My other indicator shows the linear regression slope of the source. This one finds the exponential regression slope and optionally multiplies it by R-squared and optionally annualizes it. Multiplying by R-squared makes sure that the price movement was significant in order to avoid volatile movements that can throw off the slope value. Annualizing the exponential slope will let you see how much percentage you will make in a year if the price continues at its current pace.
The annualized number is the number of trading days in a year. This and the length might need adjusting for the extra bars that might be in futures or other markets. The number does not have to be a year. For example, it can be a month if you set the number to 20 or so trading days to find how much you would make in a month if price continues at its current pace, etc. This can also be used as an alternative to relative strength or rate of change.
Informacje o Wersji
⋅
Changed the default year from 250 to 252, not a big difference. Changed the value to show percentage. This indicator is like rate of change but lags in order to give less weight to the outliers or the prices that don't fit the exponential regression line.
Why do you call exp on the slope here?
annualized_slope = (pow(exp(slope),daysInYear) - 1) * 100
The slope is already exponential since you already call log on the prices.
nlPrice = log(src)
Isn't this just simpler?
annualized_slope = pow(1+slope,daysInYear) * 100
SparkyFlary
⋅
@Skyte, The slope needs to be converted back by applying the exp function. This gives the percentage change per bar in the slope. It is then annualized by raising it to the power of however many days in a year there are(or however many bars in whatever length of time you want to see it).
This indicator comes from Andreas Clenow's book Stocks on the Move. I converted what he had in his Excel picture and description of it into an indicator, he specifically uses the exp function on there for that step. I then concluded it to be most likely correct because it looks similar enough to rate of change. It's not really available anywhere to double check, but I think I got it right..
Skyte
⋅
@SparkyFlary, yes, I think you did the same calculation as Clenow but I don't think it's necessary, since like I said you already used log on the price and this already gives you percent:
"A regression model will have unit changes between the x and y variables, where a single unit change in x will coincide with a constant change in y. Taking the log of one or both variables will effectively change the case from a unit change to a percent change."
The reason I asked in the first place is just that I'm using a python screener and I use the formula without the exp() (to have it simpler and I guess more accurate because of less calculations) so if you decide to change it the numbers will be more alike^^, but I guess it doesn't really matter that much.
Oh and btw thanks for writing this indicator. In any case it's nice to see the Clenow results in TradingView.
SparkyFlary
⋅
@Skyte, you could cut out the annualized slope and just use slope, that could save a little extra step in the math. Doing this however "annualized_slope = (pow(exp(slope),daysInYear) - 1) * 100" gives a number that a human can understand and relate to by knowing how much the price is going up in the form of a percentage in a given daysInYear length in time. If you're somehow very constraint on the computation you could just leave the slope part but it probably wouldn't be as useful as an indicator on a chart. I'm not too well versed in python but you might want to double check their math on that return function they did, it looks like they were trying to annualize it so it should look similar to mine hopefully.
andyong558
⋅
@SparkyFlary, hi, do you happen to have the code for thinkscript?
SparkyFlary
⋅
@andyong558, yes it can both screen and rank them all in one go, very useful
tankanghao
⋅
@SparkyFlary, I have compared Teddy Koker's code with yours on the definition of the function for annualised slope. His function returns the following: return ((1 + slope) ** 252) * (rvalue ** 2)
Notably there is a missing "minus 1" term prior to multiplication with the R2 term. So i would like to check which code is correct ? Appreciate your advice. Thank you.
SparkyFlary
⋅
@tankanghao, I got mine directly from the book, which had a picture and a somewhat step by step guide which I compared to and ended up looking similar. If I play around with the formula to match Koker's somehow with some different combinations on here with the +1, the histogram never goes negative and or the percent number gets too big and unrealistic. It's hard to say what he meant by adding 1 then putting it to the power of 252 because he didn't include taking the power of the exp of the natural log of price, maybe python did that for him automatically who knows, but I don't think his looks right without getting into python.