Dear Nonlinear,

I think, it is perfectly right to differentiate between up- and
downward trends.
By and large downward seems in general much faster.
This would speak for a difference in parameters.
I always wanted to look at what happens if you do separately optimize
a upward-only
and a downward-only strategy and compose it later (going flat if both
are opposing)
and compare this with two-sided optimization.
But I never got around to doing this systematically. Given the
intensity you work on it, you might
want to take a look.

Cheers
  Klaus


On Mar 8, 5:16 pm, nonlinear <[email protected]> wrote:
> You are probably paying attention to the "Bias" metric in JBT when
> backtesting and optimizing the strategies. The Bias metric reflects the
> number of long trades vs the number of short trades over the strategy's test
> period. What it does *not *tell you is what percentage of profits is
> attributable to long and short trades. Over the last weekend, I made an
> interesting discovery. Despite the market run-up in the last 6 months, most
> of the gains made by my sample strategies were from the short trades. I
> guess there is something in the market dynamics which makes it possible to
> capture shorting opportunities better than the long opportunities by my
> strategies.
>
> For the next release, I am going to make these changes:
>
> 1. Rename "Bias" to "Trade Bias" and add a new metric "Profit Bias" which
> will measure the percentage of profits attributable to the long/short
> trades. Perhaps you can thing of better names for this metrics.
>
> 2. The sample strategies as they are right now make an assumption of the
> symmetry of the indicators. For example, "enter long" when indicator is
> greater than +20, enter short when indicator is less than -20, go flat when
> indicator is around 0. Given my discovery, it looks like this symmetry is
> artificially imposed, and it constrains my models. I've refactored my
> strategies to break this symmetry, and I am running them live this week.
> I'll include the corresponding sample strategies in the next release.
>
> Feel free to contribute your thoughts in this subject.

-- 
You received this message because you are subscribed to the Google Groups 
"JBookTrader" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/jbooktrader?hl=en.

Reply via email to