When to pull the plug on your system

Trading Discussion
#1

I wanted to start a discussion on how various traders here decide when their system of choice is no longer working and requires re-evaluation.

My personal thoughts on system development and evaluation:

  1. First start by backtesting or forward testing your set of rules over several trades. The number of trades you use to evaluate your system will depend on the time frame you trade in. Day traders, for example, might need several hundred trades before the sample size is large enough to cover several market conditions.

Assuming you are satisfied with results and have been trading it, how do you decide it is no longer working for you? (And please don’t say because you’re losing money). Although losing money is obviously the result which forces us to re-evaluate or dump a system, i am looking for specific and objective measures of evaluation.

My own idea involves looking at things mathematically. If, for example, my backtesting over the last 5 years shows a maximum drawdown of 20%, then i would start questioning my system if i fell into a drawdown of at least that much plus, say, 5-7% to account for errors during live trading.

Anyway, i think people can see what i’m getting at. Too many systems are prematurely dumped because of emotional factors rather than objective factors.

I’d love to hear people’s thoughts on this

#2

If you have thoroughly tested, backtested & paper traded
a system, which you are happy will show a profit, & are now
live trading it, there should be no reason to completely dismantle it.

Every trader should know that there system has it’s faults,
limitations etc. These would be found out from the testing
stage. The need then would be to fine tune the system to
extract as much profit as possible.

As an analogy, the soccer (or baseball) team manager who is
successful doesn’t completely change his squad for the new
season, he tries to find ways to compliment the players he has,
to fine tune the team.

The same is true of forex trading systems.

The time for dumping a system is before it is live traded,
not after.

:wishes:

#3

I agree with pretty much everything you said. Systems can always be tweaked and improved no matter what but systems that tested well don’t necessarily continue to work forever. Just because you tested the last 5 or 10 years that proved to be profitable doesn’t mean at all that the system will continue to work for the next 10. Markets change all the time and with those changes some things that used to work in the past won’t necessarily continue to work into the future. A system that was optimal may eventually become sub-optimal, no matter how much backtesting you have done.

My question is basically this: Although a well-tested system should make the trader fully aware of the system’s strengths and weaknesses, there could come a point when weaknesses start to outweigh the strenghts, especially if market conditions drastically change. I am not saying that a system would have to be completely dismantled but at what point should a trader begin to at least question the viability of his/her system of choice, despite the fact that it tested well and has proven profitable in the past? What are the alarms you use or consider to indicate that something could be off?

#4

I agree with pretty much everything you said. Systems can always be tweaked and improved no matter what but systems that tested well don’t necessarily continue to work forever. Just because you tested the last 5 or 10 years that proved to be profitable doesn’t mean at all that the system will continue to work for the next 10. Markets change all the time and with those changes some things that used to work in the past won’t necessarily continue to work into the future. A system that was optimal may eventually become sub-optimal, no matter how much backtesting you have done.

My question is basically this: Although a well-tested system should make the trader fully aware of the system’s strengths and weaknesses, there could come a point when weaknesses start to outweigh the strenghts, especially if market conditions drastically change. I am not saying that a system would have to be completely dismantled but at what point should a trader begin to at least question the viability of his/her system of choice, despite the fact that it tested well and has proven profitable in the past? What are the alarms you use or consider to indicate that something could be off?

5 to 10 years whoa!! I am a daytrader it’s difficult for me to see the next
5 to 10 days.

But seriously the market is an ever changing fluid instrument,
therefore we need to be the same as traders. I feel that too
many people flit from one system to the next, you see it everyday
on this forum, without a thought for testing etc.

The point at which a systems viabilty should come into question
is when it is not performing to specified parametres, ie not realising
as much profit as it did in the past.

I will now put IMHO, because I am waiting for somebody with a lot
more trading experience than myself to chime in.

Great thread though, thought provoking.

:wishes:

#5

My new system right now is not to have a system…It seems like if i make it simpler and use more Fundmental news then indicators, i do good.

I guess everyone has to go what works for them?

#6

Hmmm… really interesting question. You’re right Pipbull - it’s not about just when you’re losing money and I’ve not been trading long enough to be definitive about this, but my thoughts…

Your logic - (max drawdown plus a few % extra for non-perfect trades) makes perfect sense to me, but a lot must depend on why your system works as to what to look out for. Most systems I’ve looked at or thought of are basically ways of following trends, so it follows if the market is ranging, most systems are going to do badly. If you’re into swing trades, and enter into a long period of ranging, like '04-'05 for GBPUSD for example, you might experience high drawdowns. But if you’ve identified those weaknesses in back-testing, then at least you’ll know why when it’s happening.

Then at least the decision is not to drop or modify, but to step aside and work out when the an overall up or down trend will kick off.

Then there’s things like the sub-prime saga. For me and my strategy at least, it’s made it much harder to trade because of the violent swings - jittery in a word. But I wouldn’t change my strategy yet because there’s only so long people can be jittery for… he said, half hoping!

Best of luck

#7

I don’t mean to be flippant, but perhaps the simple answer is when you have found a better one?

This means doing the appropriate testing, yada, yada, yada. This of course coming from someone who doesn’t yet have a system… :wink:

#8

The thing is that there is always a better system out there. But “better” is a relative and subjective term depending on what satisfies [U]your [/U]performance expectations.

Ideally, we need to all find a system we are comfortable and confident in (through testing of course) and stick with it until it doesn’t work or needs modification. This is the ultimate exercise in discipline i find: to stick with a system for an extended period of time in spite of perhaps being bored with it.

If you are always looking for a better system then your search will never end and you might never get started. IN the beginning phases of the learning curve, it is best to find one system and stick with it. When you master one or become comfortablle enough with it, then you can add to your toolbox.

#9

I think this is a very thought provoking question and I think looking at the responses provokes the same response in all of us - We wish we knew! In reality we ditch models when we lose faith in them although I will put them on a demo platform to see if they come good again and then try to work out what happened. I think your method of a certain factor outside of historical norms is a sensible way to go

#10

The problem is ofcourse the human emotion, it can work in our advantage but it also can be a strong disadvantage.
What I am trying to say is when you are testing your system and it don’t work for you try letting the system test by someone you know?
In this case maybe some mistakes can get out of it and the system begins working again…