[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: winning system ?



PureBytes Links

Trading Reference Links

--- In amibroker@xxxx, "David Holzgrefe" <dtholz@xxxx> wrote:
> Maybe a few other could run it on there indexes to see if performs 
the same ?
> 
> an give me some feed back as the results seem a little to good ..
> 
> 
> Thanks David 
> 
> prevclose = ref( close, -1);
> signl = iif( close > prevclose , 1, 0 );
> 
> BUY = CROSS(bbandbot(close, 15, 2),close) AND RSI(14) < 35 AND 
signl = 1;
> 
> SELL = CROSS(bbandtop(close, 15, 2),close) AND RSI(14) > 70;

> an give me some feed back as the results seem a little to good ..
> 
> 
> Thanks David 
> 
> prevclose = ref( close, -1);
> signl = iif( close > prevclose , 1, 0 );
> 
> BUY = CROSS(bbandbot(close, 15, 2),close) AND RSI(14) < 35 AND 
signl = 1;
> 
> SELL = CROSS(bbandtop(close, 15, 2),close) AND RSI(14) > 70;

The major problem with this simple system is the huge drawdowns 
between those "winning" trades. Look at the last few months for 
AAPL, Apple Computer.
The system did its last buy in late May 2000 at ~45, the stock then 
went to 14 in Dec 2000 before rallying. No one should or could live 
with a drawdown like that....

The problem is one that is commonly seen in a simple indicator based 
system. You get the initial trade entry and then the exit conditions 
are not satisfied for months or years. Meanwhile the price goes 
wrong and you are still in the stock. Sure it backtests "well" with 
nice profits and a hit rate in the 70-90% area, but it is a bad 
system.

The thing that most back tests is not telling you is the maximum 
drawdown (maximum adverse excursion) during the trade. Usually you 
are provided data that lists the drawdown from the entry point. 
While this is a useful number in its own right it can mislead you 
drasticaly.

In the AAPL example, price got up to the 63 and change area. If you 
had bought at the 45 price and exited at 55-60 you would have had a 
nice trade. However, once the price got to 63 it went to 14. 
Maximum drawdown for this trade was 63-14 = 49, really ugly....

With this system there is no assurance the price will exceed the 
upper BB with an RSI above 70. Now that may be a good exit point but 
it may never be reached. So, the system needs to have other exits 
added to it until you are guaranteed there will be a timely exit that 
controls the max drawdown.

As an aside, I have not looked at the work done recently by Dimitris 
but I suspect the same problems may be there as well. Several years 
ago I did a lot of work with divergences. They can provide really 
nice entries, but will miss some entries because there may not be a 
divergence. Even worse, they will miss exits because the divergence 
may not occur. So you have to also include fixed exit levels or 
trailing stops or some other exit that will signal for "every" exit 
required. Again I have not had time to look at his work but I would 
look for this possibility.

Bottom line, back testing is a tricky business and one should always 
examine the signals applied to the chart to see if the signal "makes 
sense" by not missing entries or worse, missing exits. Take care of 
the big drawdowns and the system profits will usually be fine.

If you have questions on this, feel free to ask....
Trader