Exp Regression Slope Annualized with R-squared TRADINGVIEW

This topic has 20 replies, 6 voices, and was last updated 3 years ago by RP-Markets.

Viewing 15 posts - 1 through 15 (of 21 total)

1 2 →

03/26/2021 at 10:01 PM #165457

Vinks_o_7

I was wondering if it was possible to convert such indicator for PRT: this is a useful tool to build a portfolio by ranking stocks by strength based on Andreas Clenow’s book “Stocks on the Move” .

Here’s the strategy:

1. Determine the 90 day annualized exponential regression slope for each stock
2. Determine the 90 day R-squared value for each of the stock
3. Multiply #1 * #2 and then sort from highest to lowest —> screener PRT
4. Buy the top stocks based on the formula ((AccountValue * 0.001)/ATR(20))

Thanks.

//@version=4
// This source code is subject to the terms of the Mozilla Public License 2.0 at https://mozilla.org/MPL/2.0/

// Exponential Regression Slope Annualized with R-squared Histogram
// by SparkyFlary

study(title="Exponential Regression Slope Annualized with R-squared Histogram", shorttitle="Exponential Regression Slope Annualized with R-squared", format=format.percent, overlay=false)

length = input(90, title="Slope length")
rSquaredThreshold = input(0.0, title="R-squared threshold", minval=0, maxval=1)
withRsquared = input(true, title="Multiply by R-squared?")
withAnnualize = input(true, title="Annualize slope?")
daysInYear = input(252, title="Number of bars in a year(if slope is annualized)")
src = input(close, title="Source")

//natural log of price
nlPrice = log(src)

//alexgrover's regression slope code
x = bar_index
y = nlPrice
x_ = sma(x,length)
y_ = sma(y,length)
mx = stdev(x,length)
my = stdev(y,length)
c = correlation(x,y,length)
slope = c * (my/mx)
//inter = y_ - slope*x_
//reg = x*slope + inter

exp_slope = slope
annualized_slope = (pow(exp(slope),daysInYear) - 1) * 100
rSquared = pow(correlation(cum(1), nlPrice, length), 2)
slopeAnnualized_with_Rsquared = withAnnualize ? annualized_slope * (withRsquared ? rSquared : 1) : exp_slope * (withRsquared ? rSquared : 1)

trend = rSquared>rSquaredThreshold
rising = slopeAnnualized_with_Rsquared>slopeAnnualized_with_Rsquared[1]
falling = slopeAnnualized_with_Rsquared<slopeAnnualized_with_Rsquared[1]
up = slopeAnnualized_with_Rsquared>0
down = slopeAnnualized_with_Rsquared<0

up_rising_color = color.new(color.green, 0)
up_falling_color = color.new(color.green, 50)
down_falling_color = color.new(color.red, 0)
down_rising_color = color.new(color.red, 50)

col = trend and up and rising?up_rising_color:trend and up and falling?up_falling_color:trend and down and falling?down_falling_color:trend and down and rising?down_rising_color : color.gray

plot(slopeAnnualized_with_Rsquared, title="Annualized slope x R-squared", style=plot.style_histogram, color = col, linewidth=3)

//@version=4

// This source code is subject to the terms of the Mozilla Public License 2.0 at https://mozilla.org/MPL/2.0/

// Exponential Regression Slope Annualized with R-squared Histogram

// by SparkyFlary

study(title="Exponential Regression Slope Annualized with R-squared Histogram", shorttitle="Exponential Regression Slope Annualized with R-squared", format=format.percent, overlay=false)

length = input(90, title="Slope length")

rSquaredThreshold = input(0.0, title="R-squared threshold", minval=0, maxval=1)

withRsquared = input(true, title="Multiply by R-squared?")

withAnnualize = input(true, title="Annualize slope?")

daysInYear = input(252, title="Number of bars in a year(if slope is annualized)")

src = input(close, title="Source")

//natural log of price

nlPrice = log(src)

//alexgrover's regression slope code

x = bar_index

y = nlPrice

x_ = sma(x,length)

y_ = sma(y,length)

mx = stdev(x,length)

my = stdev(y,length)

c = correlation(x,y,length)

slope = c * (my/mx)

//inter = y_ - slope*x_

//reg = x*slope + inter

exp_slope = slope

annualized_slope = (pow(exp(slope),daysInYear) - 1) * 100

rSquared = pow(correlation(cum(1), nlPrice, length), 2)

slopeAnnualized_with_Rsquared = withAnnualize ? annualized_slope * (withRsquared ? rSquared : 1) : exp_slope * (withRsquared ? rSquared : 1)

trend = rSquared>rSquaredThreshold

rising = slopeAnnualized_with_Rsquared>slopeAnnualized_with_Rsquared[1]

falling = slopeAnnualized_with_Rsquared<slopeAnnualized_with_Rsquared[1]

up = slopeAnnualized_with_Rsquared>0

down = slopeAnnualized_with_Rsquared<0

up_rising_color = color.new(color.green, 0)

up_falling_color = color.new(color.green, 50)

down_falling_color = color.new(color.red, 0)

down_rising_color = color.new(color.red, 50)

col = trend and up and rising?up_rising_color:trend and up and falling?up_falling_color:trend and down and falling?down_falling_color:trend and down and rising?down_rising_color : color.gray

plot(slopeAnnualized_with_Rsquared, title="Annualized slope x R-squared", style=plot.style_histogram, color = col, linewidth=3)

Vinks_o_7

Participant

Topics: 9

Replies: 24

Been thanked: 0 times

03/26/2021 at 10:12 PM #165459

Vonasi

Vinks_o_7 – Please use the ‘Insert PRT Code’ button even when inserting non PRT code as it differentiates it from the other text and makes your post far more readable. Your post has been edited.

Vonasi

Moderator

Topics: 192

Replies: 5088

Been thanked: 1811 times

03/28/2021 at 7:49 PM #165563

Vinks_o_7

sorry Vonasi, thank you.

Vinks_o_7

Participant

Topics: 9

Replies: 24

Been thanked: 0 times

03/29/2021 at 8:45 AM #165600

Nicolas

Please post a picture of the indicator.

Nicolas

Keymaster

Topics: 81

Replies: 20863

Been thanked: 4773 times

03/29/2021 at 5:28 PM #165660

Vinks_o_7

Hi Nicolas,

Here’s a picture and a direct link to the page

The idea is to use it as a screener to make a stock selection.

Thanks again.

Vinks_o_7

Participant

Topics: 9

Replies: 24

Been thanked: 0 times

03/30/2021 at 8:07 AM #165704

Nicolas

I can help you well, but help me in return by posting the picture yourself in this topic please. I have a lot of things to answer again this morning. Thank you.

Nicolas

Keymaster

Topics: 81

Replies: 20863

Been thanked: 4773 times

03/30/2021 at 5:02 PM #165767

Vinks_o_7

Hi Nicolas

Thanks for your feedback.

I did it two times but for whatever reason, the pic did not show up.
I uploaded a screen capture (PNG file, 75ko)…same problem.

If you click the link, you can see the picture/detail: https://fr.tradingview.com/script/QWHjwm4B-Exponential-Regression-Slope-Annualized-with-R-squared-Histogram/

Vinks_o_7

Participant

Topics: 9

Replies: 24

Been thanked: 0 times

04/08/2021 at 8:49 PM #166578

Vinks_o_7

too bad that nobody’s interested in this winning strategy…

Vinks_o_7

Participant

Topics: 9

Replies: 24

Been thanked: 0 times

09/16/2021 at 4:06 PM #177735

RP-Markets

Hello Nicholas

It would be really really nice if you could implement this screener. Actually you don’t need the indicator itself in the chart, but as a criterion to screen a list with it.

The annualized exponential regression multiplied by the R² value is then simply the sorting criterion, similar to how you can sort lists by RSL. However, if the stock price intersects the 100 day average, the indicator gets the value 0 and you get out of a stock.

Advantage of the method of Clenow is that one finds little volatile values, which run like clockwork upward.

Many thanks in advance.

Greetings Ronny

Translated with http://www.DeepL.com/Translator (free version)

RP-Markets

Participant

Topics: 6

Replies: 21

Been thanked: 2 times

09/16/2021 at 4:07 PM #177736

RP-Markets

picture

RP-Markets

Participant

Topics: 6

Replies: 21

Been thanked: 2 times

09/16/2021 at 6:29 PM #177742

jerome148

Hi,

Here is the code for the screener (might be incorrect so please check). Note that only the 50 first results are returned in the screener. It makes it difficult to verify that the stocks you own are still in the 20% of the stock universe. There are workarounds though (such as creating multiple screeners with a condition on the momentum value).

Cheers

Jerome

val = (EXP(250*LOG(EXP(LinearRegressionSlope[90](log(close)))))-1)*R2[90](log(close))
capi = Average[90](Volume)*Average[90](Close) > 50000
bullish = close > Average[100](Close)
testVol = abs(close/close[1]-1) > 0.15
volatile = summation[90](testVol)
penny = Close < 5
SCREENER[bullish and capi and  not volatile and not penny](val as "Momentum")

val = (EXP(250*LOG(EXP(LinearRegressionSlope[90](log(close)))))-1)*R2[90](log(close))

capi = Average[90](Volume)*Average[90](Close) > 50000

bullish = close > Average[100](Close)

testVol = abs(close/close[1]-1) > 0.15

volatile = summation[90](testVol)

penny = Close < 5

SCREENER[bullish and capi and not volatile and not penny](val as "Momentum")

1 user thanked author for this post.

Swingueur

jerome148

Participant

Topics: 0

Replies: 4

Been thanked: 1 time

09/17/2021 at 7:36 AM #177759

RP-Markets

Thanks a lot. 🙂

RP-Markets

Participant

Topics: 6

Replies: 21

Been thanked: 2 times

09/17/2021 at 7:43 AM #177764

RP-Markets

Would it still be possible to sort the results in descending order by the value ExpRegression*R2? (I think this should be the variable “val”)
The largest value above

RP-Markets

Participant

Topics: 6

Replies: 21

Been thanked: 2 times

09/17/2021 at 7:52 AM #177766

RP-Markets

Sorry now i see val = Momentum

but I actually forgot one point.
Can you still exclude stocks that have made a gap of more than 15% in the last few days?

Thank you very much for your work

RP-Markets

Participant

Topics: 6

Replies: 21

Been thanked: 2 times

09/17/2021 at 7:52 AM #177767

robertogozzi

As you can see from line 7, VAL is the sorting criterion in use.
As to descending order, you can choose the preferred order clicking the title of the column in ProScreener.

robertogozzi

Moderator

Topics: 69

Replies: 17151

Been thanked: 4040 times

Author

Posts

Viewing 15 posts - 1 through 15 (of 21 total)

1 2 →

Find exclusive trading pro-tools on