What are the best Volaltility indicators on WealthLab for an end-of-day breakout strategy? Or is there a strategy online?
Thank you for your help.
Thank you for your help.
Rename
ATR
HiLoLimit (community)
This one is more for a Mean Reversal system but you could adapt the basic idea for breakouts.
This one is more for a Mean Reversal system but you could adapt the basic idea for breakouts.
Historical Volatility (HV) and its relatives (StdDev, Variance, and the like)
are not recommended because they are too sensitive to (these always present) outliers.
are not recommended because they are too sensitive to (these always present) outliers.
A very good guard agaonst outliers is the moving percentile (MP, from the robust family)
Use it as a "improvement" for any existing volatility measure.
This one will improve your results.
Use it as a "improvement" for any existing volatility measure.
This one will improve your results.
Conclusion: My best guess is MP(ATR).
A high precission, high resolution solution is "Short Term Volatility" (STVola, available with the finantic.Kalman extension)
This is best suited for short term strategies.
This is best suited for short term strategies.
... and if you are prepared to get really fancy about this...
... you could adapt some Kalman filter to estimate Volatility.
I am sure this will give you the best results...
...after some research effort...
... you could adapt some Kalman filter to estimate Volatility.
I am sure this will give you the best results...
...after some research effort...
Lets go back to the basics for a moment:
The broadest definition for a "Long Breakout Strategy" I can think of is:
"Enter if (the price movement of) the stock is doing something unusual in the upward direction."
The key here, of course is the word unusual.
To get more concrete and to focus on the topic of your original question we could say unusual means: The short term volatility is unusual high, or in other words, the short term volatility exceeds the long term volatility by a certain degree.
And here we are back to your original question, broken down in this context:
1.) What is the best way to estimate long term volatility
2.) What is the best way to estimate short term volatility and
3.) What exactly means "long term"
The broadest definition for a "Long Breakout Strategy" I can think of is:
"Enter if (the price movement of) the stock is doing something unusual in the upward direction."
The key here, of course is the word unusual.
To get more concrete and to focus on the topic of your original question we could say unusual means: The short term volatility is unusual high, or in other words, the short term volatility exceeds the long term volatility by a certain degree.
And here we are back to your original question, broken down in this context:
1.) What is the best way to estimate long term volatility
2.) What is the best way to estimate short term volatility and
3.) What exactly means "long term"
Lets start with2.) because it is the simplest one:
Measuring Short Term Volatility
Here we want to see the the effect of "outliers" so we don't care about filtering them out. This means all sorts of statistical measures are possible: Variance, Standard deviation, True Range, and so forth.
One of the quickest was to measure short term volatility is simply the latest price move: ROC(1)
Or ultimately a stop order which even captures the intraday move the next day.
Measuring Short Term Volatility
Here we want to see the the effect of "outliers" so we don't care about filtering them out. This means all sorts of statistical measures are possible: Variance, Standard deviation, True Range, and so forth.
One of the quickest was to measure short term volatility is simply the latest price move: ROC(1)
Or ultimately a stop order which even captures the intraday move the next day.
It is more difficult to answer question 3.): How long is "long term".
If the markets heat up, a volatility level that was unusually high last week could be the new normal this week. If your baseline volatility measurement comes form the "last two month" you probably judge to many das as unusual and generate too many "false signals".
This means we need a way to measure volatility as fast as possible, but without too many fluctuations, i.e. we want a "stable" measurement which can adapt quickly to changing conditions in the market - not simple.
Usually I create a chart with my latest "volatility indicator" and experiment with things like "lookback period" until the chart looks good, i.e. the volatility indicator looks smooth enough but changes quickly if prices show higher (or lower) volatility for some time.
If the markets heat up, a volatility level that was unusually high last week could be the new normal this week. If your baseline volatility measurement comes form the "last two month" you probably judge to many das as unusual and generate too many "false signals".
This means we need a way to measure volatility as fast as possible, but without too many fluctuations, i.e. we want a "stable" measurement which can adapt quickly to changing conditions in the market - not simple.
Usually I create a chart with my latest "volatility indicator" and experiment with things like "lookback period" until the chart looks good, i.e. the volatility indicator looks smooth enough but changes quickly if prices show higher (or lower) volatility for some time.
1.) Estimating Long Term Volatility
It is an often overlooked fact that most statistical computations are based on the assumption of normally distributed values.
On the other hand it is a well-known fact that (changes of) financial time series do not show a normal distribution, but have too many outliers, which form fat tails.
Taken together this means that most statistical computations return other results than expected (invalid results to be more precise) when done for financial time series.
This applies to arithmetic mean, standard deviation, variance, and so forth. Or - translated to indicators - this applies to:
EMA, SMA and most other moving averages (internally based on some form or arithmetic mean)
HV (Historical Volatility based on Variance)
Bollinger Bands (based on both of the above)
and so forth.
There are (at least) two methods to cope with these difficulties:
a) robust statistics
b) transformations of the original distribution
It is an often overlooked fact that most statistical computations are based on the assumption of normally distributed values.
On the other hand it is a well-known fact that (changes of) financial time series do not show a normal distribution, but have too many outliers, which form fat tails.
Taken together this means that most statistical computations return other results than expected (invalid results to be more precise) when done for financial time series.
This applies to arithmetic mean, standard deviation, variance, and so forth. Or - translated to indicators - this applies to:
EMA, SMA and most other moving averages (internally based on some form or arithmetic mean)
HV (Historical Volatility based on Variance)
Bollinger Bands (based on both of the above)
and so forth.
There are (at least) two methods to cope with these difficulties:
a) robust statistics
b) transformations of the original distribution
a) robust statistics
There is a set of statistical calculations which are not influenced by outliers or, even better, makes no assumptions about the underlying distributions.
(see Robust Statistic on wikipedia: https://en.wikipedia.org/wiki/Robust_statistic)
You can use this in the context of trading strategies by replacing normal statistics by robust statistics. Some examples:
Use Median instead of arithmetic mean.
Use Interquartile range instead of standard deviation.
It is possible to calculate a moving median with the MP(50) indicator.
A moving interquartile range would be MP(75)-MP(25).
There is a set of statistical calculations which are not influenced by outliers or, even better, makes no assumptions about the underlying distributions.
(see Robust Statistic on wikipedia: https://en.wikipedia.org/wiki/Robust_statistic)
You can use this in the context of trading strategies by replacing normal statistics by robust statistics. Some examples:
Use Median instead of arithmetic mean.
Use Interquartile range instead of standard deviation.
It is possible to calculate a moving median with the MP(50) indicator.
A moving interquartile range would be MP(75)-MP(25).
b) transformations of the original distribution
The basic idea goes as follows:
Whenever you encounter an outlier, i.e. a value which is way too far away form the average, you replace this value by another value which would be considered a heavy outlier in a normal distribution.
Lets say we encounter a value which is 5 sigma away form the the mean (a heavy outlier). We'd correct it to be 2.7 sigma away form the mean (an acceptable outlier in a normal distribution).
All values between minus two sigma and plus two sigma go untouched.
After such a transformation you get a normally distributed set of values.
Such a transformation is done by the B2N indicator ("Back to normal").
The basic idea goes as follows:
Whenever you encounter an outlier, i.e. a value which is way too far away form the average, you replace this value by another value which would be considered a heavy outlier in a normal distribution.
Lets say we encounter a value which is 5 sigma away form the the mean (a heavy outlier). We'd correct it to be 2.7 sigma away form the mean (an acceptable outlier in a normal distribution).
All values between minus two sigma and plus two sigma go untouched.
After such a transformation you get a normally distributed set of values.
Such a transformation is done by the B2N indicator ("Back to normal").
OK, just bite the bullet and format a blog article we can post here!
Have a look at this TASC article below from 2016. Never used it however by today, I would prefer the "volatility contraction" approach before the volatility breakout.
http://traders.com/Documentation/FEEDbk_docs/2016/06/TradersTips.html#item6
http://traders.com/Documentation/FEEDbk_docs/2016/06/TradersTips.html#item6
QUOTE:
format a blog article
What would be the preferred format?
Word doc? Markdown? Other?
Markdown
Thanks so much for all the input :-)
À propos, there is an interesting article in the April issue of TASC: "Average True Range Breakouts"...
Where can I find this please? :
DrKoch7
16. März 2023, 21:56 - 2 days ago
#2
HiLoLimit (community)
DrKoch7
16. März 2023, 21:56 - 2 days ago
#2
HiLoLimit (community)
Have you tried searching it in WL?
oh sorry, missed that.
found it, what's the best condition block to use it?
Thanks for your help
Thanks for your help
QUOTE:
... adapt some Kalman filter to estimate Volatility. ... will give you the best results.
Just to clarify, we are talking about taking the "volatility" (measured robustly somehow) of the residual value, where ...
CODE:My question is, how many degrees of freedom are you figuring in the Kalman filter above? Is it just one first-derivative term (like WL's Kalman indicator uses) or are you including a second-derivative term as well?
residual = raw price - Kalman(raw price)
A traditional moving average will only have one coefficient (one degree of freedom, which is the moving slope), so it's obviously going perform poorer than a model based on the Kalman filter with two "significant" coefficients (two degrees of freedom).
Your Response
Post
Edit Post
Login is required