How do you know when a system stops working?

[BadgerMatt]

I got interested in trading about 6 years ago. I went on a few courses and quite quickly started spreadbetting, generally using a system which I paid someone to teach me. I found out quite a lot, but barely did any of my own research. I actually made some money for about 6 months and I didn’t lose it. But I gave up and have barely traded at all for about 3-4 years. I couldn’t have told you why at the time, but I now realise that I wasn’t comfortable with the risks I was taking. I was risking about 2% of my fund per trade, all of which I could afford to lose. But I really didn’t want to lose it (who does?) and got jittery whenever I got about 3 losses in a row. The system may or may not have been a good one (indeed my limited track record suggests that it was profitable one), but I didn’t have faith in it. And why should I? I had just taken someone’s word for it. I now realise that in order to have complete confidence in a system you have to know it inside out by at least thoroughly back and forward testing it.

My interest in the markets has been reignited recently and I have been avidly reading these forums for several months. And I am planning to start trading again. But this time I am going to write up a full trading plan and take responsibility for the methodology I use. I am going to spend the next few months forward and back testing several methodologies (including the one I was taught) before deciding on one to trade with for real.

But..... I am rapidly reach the conclusion it is not like a roulette wheel where I can work out what the odds are of any particular strategy working, and sticking with the one which averages out and makes you the most money. Market conditions change and past performance is not an indicator of the future. I could find a great strategy which consistently have a 1:10 risk: reward ratio and a 10:1 win/loss rate over a period of years, but who is to say that it will continue to do so? Unlike a roulette wheel or a fair dice you can’t guarantee that those rates will continue.

So I conclude that the only way to be a successful trader is to focus on managing your risk. As many people on this forum have said you don’t necessarily have to win on a majority of your trades in order to make money. That’s not to say that I plan to pick my entries randomly. As I said I plan to thoroughly backtest and forward test methodologies and use one which has the best track record. Unlike before I need to be able to trust it in order to get through the losing streaks.

So here is my question.... I know that I will need to take losses along the way, and all successful systems will involve this. But no matter how much faith I have, I can’t rule out the possibility that at some point the market will fundamentally change and the success probabilities for my system in the past no long hold true. But how can I identify this? I wonder whether there is a way of working out statistically how many losing trades I might reasonably expect to get within that range of probabilities? Does anyone do this, and if so how?

I’d be grateful for any thought anyone has. Apologies if any of this sounds incoherent - it’s the first time I’ve started to put my thoughts together on paper, which has been useful in itself.

 

[the hare]

it might be worth taking a look at statistical resampling techniques. David Aaronsons Evidence Based Technical Analysis is well worth a read in this regard.

 

[Rhody Trader]

So here is my question.... I know that I will need to take losses along the way, and all successful systems will involve this. But no matter how much faith I have, I can’t rule out the possibility that at some point the market will fundamentally change and the success probabilities for my system in the past no long hold true. But how can I identify this? I wonder whether there is a way of working out statistically how many losing trades I might reasonably expect to get within that range of probabilities? Does anyone do this, and if so how?

The two things I would look at to determine whether a system is still functioning is 1) whether the market is still producing the types of moves the system is meant to trade, and 2) whether the system is still getting in on those trades. If the market has changed, then you need to find a new approach. If the system isn't getting the trades, then you need to find a new system - or adjust the one you've got.

 

[bbmac]

I would refer you to this thread. http://www.trade2win.com/boards/first-steps/54890-why-high-probability-trading-so-important.html

You are right to develop your own trading edge, only then can you have confidence to see you through the bad times as you put it. Know it's historical strike rate (winning trades as a % of total trades) across a large sample and know therefore whether any losing runs fall within it's typical distribtion. The table below gives the most probable maximum consecutive losing runs at the strike rates listed...The typical consec losing runs will of course be less and typically half of the likely maximum...in knowing this you can set your risk accordingly to what you are comfortable with, should a drawdown ocur.

Remember too that having a trading edge is not the same as executing that trading edge such that delivers in a real a/c what it did over a back test or demo period...this is where you will find the battle lies, as you have already alluded too in your opening post.

As for the question posed in the thread title How do you know when a system stops working (assuming you are executing it with impunity) the simple answer is when it does not perform over a sample as it did typically across pervious such samples. Allow yourself a margin either way of those typical returns expected. Also be alert to the distribution and alert to anything out of the normal distrinbution as described above.

G/L

 

[leonarda]

I got interested in trading about 6 years ago. I went on a few courses and quite quickly started spreadbetting, generally using a system which I paid someone to teach me. I found out quite a lot, but barely did any of my own research. I actually made some money for about 6 months and I didn’t lose it. But I gave up and have barely traded at all for about 3-4 years. I couldn’t have told you why at the time, but I now realise that I wasn’t comfortable with the risks I was taking. I was risking about 2% of my fund per trade, all of which I could afford to lose. But I really didn’t want to lose it (who does?) and got jittery whenever I got about 3 losses in a row. The system may or may not have been a good one (indeed my limited track record suggests that it was profitable one), but I didn’t have faith in it. And why should I? I had just taken someone’s word for it. I now realise that in order to have complete confidence in a system you have to know it inside out by at least thoroughly back and forward testing it.

My interest in the markets has been reignited recently and I have been avidly reading these forums for several months. And I am planning to start trading again. But this time I am going to write up a full trading plan and take responsibility for the methodology I use. I am going to spend the next few months forward and back testing several methodologies (including the one I was taught) before deciding on one to trade with for real.

But..... I am rapidly reach the conclusion it is not like a roulette wheel where I can work out what the odds are of any particular strategy working, and sticking with the one which averages out and makes you the most money. Market conditions change and past performance is not an indicator of the future. I could find a great strategy which consistently have a 1:10 risk: reward ratio and a 10:1 win/loss rate over a period of years, but who is to say that it will continue to do so? Unlike a roulette wheel or a fair dice you can’t guarantee that those rates will continue.

So I conclude that the only way to be a successful trader is to focus on managing your risk. As many people on this forum have said you don’t necessarily have to win on a majority of your trades in order to make money. That’s not to say that I plan to pick my entries randomly. As I said I plan to thoroughly backtest and forward test methodologies and use one which has the best track record. Unlike before I need to be able to trust it in order to get through the losing streaks.

So here is my question.... I know that I will need to take losses along the way, and all successful systems will involve this. But no matter how much faith I have, I can’t rule out the possibility that at some point the market will fundamentally change and the success probabilities for my system in the past no long hold true. But how can I identify this? I wonder whether there is a way of working out statistically how many losing trades I might reasonably expect to get within that range of probabilities? Does anyone do this, and if so how?

I’d be grateful for any thought anyone has. Apologies if any of this sounds incoherent - it’s the first time I’ve started to put my thoughts together on paper, which has been useful in itself.

I look at it in a simple manner, quite simply if you "know why your system works", then you should know "when it won't"..!
Importantly, knowing why your system works, isn't for example "because my backtest said it should", for obvious reasons!
For example if your trading has some fundamentals based rules, then presumably you understand why those fundamentals give you the edge, and when the markets change such that fundamental edge isn't enough.

 

[BadgerMatt]

Many thanks all. Some very useful advice. I am thinking that I am going to have to get to know my system much better than I had appreciated. I have previously thought about it terms of having a clear set of rules which answers all of the "what if..." questions. But I am starting to realise that I also need to have an understanding to know when the market conditions have changed so as to make it no longer 'work'. I am not planning to use fundamentals at all, so I guess that means I'll need to be able to recognise when the charts start doing something they haven't done before.