|
PureBytes Links
Trading Reference Links
|
John Sweeney S&C wrote:
>
> Remember the thread earlier this month where people thought you should
> throw out the extraordinarily large gains when assessing a system's
> performance? I don't think you should. Those are the elephants you're
> hunting. If they never occur, you've got a system/tradable with no real
> upside potential.
>
Perhaps even a trend following system testing should *sometimes* discard
the elephant.
It would be a judgement call but maybe it depends on how you got in the
"big one" and its character. In other words suppose you have a trend
following system that gets in after about 10 days of "trending"
behaviour. If a trend goes gradually for 10 days you get in, and the
trend eventually accelerates then it's reasonable to keep the trade, as
it's a fair test of your exits' ability to ride a long trend. OTOH, if
you got in on the 10th day and then the 11th day a huge gap occurred up
that was the bulk of the big profits, then you shouldnt count it,
because it's almost random chance that you were in to experience the
gap.
This probably shows if you vary your parameters, but if you choose the
most profitable parameter, and if it had one of the "chance" type
trades, perhaps you should discard it.
Maybe this is too subjective?
{see Kaufman Smarter Trading for some discussion of "price shocks" and
what to do about them in trading and testing}
Conrad Bowers
[definately not a systems expert]
|