A little further information to my post above. The period involved is 2006, ie the trading period started on 1 January 2006 and is based on trading Australi's top 300 stocks. I am trading CFDs with a $20,000 account giving me total trading capital (and risk) of $200,000. Results are quite similar (although obviously the profits are comparatively higher) over longer periods.

Thanks Jose. That was my view as well, particularly because of the significant difference in the drawdowns between the two.

My problem, is that System A is based on not having any form of stop loss and the MAE is occasionally near 100% as a result. System B uses a 3*ATR stop loss and has significantly lower MAEs. Otherwise, the trading rules for both systems are identical.

When I do the backtesting over a longer period (1 January 2000 to present) I get very similar results with a maximum drawdown of 5.57% for System A (no stop) and a maximum drawdown of nearly 54% with System B (stop included). These drawdown percentages, of course, are based on my $20,000 margin capital not my full risk capital of $200,000.

So where does this leave me? Do I opt for the 'better' system without a stop loss at all and accept that I will get hit on individual stocks from time to time? Do I try and develop a better stop loss (although I am attacted to volatility based stop losses like the simple ATR one I am using to test).

All advice welcomed with glee!

John

Hmmm. Something seems amiss. WHY does the 3*ATR stop hit performance so hard?

If you'd like another viewpoint, using US data, drop me an email (JohnathanStein@yahoo.com) w/ the exploration. I'm curious enough to run the TS Monte Carlo on a couple of datasets (components of SP500, DJ30, Nasdaq 100).

Bullstock

To quote Wikipedia, "Monte Carlo methods are a widely used class of computational algorithms for simulating the behavior of various physical and mathematical systems, and for other computations."

In my terms, it allows you to test a trading system without the restriction that a series of trades will always have the same outcome. For example, if your trading system produces ten alerts on a given day but you only have capital to trade, say, four of those alerts then Monte Carlo analysis will consider all ten possible trades (but only select four) and the subsequent effect on your outcome rather than just accept the first four. It will typically run a simulation many thousands of times and consider a multitude of different paths. Those results are then averaged and you can have significantly more confidence in the outcome as a result.

I use TradeSim (Enterprise) for my backtesting. http://www.compuvision.com.au/

I think Jose's point was that I should develop high quality exits and not rely solely on stop losses to get me out of trades. His point is well made and I fully agree. He may wish to add more.

Hope this is of some help.

Regards

John

Dear John, Jose,

I have some more questions:-

1) Does the analysis results, posted at the beginning of this thread, come from TradeSim?

2) Can TradeSim analyse intraday data?

3) Re "I think Jose's point was that I should develop high quality exits and not rely solely on stop losses to get me out of trades. His point is well made and I fully agree.", will you two give any advice or suggestion on developing high quality exits? .... I know that it is a broad question, but I just can't figure out how to tackle this question.

Regards,

Bullstock

Dear Jose,

What you are saying is that a trader can go into trade if he is aware of the underlying market trend, then he can use your kits to exit with higer quality. I have visited your links and there seem to be somewhat complicated and a lot of choices too. There are some questions that I would like to ask.

Can you pinpoint it to one or two choices out of the kits?

Would MACDH Divergence be the best choices?

I am think of developing a trading system using Forum Latch with trailing stops. Can I combine the above mentioned high quality exit, e.g. MACDH Divergence sell signal for sell exit and similarly for buy exit, together with the trailing stops as a sa[censored]uard?

Regards,

Bullstock

Thanks Jose.

Well, with the understanding of limited knowledge in coding, I think I should look for trading system(s), readily available in the market, that ought to be profitable. Do you and any other experienced traders have any comment?

Bullstock

SMG, TradeSim cannot test intraday signals, but is worlds apart from MetaStock's System Tester.

Is there any third party software testing intraday data apart from MS EST?

Good luck in finding one. ;)

Got the feeling that it is not that easy to locate profitable trading system(s) that work with intraday data especially. Am I right? Could any experience MS user comment also?

Bullstock

Jim,

Thanks for your advice. Your post has raised some good questions to think about.

I would like to find out workable intraday trading system as I have found that some workable EOD systems won't work in intraday data. This makes me a it confused and don't know the reasons. Well, there are systems selling in the market and I just curious whether there are systems working in intraday data. I think there should be some trading systems working in intraday data, otherwise all buyers would be cheated. Therefore seek advices here in the forum. What I am saying now is to search and analyse the trading system for intraday data, but not testing them with real money.

Another issue is mechanical versus discretionary: this is a very good question! Discretionary would mean an introduction of human factor or experience if you prefer. For me, I tend to believe in mechanical trading with signals given by a system. This need to conduct blacktesting. A good blacktesting result that may give a better idea whether the system works or not. You may say the market may change due to the fuild nature of market. But, will you believe in a system that produced huge losses or consistent losses in the past!? A system with a good blacktesting result may mean that the system would likely work in the future and is free from human factors (including emotion).

I have read your posts elsewhere, knowing that you have bought some add-ons or plug-ins. I wonder if you have blacktesting them with EOD / intraday data. If you have, could you share the reults with us.

To me, blacktesting results are somewhat scientific to believe in! Someone may say a system worked perfectly yesterday that yielded a huge profit. But I would ask, what would be the yield if you use it for the past year. A single or a few profitable trades yielded from a system won't be give a solid evidence whether it works in the long run or not.

cheers

Bullstock