The Perils of Optimisation

TWI

Senior member
Messages
2,559
Likes
269
The Perils of Optimisation
 

Attachments

  • Optimisation.JPG
    Optimisation.JPG
    57.9 KB · Views: 496
Indeed FC, counter trend system on Dax. I did not actually trade this system, it was something somebody who had recently got to grips with TS posted to me and that I decided to watch. It worked a bit better on FTSE but still down vs. Aug peak. It seems they fell into the usual system trader trap when one gets first access to the evil temptation of the optimise button even though they swore it was not over optimised.
I posted to show how easy it is to go horribly wrong. Expect a lot of commercial systems may perform like this.
 
And perhaps the perils of thinking that the most recent data is the most important.
5 years minimum imho, including bull, bear, flat, wild and boring periods.
By the time you understand the latest sentiment, it's usually over.
Glenn
 
I think time period may depend on the granularity of data. The system above worked off 10 minute bars. Is that not equiv to about 60 times the number of daily bars?
I just think it is unstable "peak value" variables or too many variables that may be the problem. I should think this system may be workable as the idea behind the signals is actually fairly sound it is just that the 8 variables have been totally optimised for the Jan-Aug data.
 
FetteredChinos said:
counter-trend system by any chance?

I'm curious as to why/how you guessed it was a counter-trend system. Was it because the equity curve looked very flat (until it wasn't!) or was it because it abruptly tanked, or is there some other reason?
 
well, counter trend systems usually produce flatter equity curves..

with hit rates of 70% plus, lots of small winners usually mean drawdown during stable periods is pretty limited.

then of course, when things go wrong, they go wrong in style - viz being on the wrong side of the trend in the last couple of months..

last couple of months have been pretty trendy (tank tops, combat trousers) which is what made me think it was a counter trend strategy.


they usually make a killing when the market is range bound.

i suppose the simplest way is to trade 2 systems simultaneously. one trend following, the other counter trend..

should smooth out the equitry curve a treat..

FC
 
"I think time period may depend on the granularity of data. "
If your data sample was taken in, say, a bull market then that's when it will work better, no matter whether intraday or daily etc.

"By the time you understand the latest sentiment, it's usually over."
That's why people buy tops and sell bottoms. Systems are the same.
Glenn
 
FC,t walker good posts
I trade the Dax, and used to be a pure trend trader, then more counter trend in the last year or so.

When I first tried trading a mixture of trend and counter trend (for same reasons as you say above) I found it very hard to be mentaly dextrous enough to buy and hold for trend, then to counter trend on smaller more limited targets. At the time, there were enough trend moves for it to not matter standing aside whilst waiting solely for the trends. Then volatility gradually dried up. So traded both.

I still find I am lacking in mental dexterity to switch easily between the two, particulary when you get a couple of range bound days. Theoreticaly, I try to mark lines on the chart where an imaginary sign pops up and the counter trend "trading tools" get locked up in a cupboard, to be replaced by the trend day tools.
My equity curve hasn't suffered as badly as the system above, but there has been quite a bit of pain all the same.
 
Top