Deadline June

[Adamus]

The only alternative that I've seriously checked out at this point is MB Trading, who have more respect than IB in all those reviews. However I was totally unable to get hold of their historical data for backtesting.

They say they have 15 years worth of tickdata available but when I asked for it, offered to buy it, they said it wasn't possible. I think the salesman was talking rubbish.

 

[Adamus]

Found that optimising over a shorter period does give better results. 30 days appears to be best.

Testing optimising each MA of the 3 Ducks seperately. Playing with fire maybe. It doesn't work with short periods. I'll try a couple of long periods, e.g. 5 years.

 

[Rossini]

Optimising your optimisation... Lol.

 

[Adamus]

Hey, you were the one who suggested optimising the different MAs seperately.

When you put it like that, it sounds bad, but if I apply the same precautions as I do to primary optimisation, e.g. lots of blind tests / walkforwards, then I should be alright.

I have a lot of data to test it out on. I've got to have enough tests to decide on the best optimisation period and walk-forward period that works - and for reasons that make sense - and then I've got to make sure I keep enough data to try out the whole process from scratch on other data.

 

[Rossini]

Hey, you were the one who suggested optimising the different MAs seperately.

When you put it like that, it sounds bad, but if I apply the same precautions as I do to primary optimisation, e.g. lots of blind tests / walkforwards, then I should be alright.

I have a lot of data to test it out on. I've got to have enough tests to decide on the best optimisation period and walk-forward period that works - and for reasons that make sense - and then I've got to make sure I keep enough data to try out the whole process from scratch on other data.

And I stand by that. I just thought it a catchy line, I don't think it's a bad thing to do. As you say, as long as you have enough data then this should avoid over-fitting.

 

[Adamus]

And I thought "over-fitting" was when the man in the trousers dept takes a bit too long measuring your inside leg.

 

[Rossini]

Happens to me all the time...

 

[Adamus]

Trading Sessions

Just thought I can try applying trading times to my systems.

I'm watching the Forex Morning Trade crowd putting their punts on the GBP one way or the other at 06:30 - they've got a whole system based on the characteristics of a trading session.

There's also research about which pairs trade best when, and it actually goes into what day of the week is best too. It's all a bit questionable since the authors never discuss how they decided what their "trends" were - the test whether the trend which started in that period was profitable or not - by when or by how much was no clear. forexcheatsheet.pdf

I want to get the times together for each forex pair to restrict trading entries. Minimum should be one session. Max can be all sessions. Defining the sessions should be interesting.

Then when there is a significant performance increase when I apply the timetable restrictions, I'll can do that. If not then nothing lost. I'll post the pairs, sessions, and times here later.

 

[Adamus]

Checklist for good systems, v.1

I failed to notice that my unoptimised version of the 3Ducks loses money in the long term on long trades.

I never cease to surprise myself at my sloppiness. How did I fail to notice that it makes a loss on long trades? The result was sitting there in the middle of the screen, literally, when I found my "final" version of the system and started optimising it.

I'll need to keep it in the lab for a bitter longer to tinker with it and find out what happens with long trades. I re-optimised it for long trades only and short trades only, which is how I found out.

I need a checklist to go through when system building so I don't do this sort of thing again. My girlfriend is a virologist and has a written protocol for every experiment she carries out - I should do the same. I'm going to start a checklist here of all the points that have to be met when I think I've found something.

System Building Checklist v1

* Is the system profitable both long only and short only?
* Is the system profitable over 10 years?
* Is the system profitable on unseen periods long enough to be more than chance?
* Is the profit spread out across the period, i.e. is there a nice equity curve?
* Is max drawdown less than 50%?
* Is longest drawdown less than 18 months?
* Does it work across all major pairs?

That's a start. I need to define more where and when I mean optimised or unoptimised, and where I mean across a basket of instruments or per instrument.

Anyone with any suggestions please add them in.

 

[Adamus]

Can find everything except a way to make long trades profitable

I think I need to study some more tweakology because I am looking at these charts of long trades with the 3****s and I can see a hundred potential new systems but I can't see any way of making the long trades profitable.

re EUR-GBP: currently milling around 0.8900 and unable to stay above it. Fading the double zeros - anyone heard of that?

 

[Adamus]

Re: Checklist for good systems, v.1

System Building Checklist v2

* Is unoptimised system profitable both long only and short only?
* Is unoptimised system profitable over 10 years?
* Is unoptimised system profitable on unseen periods long enough to be more than chance?
* Is the profit spread out across the period, i.e. is there a nice equity curve?
* Is max drawdown less than 50% of total profit?
* Is longest drawdown less than 18 months?
* Does it work unoptimised across all major pairs?
* Are there any stupidly profitable trades, i.e. >5 x bigger profit than avg?
* Are there any stupidly long trades, i.e. >5 x longer than avg duration?
* Are there any massive losses?

 

[Adamus]

USD-CHF-long-only defies all attempts to make it profitable. The 3Ducks is whipsawed the whole backtest period with any combination of 3Ducks code I can think of.

I tried optimising the MAs seperately.

I tried entering only when it breaks through resistance, or resistance plus 10 pips, or just when it breaks through the top MA.

I tried exiting when it dips below the top MA, or the hourly MA, or support.

I tried using Lows touching the line, I tried specifying it had to be a whole OHLC on the other side of the line.

I tried shorter initial stops, I tried longer stops.

I tried time-based stops.

I even tried targets.

I even tried curve-fitting it.

It's the alter ego of GBP-USD, which never makes a loss however I try to apply the 3Ducks.

 

[meanreversion]

This is where your mechanical system becomes discretionary, as you get to choose the markets you trade.

USD/CHF is not nearly as liquid as EUR/CHF. Or, to put it another way, USD/CHF is not a "real" currency pair, in the same way that GBP/JPY is not.

Perhaps you should stick to the pairs which are the most actively traded, EUR/USD, GBP/USD, EUR/GBP, AUD/USD, USD/JPY, USD/CAD.. that's probably about it for the liquid ones.

 

[Adamus]

I didn't realise USD/CHF was so illiquid. I guess I should know this stuff. But I also don't think there's a correlation between liquidity and profitability - in backtesting at least.

I think there's a lot of scope for huge spikes knocking your system out of trades and appalling fills in the cross-currencies, but how much I'm not sure and whether it would destroy the profits I also don't know. But then I've actually had my worst fills on GBP/USD.

I'm considering trading stuff like GBP/AUD and AUD/JPY to balance the USD exposure.

Currently it's not so bad with the leverage I'm using but with 150,000 to 200,000 per trade the exposure is more obvious.

EUR/USD, GBP/USD and EUR/GBP is a great trio. Maybe I should concentrate on other systems and leave the 3Ducks as it is instead of constantly trying to improve stuff.

PS in fact the more I think about it the more I think that a system like the 3Ducks should work on a currency pair regardless of the technical details of the market. It might not be directly traded from e.g. AUD to CAD, but if AUDUSD moves a little one way and USDCAD moves a little the other way, those moves may not be tradeable , but the resulting move in AUDCAD could be.

 

[Adamus]

Seeds of a system

I'm looking at the currency markets opening up 22:15 Sunday London time after the G20 finance meeting produced an agreement (which sounds like hot air to me).

Some of the gaps on the cross currencies charts are gaping chasms! There must be a system in there somewhere.

I got caught on the wrong side of the NZD/USD, not because my system was long, which is was, but I tried to close all my positions on Friday evening and forgot that the kiwi stops trading on IB at 20:00, unlike everything else. If anyone else was my employer except me they would have fired me by now. So I lost 300 bucks on that one - well, I would have done if I closed it immediately. I put in a stop in the hope of trailing it down because I can't see that gap remaining unfilled.

But those gaps - cable gapped down by 10 points and AUD/USD gapped up by 80 points and on the GBP/AUD chart it looks like the gap is something like 110 points. Most of the crosses gapped 40 points.

There must be something in there that's tradeable. I guess the best strategy would be some option spread but there might be some kind of serial correlation in there. Worth investigating.

 

[brettus]

Re: Seeds of a system

I'm looking at the currency markets opening up 22:15 Sunday London time after the G20 finance meeting produced an agreement (which sounds like hot air to me).

Some of the gaps on the cross currencies charts are gaping chasms! There must be a system in there somewhere.

I got caught on the wrong side of the NZD/USD, not because my system was long, which is was, but I tried to close all my positions on Friday evening and forgot that the kiwi stops trading on IB at 20:00, unlike everything else. If anyone else was my employer except me they would have fired me by now. So I lost 300 bucks on that one - well, I would have done if I closed it immediately. I put in a stop in the hope of trailing it down because I can't see that gap remaining unfilled.

But those gaps - cable gapped down by 10 points and AUD/USD gapped up by 80 points and on the GBP/AUD chart it looks like the gap is something like 110 points. Most of the crosses gapped 40 points.

There must be something in there that's tradeable. I guess the best strategy would be some option spread but there might be some kind of serial correlation in there. Worth investigating.

The gap usually gets filled, which on EURUSD you can see has filled. You should test that.

Can i put things in perspective a little bit in terms of risk? The average hedge fund will make 11% per year on its money, and you would like to think that they have a better understanding on how to manage money and risk. I'm not totally sure of your risk in terms of percentage per trade, but rather intrigued. Risking 2% per trade can be suggested to be a comfortable amount of risk, as long as you dont have 10 trades on at once, because that is 20% risk of your account. On paper it might well within the drawdowns of the system, but psychologically damaging. How do you deal with it because you arent trading a small account, well larger that probably 95% of anyone on this forum?

 

[Adamus]

And the NZD/USD gap got filled, although my trailing stop was 10 points away when it rebounded sharply and took my stop out.

LTCM was a hedge fund. I don't assume that they knew what they were doing. Lehman Brothers - did they know what they were doing? So the average hedge fund is probably run by Beavis and Butthead.

But that's beside the point regarding risk. I think it depends on your perspective.

I started out having £100K and I spent half of it before I even put a trade on. So I realised at that point that I was kidding myself that I was a trader but never actually putting my money on the line. The systems I have would allow me to succeed unless I'm grossly unlucky. I use fixed fractional money management to work out what my leverage should be, following Ralph Vince. You work out what the worst trade is going to be (Ralph Vince takes what your worst trade was in backtesting which underestimates it).

The drawdowns can be much more than 20% per instrument, but hopefully not much more on the basket. Even so, with forex I can adapt my lot size down to the precise % that Vince's optimal f formula says it should be, and I can keep going through drawdowns until the lot size falls to the IB minimum.

But having said that, I just halved my leverage to what it will be when I introduce a second system. So optimal f requires cohones grandes, and I couldn't hack it. It'll be optimal f when my second system goes online though. Perhaps.

 

[DionysusToast]

You have time risk too. You have lots of options at this point, which is fantastic. You could learn how to trade. You could continue with automation. You could give up on 3 ducks or you could carry on with it.

Isn't it time to cast aside what you want to believe about the 3 ducks and start looking at the reason the 3 ducks exists ? It's really a vehicle for selling training courses. Surely there are better models out there.

 

[meanreversion]

I suppose my question would be "why do you believe an automated system should work"?

For example, many people look at support and resistance; it thus becomes self-fulfilling in some shape or form. Many people use TA on EUR/USD, so there might be some edge in that.

But what about EUR/ZAR - is it worthwhile using TA in that "pair"? After all, it is not a real currency pair, it's simply a function of EUR/USD and USD/ZAR.

Do you see where I'm going with this? Would you expect 3 Ducks to work on BRL/JPY? If yes, why?

 

[Adamus]

Peter, I feel you don't believe in the 3 Ducks. It's not too late, you too could still be saved. Close your eyes and imagine your aunty's living room wall, with the pink flock wallpaper. See above the mantelpiece? Those three ducks? They are lining up for a reason, trying to tell you something. Hallelujah!

As for USD/ZAR, there's no data out there for it which I can get my hands on so I wouldn't know if it works on that or not. I suspect it would though. As for the cross EUR/ZAR, I guess you missed what I added as a postscript here:

PS in fact the more I think about it the more I think that a system like the 3Ducks should work on a currency pair regardless of the technical details of the market. It might not be directly traded from e.g. AUD to CAD, but if AUDUSD moves a little one way and USDCAD moves a little the other way, those moves may not be tradeable , but the resulting move in AUDCAD could be.