|
PureBytes Links
Trading Reference Links
|
Dark Hacker wrote:
> I was musing a bit today about system performance. Let's say that you
> code up a system that produces hideous buy/sell signals... and I mean
> all the wrong signals at all the wrong times. I'm sure we all have
> well maybe one or two of these lying around :-)...
>
> Ok, now turn this around as a contrarian indicator. Sell when it
> says buy, buy when it says sell. So if it looses 89 trades out of
> 100 and wins only 11, by turning the logic around you should have
> winners 89 percent of the time ...
>
The problem usually is ( and I think Schwager mentioned this in one of his
books) that the 89 losing trades, when you add up transaction costs are
only "marginal" losers and if you then take potential drawdowns into
account, not much of a candidate for reverse engineering. An easy way to
find out if it pays to reverse the system is to insert the "system equity"
plot on the chart.
If you can see the equity curve continuously going down into negative
territory and not taking you for a rollercoaster ride, then you may have
something.
On the same line of thought, what if your system produced 89 marginal
winners and 11 BIG losers? That to me would make sense investigating as a
possible candidate for reversing the signals (again depending on other
factors).
Carlos Lourenco
|