadratic Semaphore
A quadratic regression is the process of finding the equation of the parabola that best fits a set of data.
A quadratic regression is the process of finding the equation of the parabola that best fits a set of data.
Exactly.
Thanks for your quadratic regression program (i suppose you’ve used gradient to estimate the parameters). Information is welcome.
