The top chart is a TradeSim run of the main system I use and the bottom chart is actual performance from 2003. They look pretty similar if you can mentally adjust the Y axis.
There are a number of things that I will mention;
1. I trade more than this one system so comparison with only one system is probably not totally accurate.
2. I pull money out and put money back in. The lift in the current year is because I put more money in.
3. I have pretty much pulled most of the profits out for various frivolous pleasures - the TradeSim run did not pyramid profits. I pumped more money in early this year.
4. Discretion is designed into the system - every run I do on TradeSim will give a slightly different result.
I think that TradeSim is an excellent product. The data for the run above was output from AmiBroker (thanks Glenn for the code which is available on the Yahoo site). The system used for the test above was actually designed using Metastock. The software doesn't necessarily make us better traders.
I read Way of the Turtle and the last couple of pages crystallised the thought that has been lingering in the back of my mind for some time.
The best way to trade systems is to write, test, optimise, walk through individual trades and immerse oneself in the process to design a system that suits the individual.
The more confidence one has in the system the more likely it is to work.
Giving a system to someone, even with a lot of coaching, will probably only work if that person pulls apart the system, changes aspects of it, and puts it back together.
Many don't have the time, effort or capital to put into the process.
Enough lecturing,
stevo
5 comments:
Hi Stevo,
Do you do periodic re-optimisations for any of your systems?
Or is your main system pretty much still in the exact same form as when it was first traded in 2003?
Do you think there is a place for re-optimisation in system development or is it only required when the system stops working?
Nizar.
I don't do re-optimisations on any systems at this stage, so the main system hasn't changed since 2003.
I think that there is a place for re-opts in trading systems. It would need a bit of work to come up with a routine approach to this issue.
Such questions as how often, over what time period, what to optimise, what criteria to target for, what to do with currently open trades etc would make me reluctant to get too carried away.
Imagine having 3 versions of one system running at the same time. Old trades are still on version 1 - or do we use the new version 2/3 exit criteria?
I would suggest that each optimisation is really creating a different system that is trying to adapt to the latest market characteristics. But the latest market characteristics may change tomorrow.
I prefer to freeze changes on the current system and go and work on a new one that could be introduced when I am happy with it. I am running 3 systems currently, although the original system still dominates.
I think I would find it difficult to alter aspects of the original system since I didn't optimise them much in the first place. I spent quite a bit of time looking at how the exits and entry criteria could vary depending on the stock's characteristics. I also worked on an ON/OFF switch using an index filter.
Ideally the system adapts to the market over time, which is why I like things like bollinger bands and average true range with a multiplier that varies depending on the stock's character.
It is still an interesting idea - maybe I will do a minor re-optimisation every 5 years!
stevo
Hi Stevo.
Yes i agree, I found ATR based exits to work excellent on weekly charts.
Iv just bought a copy of Leon Wilson's breakthrough trading. He uses MS/Tradesim!!! Its nice when somebody speaks your language!
Anyway, i wanted to ask you, if you have some time, it would be great if you could comment on my last post on the Robustness thread over on ASF.
Thanks.
Nizar.
What you say about finding a system, pulling it apart and putting it back together again, with some changes, rings so true.
I'm too much of a sceptic to ever follow some lucky b@stards system over the edge of a cliff like a lemming.
Re: re-optimization, when I look at the parameter values I've selected in the system I'm working on most of them are really round numbers ie. 1/4, 1/2, whole years, 25%, 100% etc. I ran various optimizations of them in Amibroker to see if any particular settings were better than others and whether the ones I'd chosen were sitting on the edge of any cliffs in the 3d optimisation. I actually can't foresee a need to re-optimise.
ASX.G
"I actually can't foresee a need to re-optimise."
That said I understand why people do it. It seems Howard Bandy's book might have some good detail around this.
Post a Comment