How stable is any system?

[mathsfreak]

I have recently read that what worked in the past may not work in the future. With this in mind how stable do you think a typical system is? (or is there no such things as a typical system?)
Do you think it is a good idea to have a well tested backup system in case one doesnt work as well as it used to?
At what point do you stop using a system? I realise the importance of sticking with a system but if it used to consistently give you a 50% strike rate every month for a year and then suddenly it drops to 25% should you stay with it or switch to another but monitor the situation to see if it improves?

 

[TWI]

Use a money management system that will ensure if a system does not work the risk taekn by that system declines accordingly. when it starts to work again then the risk accordingly increases. This has to be done in the context of its tytpical win/loss profile. It is best to have a portfolio of systems that use different characteristics of price and timeframe.

 

[Rhody Trader]

The big issue with systems is what happens when market conditions change. Many trading systems are only back-tested (knowingly or not) over certain conditions, such as relatively low volatility. If volatility rises, as it seems to be doing across many markets, then systems which performed quite well under previous conditions might fall apart under the new ones. That's why you often hear the term "robust" when it comes to system design. Robust systems may not show results as good as more specifically refined ones, but are more likely not to suffer as badly when things change, and indeed may be able to ride them out.

 

[TWI]

It is not a difficult thing to adjust a system to cope with changes in historic volatility. All parameters can be calculated based on volatility or ATR or something similar. Definitely worth investicating this rather than sticking with rigid risk/reward irrespective of vols.
One thing that may change and really sc*ew systems over is a tranfer from breakout to mean reversion. It is no easy to have a system that can efficiently change from one to the other this is why have to have a nuber of systems and allow risk management to sort out the good performers from the ones that are failing.

 

[Glenn]

I have recently read that what worked in the past may not work in the future. With this in mind how stable do you think a typical system is? (or is there no such things as a typical system?)
Do you think it is a good idea to have a well tested backup system in case one doesnt work as well as it used to?
At what point do you stop using a system? I realise the importance of sticking with a system but if it used to consistently give you a 50% strike rate every month for a year and then suddenly it drops to 25% should you stay with it or switch to another but monitor the situation to see if it improves?

If any system does not account for the effects of Chaos theory, then it will only work for a time because it is imperfect and does not adapt imho. Hence the need for multiple systems.
If you can grasp the way that Chaos affects the tide, then you only need one 'system' and picking the right boat is pretty straightforward if you are not trading the tide itself. Sometimes merely a log will do.
Glenn

 

[hanzahar]

in my opinion there isn't any 1 system for all markets

u need system for trending,ranging and well flat market...

market 80% ranging so having a scalping system is really going to help

 

[firewalker99]

It is not a difficult thing to adjust a system to cope with changes in historic volatility. All parameters can be calculated based on volatility or ATR or something similar. Definitely worth investicating this rather than sticking with rigid risk/reward irrespective of vols.
One thing that may change and really sc*ew systems over is a tranfer from breakout to mean reversion. It is no easy to have a system that can efficiently change from one to the other this is why have to have a nuber of systems and allow risk management to sort out the good performers from the ones that are failing.

I agree adding a parameter like the ATR or the average daily range will help your system adopt to the volatility cycles which markets go through.

I'm not quite clear on your second comment however. Could you explain a bit more what it is you mean exactly by "a transfer from breakout to mean reversion"? Thank you!

 

[firewalker99]

in my opinion there isn't any 1 system for all markets

u need system for trending,ranging and well flat market...

market 80% ranging so having a scalping system is really going to help

My system works best in ranging days... as the majority of the days are ranging that helps me. However I need to be careful in trending days not to lose too much. That's why having an extra signal or filter to assist you in these days will greatly improve your system.

For instance, I look for trending days after an important support or resistance level has been broken.

 

[firewalker99]

I have recently read that what worked in the past may not work in the future. With this in mind how stable do you think a typical system is? (or is there no such things as a typical system?)
Do you think it is a good idea to have a well tested backup system in case one doesnt work as well as it used to?

I think you raise a good question there. It is one I have struggled with to when I first started to work on my system. How much time and effort do I have to put in, if my system will only work for a couple of months? Will it do as well in a bull market as in a bear market? Will I be able to tweak and adjust it to fit the ever changing market circumstances?

Now that I have been trading my system for a couple of months, I've been getting more confidence in it, but who knows if I'll still be profitable within a couple of years? Perhaps some more experienced traders can shed a light on this and tell us how long they have been trading their system, and how many adjustments they had to make in all those years. Be looking forward to some suggestions here.

 

[alexander]

I have recently read that what worked in the past may not work in the future.

This is exactly right and most system traders have expereinced that with a lot of pain.

With this in mind how stable do you think a typical system is? (or is there no such things as a typical system?)

I would use the word "robust" instead because if there is an issue of stability then you should not trade it in the first place. BUt this is technical detail. The answer is that whichever word you use to describe the same thing, it appears that ALL systems sooner or later will become victims of the law of large numbers and their performance will revert to zero mean minus commissions of course, which makes it a highly negative performance after some time.

Do you think it is a good idea to have a well tested backup system in case one doesnt work as well as it used to?

I wonder if you meant "back tested" or actually a tested backup system as an alternative to the one used.

At what point do you stop using a system? I realise the importance of sticking with a system but if it used to consistently give you a 50% strike rate every month for a year and then suddenly it drops to 25% should you stay with it or switch to another but monitor the situation to see if it improves?

Good question. At what point do you stop using a system? Very good question but maybe not the proper question to ask. The key question is at what point can you determine that a system does not behave as it was expected when designed. This can only be done using real money under real market conditions. You must collect data like:

1. success rate
2. average win, average loss and then get the ratio
3. profit factor
4. consecutive losers
5. max open equity drawdown
6. Sharpe ratio
7. etc.

Then compare to the tested system. If the results diverge widely for the worse then the system must be abandoned. At what point? I think after 30 trades or so to have some statistical significance. But that depends on many factors. YOu are alone in that like every one else. Good luck.

Alex