1

In the screenshot, you will see the daily prices of Nasdaq. Each candle has a High, Low, Open and Close price. I have drawn a regression line with a 2 standard deviation channel on either side. How would I go about determining the following odds:

Price reversing back to the mean from the top of the channel Price reversing back to the mean from the bottom of the channel

Here is the relevant data in CSV format:

Date,open,high,low,close
2020-04-22,8638.04,8791.67,8584.55,8606.75
2020-04-23,8606.75,8786.69,8503.14,8773.93
2020-04-26,8773.93,8904.59,8736.59,8824.04
2020-04-27,8824.04,8953.96,8662.47,8707.38
2020-04-28,8707.38,9128.59,8707.38,9128.17
2020-04-29,9128.17,9154.75,8866.86,8879.61
2020-04-30,8879.61,8888.91,8682.93,8727.01
2020-05-03,8727.01,8838.2,8568.02,8820.48
2020-05-04,8820.48,9030.75,8810.19,8945.63
2020-05-05,8945.63,9068.2,8890.4,8957.83
2020-05-06,8957.83,9137.67,8942.08,9129.77
2020-05-07,9129.77,9246.73,9115.09,9225.91
2020-05-10,9225.91,9346.54,9127.77,9289.57
2020-05-11,9289.57,9354.5,9048.46,9050.96
2020-05-12,9050.96,9212.14,8886.47,9022.94
2020-05-13,9022.94,9113.27,8856.21,9097.25
2020-05-14,9097.25,9156.71,8933.83,9103.21
2020-05-17,9103.21,9369.66,9103.21,9324.09
2020-05-18,9324.09,9424.65,9291.39,9305.1
2020-05-19,9305.1,9502.52,9280.79,9499.07
2020-05-20,9499.07,9515.22,9355.88,9363.9
2020-05-21,9363.9,9422.59,9246.58,9410.25
2020-05-24,9410.25,9541.35,9395.05,9534.01
2020-05-25,9534.01,9608.78,9376.65,9416.38
2020-05-26,9416.38,9512.9,9177.67,9446.79
2020-05-27,9446.79,9569.72,9324.69,9462.54
2020-05-28,9462.54,9586.35,9376.22,9580.78
2020-05-31,9580.78,9609.06,9454.77,9596.79
2020-06-01,9596.79,9673.83,9509.08,9661.21
2020-06-02,9661.21,9730.72,9638.96,9692.7
2020-06-03,9692.7,9744.5,9574.55,9645.42
2020-06-04,9645.42,9847.5,9603.96,9810.66
2020-06-07,9810.66,9902.49,9748.95,9884.84
2020-06-08,9884.84,10006.7,9813.53,9963.26
2020-06-09,9963.26,10157.12,9960.27,10088.48
2020-06-10,10088.48,10108.41,9585.13,9621.5
2020-06-11,9621.5,9849.63,9495.38,9646
2020-06-14,9646,9816,9381.75,9811.77
2020-06-15,9811.77,10014,9797.8,9969.67
2020-06-16,9969.67,10059.42,9926.92,9998.25
2020-06-17,9998.25,10041.38,9879.25,10003.09
2020-06-18,10003.09,10125.67,9929.69,9932.92
2020-06-21,9932.92,10147.67,9856.86,10134.79
2020-06-22,10134.79,10309.42,9985.74,10190.49
2020-06-23,10190.49,10255.2,9941.56,10029.03
2020-06-24,10029.03,10120.51,9899.8,10109.29
2020-06-25,10109.29,10134.14,9837.73,9880.49
2020-06-28,9880.49,10010.65,9743.03,9999.65
2020-06-29,9999.65,10184.18,9953.2,10151.46
2020-06-30,10151.46,10321.62,10088.89,10268.39
2020-07-01,10268.39,10433.51,10259.41,10360.82
2020-07-02,10360.82,10400.9,10320.4,10338.29
2020-07-05,10338.29,10626.21,10338.29,10610.1
2020-07-06,10610.1,10706.55,10517.89,10538.77
2020-07-07,10538.77,10685.05,10517.73,10683.92
2020-07-08,10683.92,10786.46,10572.29,10735.73
2020-07-09,10735.73,10854.72,10637.74,10850.22
2020-07-12,10850.22,11070.48,10574.11,10610.96
2020-07-13,10610.96,10705.54,10371.63,10656
2020-07-14,10656,10778.78,10563.94,10701.21
2020-07-15,10701.21,10710.39,10488.15,10546.97
2020-07-16,10546.97,10681.67,10535.74,10636.56
2020-07-19,10636.56,10972.65,10558.93,10959.53

Many thanks.

enter image description here

Grantx
  • 13
  • 2
  • 1
    if you want help from stats community you need to explain and clearly define your terms: price reversal, mean, channel etc. these are very TA domain specific terms. alternatively, you can go to https://quant.stackexchange.com/ site. you may not like what they'll tell you though. TA is nonsense, in my opinion, but it can be great to establish this on your own by actually back testing different TA techniques. – Aksakal Jul 20 '20 at 20:56
  • Why linear regression and not logistic regression? – Tim Jul 20 '20 at 21:19
  • I don't know @Tim. Why would I use one over the other? I'm here to learn. – Grantx Jul 20 '20 at 22:15
  • @Grantx linear regression can predict anything between $-\infty$ to $\infty$ so it is a pretty poor choice if you need to predict probabilities. Logistic regression predict values between 0 and 1, so probabilities. – Tim Jul 20 '20 at 22:36
  • Thank you @Tim. You should make that the answer and I will give you the credit. – Grantx Jul 21 '20 at 09:01

1 Answers1

1

Linear regression is a model that can predict values in unrestricted domain, from $-\infty$ to $\infty$. If you want a regression model to predict probabilities, then the to-go model is logistic regression that restricts the outputs to $[0, 1]$ range.

For more details see other questions tagged as , for example the What is the difference between linear regression and logistic regression? thread.

Tim
  • 108,699
  • 20
  • 212
  • 390