Can mechanical forex systems work?

[north5]

I'm in the sam eboat rnicholl. I managed to very skillfully lose all the money I made the other day - and the system has done very little recently (+5pips). So I'm going to bow out of the challenge until some confidence returns. I'm going to call it a draw.

I think I'm going to leave my present system as is for now and move on to something else. I have been playing with MetaTrader some more and starting to like it. In this playing I developed a system that would have made $750,000 on EURUSD over the last 18 months.

Clearly this is curve fitted to the bull run - but has got me thinking. I am going to try and develop 3 strategies in MT. One for a bull run, one for bear run and one for sideways. If I can find one that runs across various makets when they show those characteristics then I can just set the strategy and let a system do the finding of entry and exit points.

Is that possible - has anyone tried this?

 

[Vrothdar]

Sounds like you want to develop a trend following and mean reversion system and use the signals from one or the other based on market conditions. If you can develop two successful systems and then throw in a mechanical method for determining whether the market is trending or not and you'd probably have achieved what you're after.

Seeing as mean reversion and trend following systems are both within the realm,s of possibility I don't see why it shouldn't be possible for it to be done.

 

[Charlton]

Sounds like you want to develop a trend following and mean reversion system and use the signals from one or the other based on market conditions. If you can develop two successful systems and then throw in a mechanical method for determining whether the market is trending or not and you'd probably have achieved what you're after.

Seeing as mean reversion and trend following systems are both within the realm,s of possibility I don't see why it shouldn't be possible for it to be done.

Vrothdar

I think this is the correct approach and it is what I referred to in my post as an "overlay". It's a way of keeping the integrity of a programmed strategy that is proven to perform well in cerntain market conditions and running it alongside other proven strategies that work well in different market conditons. The overlay code simply determines if the conditions are most appropriate for one strategy rather than another - like a switch. It may well be based on an entirely separate set of indicators or timescales that are not specifically referenced in the individual underlying strategies , for example, the reversion or trend strategies.

It means that tweaking of an underlying strategy may be carried out, if desired, without affecting any other ones. Also if one strategy starts to fail e.g. because of changing market conditions it can always be switched off, leaving the other(s) to continue making money.

Charlton

 

[north5]

I'm not sure if I'm going backwards or forwards at the moment. Am I wasting time chasing a new shadow?

I think I'm getting somewhere for the idea on this new path. Just manually traced my idea through a few charts with differing conditions and timescales and it looks promising. I've had a few late nights (this plus young son) so I'm really tired and could be seeing illusions. It also means my brain is too fuddled to think through the code. I'm going to write the idea down and take a rest for today. It shouldn't be too difficult to achieve in MT though.

I think my old system still has merit - but as it only takes 20 pips at a time must be viewed more as an automated scalper. When it hits it seems to hit pretty big - then it seems to go nowhere for a while - and then it hits in again. The good point is that there still hasn't been any major drawdown periods or failures(yet). I still can't see how I could do it in MT though.


(one last thing - I reveresed my MT strategy and ran it across the fall in USDCAD and it worked really well again)

 

[rnicoll]

I'm in the sam eboat rnicholl. I managed to very skillfully lose all the money I made the other day - and the system has done very little recently (+5pips). So I'm going to bow out of the challenge until some confidence returns. I'm going to call it a draw.

I've got the paper trading version still running, but basically it looks like the market is having a nervous breakdown. I don't expect it to get back into swing before Monday, and quite possibly mid-August. Go away in May, come back in September?

On the other hand, I'm very tempted to trade futures by hand, over long term. It looks like people are panic buying/selling futures, and there's good profits to be made from the market getting its act together.

I think I'm going to leave my present system as is for now and move on to something else. I have been playing with MetaTrader some more and starting to like it. In this playing I developed a system that would have made $750,000 on EURUSD over the last 18 months.

On an investment of... ?

Yeah, one of the things I realised early on is that if you figure out how to describe most of your analysis as indicators, then indicator based strategy runners make a lot of sense.

Clearly this is curve fitted to the bull run - but has got me thinking. I am going to try and develop 3 strategies in MT. One for a bull run, one for bear run and one for sideways. If I can find one that runs across various makets when they show those characteristics then I can just set the strategy and let a system do the finding of entry and exit points.

Is that possible - has anyone tried this?

I believe it's an extremely common next step. My system is designed specifically towards when the market's going nowhere (non-trending/cyclic/sideways), so a bull/bear market turns it into mush. I'll swap you my sideways strategy for your bull strategy if you want?

Also worth mentioning; you shouldn't need separate bull/bear strategies, I'd generally expect a strategy to trade any trend it can find, up/down or otherwise. For Forex, anyway, where there shouldn't be any up/down bias over a long period of time.


Oh, have a look at forex futures, BTW. Early results on my sideways algorithm look promising (because futures tend to overreact to news, even more than the underlying currency does).

 

[rnicoll]

I'm not sure if I'm going backwards or forwards at the moment. Am I wasting time chasing a new shadow?

I think I'm getting somewhere for the idea on this new path. Just manually traced my idea through a few charts with differing conditions and timescales and it looks promising. I've had a few late nights (this plus young son) so I'm really tired and could be seeing illusions. It also means my brain is too fuddled to think through the code. I'm going to write the idea down and take a rest for today. It shouldn't be too difficult to achieve in MT though.

My advice; get a good few night's sleep, don't worry about this until Tuesday, the markets will still be here. Not that I'm listening to my own advice, but I'm trying!

I think my old system still has merit - but as it only takes 20 pips at a time must be viewed more as an automated scalper. When it hits it seems to hit pretty big - then it seems to go nowhere for a while - and then it hits in again. The good point is that there still hasn't been any major drawdown periods or failures(yet). I still can't see how I could do it in MT though.

Mine's definitely a scalper. Makes 10-15 pips on a good trade, can lose 40 on a bad trade, and just makes it up in good/bad ratio. On the other hand, it makes fantastic profits on backtesting...

(one last thing - I reveresed my MT strategy and ran it across the fall in USDCAD and it worked really well again)

Nice. Answer a theory for me, try EURGBP and USDCHF would ya?

 

[rnicoll]

It means that tweaking of an underlying strategy may be carried out, if desired, without affecting any other ones. Also if one strategy starts to fail e.g. because of changing market conditions it can always be switched off, leaving the other(s) to continue making money.

The mad science alternative to this would be to run all algorithms at once, with an interaction layer so that if one buys and one sells at the same time you don't trade, and go with the theory that the winning stragegies will make more than the losing strategies. Solves awkward decisions about what the market's going to be doing next

 

[north5]

Nice. Answer a theory for me, try EURGBP and USDCHF would ya?

I know I was going to leave this alone - but as you asked.....

The MT strategy has a risk ration of 1:2500 and starts with $10,000. (If I move risk to 1:500 then it makes about $4m !)

FYI for the EURUSD has 400+ trades and win rate of 65%.

EURGBP works well (50 trades - 75% success).

USD CHF doesn't work (13 trades - 38% success).

I haven't tried to optimise these - just ran with the parameters from EURUSD.


I suppose the first question I need to answer for this strategy though is whether it is more profitable than simply holding over the period.

Most important thing for me though I think is trying out my idea for this overlay.

Much work to be done - but I think I will wait until Monday at the earliest to do any of it.

 

[rnicoll]

I know I was going to leave this alone - but as you asked.....

The MT strategy has a risk ration of 1:2500 and starts with $10,000. (If I move risk to 1:500 then it makes about $4m !)

:-0

FYI for the EURUSD has 400+ trades and win rate of 65%.

EURGBP works well (50 trades - 75% success).

USD CHF doesn't work (13 trades - 38% success).

I haven't tried to optimise these - just ran with the parameters from EURUSD.

That makes a lot of sense, to be honest. Suggests EUR/GBP (bane of my strategy) is heavily trending, and so works well. Would have expected better results for USD/CHF, but possibly it's been switching between trending and non-trending rapidly.

I suppose the first question I need to answer for this strategy though is whether it is more profitable than simply holding over the period.

Most important thing for me though I think is trying out my idea for this overlay.

Much work to be done - but I think I will wait until Monday at the earliest to do any of it.

Second question (which is your overlay) is: Can I spot trending market vs a non-trending market?
Third question: Can I spot the market changeover from trending to non-trending?

 

[Market Wizard]

Thanks for the well wishes.

Do you have any opinion on the idea of cutting stake by 30% after a loss and increasing it by 20% after a win. TBH It's a new addition and I just plucked the idea and figures from thin air.

I use similar sort of MM rules in my automated systems. After a losing trade i have a set figure it will trade with(minimal amount), if the next one loses also, it performs a dummy trade and only goes into real money, increasing incrementally , after the first win on the dummy trade. I haven't attempted any sort of cumulative investments yet.

What is your experience increasing by 20% after a win?

 

[rnicoll]

I use similar sort of MM rules in my automated systems. After a losing trade i have a set figure it will trade with(minimal amount), if the next one loses also, it performs a dummy trade and only goes into real money, increasing incrementally , after the first win on the dummy trade. I haven't attempted any sort of cumulative investments yet.

What is your experience increasing by 20% after a win?

I haven't seen long strings of failing trades with my strategy, but profits tend to come and go over time. I might use the average profit from the last 3 trades to calculate the amount to invest...

Oh, and for the curious my first stab at a trending & non-trending market compatible strategy can be grabbed from: http://www.trade2win.com/boards/free-systems/35693-two-way-bands.html

 

[jason0181]

Hi

I have just read through the thread and I am very curious as the the outcome of mechanical 'blackbox' systems. I have purchased one this month, and is doing well. The benefit of this system is that it works on EOD, and has large pip gains for the ammount staked via the stop loss. Typical gains are 250+ pips on GBDUSD over 2-3 days depending on market activity.

Why I am interested in this thread is that I have a basic level of programming ability, and basic level of market knowledge and am looking to develop trading systems, but not sure if I could ever trust one of my own 100%. Have any of you looked at:

Trading Systems, Backtesting, Technical Analysis, Indicators

It looks like a very basic .NET application with good backtesting ability.

I admire you all for having faith in your systems, and posting them on collective2. Well done

 

[rnicoll]

Typical gains are 250+ pips on GBDUSD over 2-3 days depending on market activity.

I'd expect to get hit hard when the market decides to do its own thing, next, then. Not that this makes the system broken at all, just... be aware, week or longer bad periods happen, from what I've seen so far.

Have any of you looked at:

Trading Systems, Backtesting, Technical Analysis, Indicators

It looks like a very basic .NET application with good backtesting ability.

Seen it, not tried it. I like NinjaTrader myself, although the lack of anything that doesn't require Windows bugs me.

I admire you all for having faith in your systems, and posting them on collective2. Well done

Faith? Bah, it's there for comic effect

 

[rnicoll]

Just a quick update... and an indicator! Been dabbling around trying to make a trend-following strategy, and so far no luck. I seem to always make interesting new cyclic strategies. Also, sure I saw a 3,000% profit on the EUR/AUD on one test last night, but this morning it only comes out at a 1,500% profit.

If that sounds too good to be true, you'll be right; odds on it'll lose 4,000% next week Certainly, it's far too unrefined to let anywhere near a live account, anyway.

Going to get some more background reading done, see if I can find some better indicators to try. MACD doesn't seem terribly compatible with code, at least in its raw form, and Bollinger Bands are much better for cyclic than trending.

Oh, yes, and there's a NinjaTrader indicator attached, incase anyone wants to try making it stabler. Y'know how there's a bunch of technical analysis that's meant to be human only, like spotting S/R lines? Well, this searches for S/R areas, by looking at how much trading has been around the current price over the last "LongPeriod". Has a second parameter, "ShortPeriod", which defines how long it smooths the prices over. Oh, and it's set up for 1 minute bars at the moment, so ya may want to drop LongPeriod a bit if you're moving to anything a little more sensible in length! If anyone has any ideas how to make the output a little more stable, I'd love to know. Considering wrapping an oscillator around the output to give an idea of how much "Friction" it has compared to the last n time periods...

 

[north5]

I know I was going to wait to try out my idea - but I did some work on it today and results look promising to say the least......

I have run it on 6 markets and minimum success rate is 75% (max so far is 90%)!

OK so there is a bug somewher that means it is only going long. The stop an profit targets are not set where I want them (they will be based on market conditions for each trade and not set to one value) so the equity curve isn't as smooth as it should be - but these are just details I will fix tomorrow.

I really think I may have cracked this now.


(I'm sure when I look a bit more closely I will find the kick in the nuts that is bound to be there - but at the moment I can't see where that will come from).

more to follow......

 

[Vrothdar]

Good luck with that North. I hope you spend a long time checking it over and over again and come to the conclusion that there isn't anything wrong with it.

 

[rnicoll]

I know I was going to wait to try out my idea - but I did some work on it today and results look promising to say the least......

Addictive, isn't it.

No, that's not code you see in the other window. It's not! It's... err... porn, I swear!

I have run it on 6 markets and minimum success rate is 75% (max so far is 90%)!

OK so there is a bug somewher that means it is only going long. The stop an profit targets are not set where I want them (they will be based on market conditions for each trade and not set to one value) so the equity curve isn't as smooth as it should be - but these are just details I will fix tomorrow.

I really think I may have cracked this now.


(I'm sure when I look a bit more closely I will find the kick in the nuts that is bound to be there - but at the moment I can't see where that will come from).

more to follow......

As I'm working through this, I'm beginning to think that there's nothing wrong with any of the strategies we've had, so far. Think of them not as complete solutions to trading, but as a way of massively simplifying the process. The biggest problem is psychological; can you deal with the stress of trading something like this?

This is becoming considerably more obvious to me, as my strategy is showing signs of good currency pairs becoming unprofitable, and less profitable ones becoming profitable, and I can see hints of why this happens.

This is not to say there isn't one solution to it all, merely that it may be considerably simpler to do most of the heavy lifting by computer, and let the human tweak the last few bits here and there.

I've got a MACD-based bounce trader now, as well, which looks interesting. High drawdowns, but that can be fixed. Also, it's in the market continually, which is different, and means that instead of paying comission on market entry and then exit, position reversal can be done with a single trade. Am going to keep working on other ideas for now, but it is, as I said, interesting.

Still no luck with trending strategies. I'm sure these were meant to be simpler :-D

Results from MACD trader test attached. All trades done without taking into account margin, and NinjaTrader does two trades to go in and then out of a market, so considerably overestimates commission.

One last thing; what would people like to see on the blog next?

 

[Vrothdar]

The biggest problem is psychological; can you deal with the stress of trading something like this?

Bang on the money. The way I plan to tackle this is by using both diversification of currency pairs and trading systems. Unfortunately I doubt very much that I'll have enough capital, initially, to be as diverse as I'd like. Ho hum.

 

[rnicoll]

Bang on the money. The way I plan to tackle this is by using both diversification of currency pairs and trading systems. Unfortunately I doubt very much that I'll have enough capital, initially, to be as diverse as I'd like. Ho hum.

I'm planning on having my flatmates restrain me by covering a wall in blu-tack and sticking me to it.

It's not a good plan, but dangit it's my plan.

Edit: Good news, my system auto-submits orders to Collective 2. Bad news; that was a test, not a live trade.

 

[north5]

OK so there was a kick in the nuts - so it doesn't work as an automated system. It does however make a nice indicator for me to judge if we are in a trend or not.

I think I'm going to take my trending system and have versions that I apply. So I may think we have a bull run on the hourlies and so set that system on, or a bear run on the dailies and so set that system on.

Yes there will be a good bit of judgement call involved but I've driven myself mad trying to find the automated way to make thiese calls and just can't do it. I think my time would be better spent thinking about how the markets are currently behaving.


This may not however be such a bad thing. So I'm going to optimise the strategy I have on various timescales and then sit back and take a startegic view of the markets and let the system do the tactical stuff of actually buying and selling.