Is there a way to create a live system that can identify and output any stock that has had a big (like 3 SD) move vs its own variability over say 6 months. How can one set it up so it outputs every trading day?
Sorry, as you can see I do not spend much time coding
A screen with the rule PctDev(2,1) - PctDev(63,1) > 3 might do the trick. You can run it every morning. You can also set the API to run it every morning . . .
Thanks Yuval. Is there a tutorial on how to use the API that you know of?
For info on the API, see the Knowledgebase:
https://portfolio123.customerly.help/en/dataminer-api
And the API specs:
https://api.portfolio123.com/docs/index.html
I’m not sure if this is accurate, but Yuval’s approach involves subtracting one volatility sample from another. For instance, if IBM’s volatility over 2 days is 5% and over 3 months is 3%, he calculates the difference to be 2%. This comparison wouldn’t pass the screening test, unlike a similar stock with volatility of 15% and 9%, respectively, which, when scaled by a factor of 3, would pass.
Correct me if I’m mistaken, but it seems you’re seeking an unusual increase in volatility that deviates by 3 standard deviations (i.e., 99% statistically abnormal). To achieve this, you need to calculate the volatility of volatility and then compare it to the volatility over a certain period. If the absolute difference is 3 or greater, it would pass the test. I’m uncertain about the p123 methodology for this but it’s a t-stat. Perhaps RTNL can attempt to clarify if I’ve misunderstood anything.
I would ratio the [percentage move of the latest closing price to the previous closing price] (absolute value if both up and down moves are desired) to the standard deviation of percentage moves over time. Perhaps:
Abs(Close(0)/Close(1) - 1)*100 / PctDev(63,1) > 3
I think PctDev(2,1) should work the same, if a ratio to the long term standard deviation is used instead of a difference.