Deadline June

[meanreversion]

Ok, first of all, think about liquidity. EUR/ZAR is not liquid. You will cross huge bid/ask and slippage will be horrendous. For that reason alone, you should not attempt to trade it mechanically.

But beyond these issues, why should 3 Ducks work? What aspect of the market is it harnessing?

Do you think it should work for anything? What about gold denominated in JPY, XAU/JPY.. would it work for that? Why should it? Is it a magic system that works on anything that moves??

 

[meanreversion]

Or put it this way - what is it you believe about markets that means a system like 3 Ducks will be profitable over time?

 

[DionysusToast]

What aspect of the market is it harnessing?

Gullible newbies ?

Adamus - it's not that I don't like it - it's just that theres thousands of systems like this on the internet and they exist to enrich the sellers, not the customers.

Now - of course, 3 ducks is free - but if you REALLY want to add the magic dust that makes it work - you have to give Captain Currency $$$ for the real goods.

Just trade live for 3 months - small money in a CFD account. You will learn more in the first couple of weeks than you have in all the months you've been trying to automate. Then come back to automation armed with all this new found knowledge.

 

[Adamus]

Re: Trading Sessions

I'm going through the results of the 3Ducks backtesting for a five year period up to Mar 2010 to see whether any of the forex pairs have clear periods in the day when they trade well or trade badly.

forexcheatsheet.pdf outlines an analysis the author did on an unnamed and unmentioned system's results. He's selling something but it's not clear what. That is beside the point as the author does quite a good job of laying out his data and analysing it. I think it's probably from the same guy who flogs the FMT system.

As well as hour of day and day of week he analysed the length of his profitable trades and throws that into the system too. I guess it's one way of doing it but it's not exactly something for the 3Ducks-based system I'm looking to tweak this way. Neither is day of week.

The trading day can be divided up into 3 main sections by geography - Far East & Australasia, Europe and North America, with a slight compilcation that there is a big overlap between European PM and North American AM.

Dividing up the trading day in terms of volume gives 5 distinct sessions:

21:00 - 23:00 GMT a couple of hours in NZ before Sydney and Tokyo open
23:00 - 07:00 GMT main Far Eastern session
07:00 - 12:00 GMT London / Europe AM session
12:00 - 17:00 GMT London / Europe PM / NY AM session
17:00 - 21:00 GMT NY PM session

These session times are given in GMT. The author notes that the session times equate to the trading times of the local stock exchanges. Basically he's forgotten about daylight saving time - in summer, these session times all shift one hour forward. As I write, we have British Summer Time in London, and Daylight Saving Time in NY. The Japanese have JST all year so they don't shift - but London, EU & NY are out by 1.

From wikipedia we have these hours for the stock exchanges:

ASX GMT+10 DST 10:00 - 16:10
Tokyo GMT+9 09:00 - 15:00
European exchanges GMT+1 DST 09:00 - 17:00 (roughly)
London GMT+0 DST 08:00 - 16:30
NYSE GMT-5 DST 09:30 - 16:00

Change these to London times regardless of DST***:
Aus & Jap 00:00 - 06:00 (Tokyo is hour later in summer with no DST)
London & EU 08:00 - 16:00
NY 14:00 - 21:00

I should look for some info about forex intraday volume distribution but right now I don't have time. Spent far too long getting excited about the new 5kg bag of Columbian coffee beans I just opened. It's organic, fair trade, harvested by virgins on the slopes of Cotopaxi while llamas are sacrificed to the Coffee God.

*** Forget the God of Coffee - those crazy Americans don't leave DST until 7th Nov, while here in Eur we swop back a week earlier. As for DST in the southern hemisphere - let's just forget about that for now.

 

[Adamus]

Re: Trading Sessions

This is what I've got now, London times regardless of DST***:

(1) 22:00 - 00:00 NZ
(2) 00:00 - 06:00 (Tokyo is hour later in summer with no DST)
(3) 06:00 - 08:00 Limbo between Far East & Eur
(4) 08:00 - 14:00 London & EU
(5) 14:00 - 16:00 London, EU & NY
(6) 16:00 - 22:00 NY only

I need more info to define periods 3 and 5 better. They might be longer and they might vary due to DST changes between summer & winter.

Period 3 is probably bad for trading while period 5 is probably good.

 

[Adamus]

Ok, first of all, think about liquidity. EUR/ZAR is not liquid. You will cross huge bid/ask and slippage will be horrendous. For that reason alone, you should not attempt to trade it mechanically.

But beyond these issues, why should 3 Ducks work? What aspect of the market is it harnessing?

Do you think it should work for anything? What about gold denominated in JPY, XAU/JPY.. would it work for that? Why should it? Is it a magic system that works on anything that moves??

Or put it this way - what is it you believe about markets that means a system like 3 Ducks will be profitable over time?

The basic premise is that markets tend to trend - do I really need to say that, apart from to Dionysus T.? Apart from maybe OJ and deep frozen concentrated silk moth cocoons, lambs wool futures and pork bellies, (and I might be wrong there too), markets will trend from one state to another. In fact anything that has a futures market will trend and offers speculators opportunities and that includes currencies obviously. (NB I'm not trading futures, just 4X)

I did say, "regardless of the technical details of the market", by which I meant if it's actually only traded via two trades on other pairs, or has massive b/a spreads, or spikes etc, the latter being why I can't trade New Zealand dollar crosses or the Swedish Krona. Or ZAR in any form as you point out. But the long term trend followers who use Trading Blox for instance will trade anything they can get a position on. A hundred markets is their ideal basket size.

 

[Adamus]

Gullible newbies ?

Adamus - it's not that I don't like it - it's just that theres thousands of systems like this on the internet and they exist to enrich the sellers, not the customers.

Now - of course, 3 ducks is free - but if you REALLY want to add the magic dust that makes it work - you have to give Captain Currency $$$ for the real goods.

Just trade live for 3 months - small money in a CFD account. You will learn more in the first couple of weeks than you have in all the months you've been trying to automate. Then come back to automation armed with all this new found knowledge.

I have traded and continue to trade in a small way on a totally unsystematic and unmethodical way and it teaches me loads, yes you are right, and the results are insignificant to my PnL, and basically it's just fun. Most of what it teaches me is stuff I try to apply in my mechanical system building.

I've still got a huge list of ideas for systems to keep me busy for years, one of which involves building and using a book of bid and ask sizes, for instance.

I can program better than I can trade. Sometimes when I'm trading myself, I can't even hit the right button.

 

[DionysusToast]

The basic premise is that markets tend to trend - do I really need to say that, apart from to Dionysus T.? Apart from maybe OJ and deep frozen concentrated silk moth cocoons, lambs wool futures and pork bellies, (and I might be wrong there too), markets will trend from one state to another. In fact anything that has a futures market will trend and offers speculators opportunities and that includes currencies obviously. (NB I'm not trading futures, just 4X)

I did say, "regardless of the technical details of the market", by which I meant if it's actually only traded via two trades on other pairs, or has massive b/a spreads, or spikes etc, the latter being why I can't trade New Zealand dollar crosses or the Swedish Krona. Or ZAR in any form as you point out. But the long term trend followers who use Trading Blox for instance will trade anything they can get a position on. A hundred markets is their ideal basket size.

OK - so markets trend & they trend a certain percentage of the time historically.

We went through a little exercise on here a while back where we looked at a purely mechanical system applied to the DAX. Applied to the DAX, this system based on simple MACD crossovers made money from 2001-2010.

The reason this worked well over the test period is that there was very little chop on the weekly DAX. This is a good example of a market that would work PERFECTLY with a coin toss strategy with a trailing stop. The coin toss + trailing stop is effectively a strategy that will work with a market that trends more than it doesn't.

Now - if we put this MACD system on Citigroup 5 min charts it would suck money out on almost a daily basis (and no you can't reverse to suck money out of Citigroup).

So - there's a system that works on a market and we know the charactersitics of the market & why the system works.

I agree that markets trend but how often do the markets you are looking at trend ? Given this information, why is the 3 ducks an appropriate strategy to trade markets that trend this percentage of the time ?

 

[DionysusToast]

I have traded and continue to trade in a small way on a totally unsystematic and unmethodical way

The biggest leaps forward will be made when you trade in a formal manner. This means figuring out what you will do, doing it and logging the results to enable you to make the necessary adjutsments.

Anything else is really gambling...

 

[JRP2891]

Just trade live for 3 months - small money in a CFD account. You will learn more in the first couple of weeks than you have in all the months you've been trying to automate. Then come back to automation armed with all this new found knowledge.

Good advice.

 

[meanreversion]

I have traded and continue to trade in a small way on a totally unsystematic and unmethodical way and it teaches me loads, yes you are right, and the results are insignificant to my PnL, and basically it's just fun.

If it's teaching you anything, it should be that you should stop trading in this way?

The casino is more fun, there tend to be women around, some of whom will serve you a chilled alcoholic beverage if you ask nicely.

I'm not sure I buy into the notion that you can apply a trend model to anything, regardless of what people on Trading Blox do. In my mind, a market has to be sufficiently traded for a trend to develop - for example, people can see USD/JPY is in a downtrend, and trade accordingly. But EUR/ZAR? No-one looks at it.. it's not a real currency pair.. I really struggle to see the value in attempting to apply TA or a trend model to it.

Out of interest, how many trades are you/will you be doing per day, in total?

 

[Adamus]

I agree that markets trend but how often do the markets you are looking at trend ? Given this information, why is the 3 ducks an appropriate strategy to trade markets that trend this percentage of the time ?

Those questions are answered by backtesting. I don't need to know the exact strength of the correlation or the exact frequency of its occurrence. I just need to look at the PnL of a backtest using a trading platform like NinjaTrader, TradeStation or whatever.

The DAX example you gave is a perfect example, other types of trending like that are the same but just not as obvious. It's correlation between one event and a subsequent event, in this case between the DAX rising at one point in time and then subsequently rising more at a later point.

In some cases the correlation doesn't involve trending at all. It just involves two events of any description.

Everything in TA is supposed to follow the same type of correlation but obviously there are so many pitfalls for the delusional and the naive that it's easy to contest that such correlations don't exist. Like you say, it requires consistency of application and discipline.

 

[Adamus]

If it's teaching you anything, it should be that you should stop trading in this way?

The casino is more fun, there tend to be women around, some of whom will serve you a chilled alcoholic beverage if you ask nicely.

Yes but you have to go out for the night in Monte Carlo and mix with a bunch of snobs, but I can just sit here and trade

It's irrelevant really. It's not as if I'm going to blow my account.

I'm not sure I buy into the notion that you can apply a trend model to anything, regardless of what people on Trading Blox do. In my mind, a market has to be sufficiently traded for a trend to develop - for example, people can see USD/JPY is in a downtrend, and trade accordingly. But EUR/ZAR? No-one looks at it.. it's not a real currency pair.. I really struggle to see the value in attempting to apply TA or a trend model to it.

I don't think you're being totally objective with this liquidity issue. Sure it affects the cost of trading, but as long as there is a market there and you can get data for it in sufficient quantity and quality then it's worth looking at. For me it seems the question boils down to whether I should backtest these cross currencies that I have 2 years of data for. I've spent the day backtesting them and got some pretty encouraging results. I'm not going to start trading them yet though, not with the 3Ducks in the configuration I have it now. It's pretty much the same problem with half of these cross currencies as it is for USD/CHF which I mentioned earlier. You have got me thinking about dropping the USD/CHF into the same group as the cross currencies, and maybe promoting one or two of the cross currencies that do really well, a bit like a sports team and squad A and squad B.

Out of interest, how many trades are you/will you be doing per day, in total?

The 3Ducks permutation that I run at the moment does about 0.15 trades per day per instrument and I'm running it on 11 instruments, so approx 1.5 trades per day. Avg trade length = 3 days.

Edit actually that was just the Long side. The short side is similar and so doubles the basket trade frequency.

 

[Adamus]

Re: Trading Sessions

This is what I've got now, London times regardless of DST***:

(1) 22:00 - 00:00 NZ
(2) 00:00 - 06:00 (Tokyo is hour later in summer with no DST)
(3) 06:00 - 08:00 Limbo between Far East & Eur
(4) 08:00 - 14:00 London & EU
(5) 14:00 - 16:00 London, EU & NY
(6) 16:00 - 22:00 NY only

Test results possibly show it could work but not convincing enough to trade. Need more trades.

 

[meanreversion]

The 3Ducks permutation that I run at the moment does about 0.15 trades per day per instrument and I'm running it on 11 instruments, so approx 1.5 trades per day. Avg trade length = 3 days.

Edit actually that was just the Long side. The short side is similar and so doubles the basket trade frequency.

3 trades a day is about manageable and shouldn't involve too much bid/ask and/or commission.

Sorry I've lost track a little, are you trading it live now?

 

[Adamus]

3 trades a day is about manageable and shouldn't involve too much bid/ask and/or commission.

Sorry I've lost track a little, are you trading it live now?

Yes, it's live.

3 trades a day doesn't tell me much in the way of stats but then each to his own.

When I'm judging whether I can run a system profitably, I look at $/trade and if it's not significantly more than commission + spread, I keep looking.

 

[Adamus]

Checklist for good systems, v.3

System Building Checklist v3

* tailor entries and exits to envisioned trade lengths, frequency, hit rate etc and don't faff with bells and whistles if your basic system is a turd
* 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?

 

[meanreversion]

You'll want a CAGR/MDD of at least 0.8, add that to your list.

 

[Adamus]

You'll want a CAGR/MDD of at least 0.8, add that to your list.

Why 0.8 specifically?

 

[meanreversion]

It's a very arbitrary number. Typically you would like to get above 1, but 0.8 and higher should be manageable. I reckon 0.6 or so, you would struggle.. i.e. the drawdowns would be far too big relative to the amount you would expect to make. If you can find something above 2, then I want to be your friend