[R-SIG-Finance] help with chart.Histogram of PerformanceAnalytics ?

[Nick White]

>From a quick look at your PDF, i'm willing to bet this is an intraday
strategy. So, before going on, I would ask whether this is a stock or
futures trade...if this is a stock trade (and you've presented gross
returns),  then your expectation is likely going to be eaten up by
transaction costs and slippage (unless you're paying rock-bottom execution
costs and you understand your model limits well).

But, for the general answer in lay terms and in the absence of numerical
data, you have a (literal) extreme leptokurtic distribution of returns with
what initially appears to be some positive skew. What this means is that
your average return clusters aggressively around a nominally positive
return (hooyah! you've found an effect!)...but you can have some big
outliers and tail events which will overwhelm your expectation from time to
time (bummer). I would also suspect that you've found an academically
'significant' effect if you were to measure it...but (1) in practical terms
it's probably not going to be nearly significant enough for the realities
of the market and (2) it might even not be significant given some of the
properties of what appears to happen with your negative returns...which
leads nicely to the quantile plot.

This is what the quantile plot shows...if your strategy's return was
normally distributed (and very few strategies are) then all the points
would lie along the diagonal line....what the quantile plot is showing you
is that the magnitude of your outliers increasingly deviates from normal
expectation as you go further out into the tails. I can't tell because the
quantile plot doesn't have scale, but eyeballing it looks like your
negative deviations from normal are more skewed than your positive
deviations. That is, your bad outliers are worse (and increasingly worsen
at a greater rate) than your good outliers do. Not a good sign.

So, without knowing details, this shows that you have a strategy where you
are probably nominally / implicitly "short vol". It shows you have the
foundation for probably having found some kind of effect, but you now need
to go and research to find the best way of exploiting the effect you've
discovered...and, most importantly, how to bet / size your trades on that
effect once you've optimized your signal. You might look at Maximum Adverse
Excursion data if this is intraday based.

But all of the above is moot if, in the real world, your execution costs
(explicit and implicit) outweigh your upside.

HTH

On Thu, Apr 2, 2015 at 5:48 AM, ce <zadig_1 at excite.com> wrote:

> Dear all,
>
> Sorry for the dump question, my statistics knowledge is limited  My
> questions is how to interpret chart.Histogram diagram ? What is best to
> maximize returns and minimize risk of my backtesting ?
> My line is :
>
> chart.Histogram(returns,methods =
> c("add.normal","add.density","add.centered","add.rug","add.risk","add.qqplot"))
>
> - add.normal :  should be regular bell curve , fat tail,skewed,
> leptokurtic, mesokurtic, platykurtic ?
> - add.density : should be regular bell curve , fat tail,skewed,
> leptokurtic, mesokurtic, platykurtic ?
> - add.risk: Is it better  it be far from peak of the curves or close?
> What is the difference between Var and ModVar ? Should they be close to
> each other, left, right of each other ?
> - add.qqplot: which shape is the best, all look the same to me?
>
> I had seen also R histogram help and related web site
> http://zoonek2.free.fr/UNIX/48_R/03.html
>
> I attach pdf file but if you cant see it there are similar examples on
> internet :
>
>
> http://i2.wp.com/2.bp.blogspot.com/-BnYSdT3jHNg/UWbBo4Utd9I/AAAAAAAAAcQ/2bb8iWPXwtE/s1600/daxbhhisto.png
>
> https://tradeblotter.wordpress.com/2013/01/18/visually-comparing-return-distributions/
>
> Thanks for reading
> ce
>
> _______________________________________________
> R-SIG-Finance at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-finance
> -- Subscriber-posting only. If you want to post, subscribe first.
> -- Also note that this is not the r-help list where general R questions
> should go.
>

	[[alternative HTML version deleted]]