Optimized Trend Trading

The concern about replication makes sense. I'm all for continuing the discussions, but I don't see ho you do that in a public forum without revealing things you would prefer not to reveal.

If you transform into a private forum somehow, please let me know.

Sincerely,
Ngawang

So you are basically claiming that spreads went up because of this particular EA...?

I can appreciate the concern over replication, but must admit to some skepticism that an EA (even with thousands running) could possibly influence large market makers. Yes, I could see some retail ops changing their spreads or spinning the odds but brokers like Dukascopy? I mean, how many EAs will there be running there when you need a minimum of 50k to even open an account?

I'd certainly like to express my interest in continuing the conversation -- I'd even be willing to provide a secure forum to do this if there is interest.

PM me if you're interested.

.. and also on Dukascopy which is an ECN?


And thanks for that, I had totally forgotten Parrondos Paradox although I was quite intrigued about it at some years ago (not because of trading but something else).

To anyone who's sceptical about my FapTurbo spread hypothesis, wondering if I should be using an 'i' instead of an 'a' in my name, and thinking it all sounds a bit X-Files, I started off with that same scepticism myself - I invite you to research the issue then by all means present me with the arguments against -

the workings of the market mean cause and effect are rarely clear cut; there's a degree of feedback and overlap in things, not to mention plain ol' coincidence - I'm happy to be proven wrong.


Just a few of the numerous things to look at would be:

-FapTurbo's evolution (and related EAs, e.g. Abeiks 3rd place win in ATC 2008).

[On a sidenote, each year for the past couple of years I've bought any EAs I can persuade the best contestants to sell - with the purpose not of trading them, but to study the algorithms they're based on to see what I can incorporate into my own modular voting based system. As was the case last year, the best EAs don't use any technical indicators whatsoever - purely price action and what the current bar's OHLC is relative to x bars ago. This year I've paid $12k+ for that info - hopefully it's of use to you ;-) ]

-the historical performance of the algorithms those EAs are based on.

-the traits of the ccy pairs those algorithms work most profitably on compared to other pairs.

-the ccy pairs affected by spread widening with both ECN and non ECN brokerages, and the relative changes in spread.

-macroeconomic issues - mapping key events to market behaviour, asking the question - why would overnight trading volume/liquidity change ?

-the idea that if you worked at Renaissance, getting paid $300k+ a year with your sole job description being 'research interesting new ways of trading anything', wouldn't it be criminally negligent if you _didn't_ study the algorithms used in FapTurbo or browse forums like this, and look at how you could apply them to a fraction of that entity's huge trading capital, given the promising near linear equity curve and excellent mid-long term track record ?

I'm confident that this site and others are visited (but never posted on) by such folks occasionally - once their interest is aroused enough to get a PoC up and running, a 2-3 man quants project and a chunk of cloud computing is approved, and the game's on its way to changing...

...that's not pure conjecture.

A few years ago I jumped pond from central to investment banking, coding trade matching, risk pricing and hedging algorithms for derivatives in what at the time was the largest IB in the UK, with access to the futures and credit derivs order book. A very stressful but colourful world, with a young prince working there alongside rough cockneys, the best traders allowed to keep 50% of their winnings, bringing in 1% of the entire bank's profits each, and hiring Page 3 girls as assistants...but that's another story !...the point being that I was amazed at the variety of research going on there, and how much leeway successful traders had to place significant trades in the market based on any idea that appealed to them.

The nuances here matter - I'm not claiming that Dukascopy deliberately widened its spreads as a direct result of FapTurbo, but possibly as a knock-on impact of those style trades taking off - I spoke to them at length at the time about this issue, but won't try and quote that verbatim six months later - obviously, it's in Dukascopy's own interests to keep spreads as tight as possible given its business model.

I'm off back to the comfort of the shadows, and my study of historical TP/SL risk reward ratios...I wish everyone the best of luck in creating a successful trading system based on the ideas set out in the threads here and elsewhere in this forum - to those PMing me, please be patient, my inbox is filling rapidly but I will reply to everyone in time before I retire this incarnation of my profile at the end of April.

kind regards

Cretan.


PS one last titbit before I'm off - consistent losers are just as valuable as consistent winners in any system !

Cretan, I completely agree that if an EA sold to the public really has "something in it", then it would be torn apart and analysed in details as soon as the news hit the right people. Speaking for my self, I would do whatever it takes to crack it if I were to believe it would worth the effort. If enough people (including institutional traders) started to trade it, the inefficiency it exploits would most likely vanish or at least decay.

But for a strategy to have such detrimental effect on the market, that the end result is significantly and permanently decreased liquidity in the "real" spot market, now that is something else...

We are in a general liquidity crisis of historical proportions, so that might be the first place to look for a bit more plausible explanation.

Guys

Please let us NOT get sidetracked onto liquidity. It is largely irrelevant to the way the systems described in this thread work. We are not scalpers. 2 pips or 3 pips spread is neither here nor there IMHO.

Attached results on my No. 2 account for 3 weeks to 24 March 2009. Please note that this account trades ONLY 2 pairs, EURUSD and EURJPY. The lot size is fixed at 0.1 for nearly the whole history.

Then on the last day I got bored and wound up the lots to 0.5 just out of curiosity (malice?). Didn't phase the system too much.

So up 15% in 3 weeks on 2 pairs with basically £1 a pip MM on a £10K account.

CB

Right, I have something better. These equity curves are from the same strategy too! Amazing no?

I can show you whatever kind of equity curve you want to see. That is not the real beef. It is silly when you post equity curves just to get attention and watch people salivate.

An equity curve coming from thin air is not getting back to the point of the thread which after 30 pages doesn't really seem to have much of a practical point. Some silly DSP posts here and there much of which are related to Ehler's and some talk about dynamic time frames. That is all it takes to make 10-20% a month with virtually no draw down. I don't know why the hell hedge funds and investment banks with a ton of top knotch brain power struggle to make 20% a YEAR and are going belly up at record pace.

Please share something constructive and deep. The rest is nonsense.

Guys,

I think that it is time to take Cretan's advice and "go back to the comfort of the shadows". This thread has served its purpose. As around 30 folk have found, there is everything here you need to put together a working system IF you are prepared to put in the effort. This last proviso is important as it keeps the bottom-feeders away.

The thread will now essentially be "underground". All those who have corresponded with me privately know who they are and have my private email. I want to focus on some serious trading, especially now that I am about to run a significantly larger account than before.

This thread has received some very complimentary comments and I would like to thank the serious contributors for their efforts. I may make the occasional post if the mood takes me, otherwise I will be available by private email.

CB

Thanks for the thread CB and have fun with the Parrondo stuff, I'm sure you're already working on this.

CB,

I totally agree with this move...

I have yet to obtain your private email but have been following this thread with much interest! I feel like I could be of assistance with this search for finding a consistently profitable trading strategy in regards to the communication between Matlab and MetaTrader.

Please PM me if you agree...


Thanks, Brent

All the best with your trading!

Please drop a note here after you have traded your system for a year or something, I'm really interested to hear how the actual trading results are after a longer period of live trading.

Excellent thread guys. I hope I'm not too late to the party. If things are being taken offline I'd like to be added to the list. I just spent the weekend furiously reading the whole thread from start to finish. I've been reading books on signal processing and attempting to get my math and concept understanding up to the level where I can implement this stuff. Is there light at the end of the tunnel?

I would like to attempt to go through all the steps outlined but the document link regarding how to run the Dickey Fuller test below (referenced in post #67) doesn't seem to work any longer. Would some kind soul re-upload it or send me a PM and email it to me? Thanks!

Link for post #67:
http://www.forexfactory.com/showpost...3&postcount=67

and broken link containing document example:
http://foquant.com/wp-content/upload...er-example.doc

CB, thanks so much for your willingness to share, and for the time and effort you had put into this thread! All the best to you in your endeaver (looks like you've got a winner )! Hope you'll drop by and give us an update on your progress in the future!

I've been reading this tread with profound interest (but I never posted until now) and following your advice I've been testing the waters with DSP, hoping eventually to create a profitable EA. I have a background in finance but I'm new to DSP so in the past few weeks I've been reading some books on DSP. Would you mind sending me your private email to my PM so that in the event this site breaks down (god forbids), I will still able to contact you and ask you questions?

Once again, CB, your help on this thread is very much appreciated.

Forzagrifo

Hello everybody,

Here is a screenschot of Goertzel cycles generated on non stationary CHIRP signal compared to cycles generated by NOXA CSSA. It seems that Goertzel is not able to resolve this signal without lag. Any ideas how to overcome this ??

All paratemeters like width of the window, number of cycles etc. are chosen by Genetic Algorithm to have a maximum profit of the strategy but I don't think it matters in this case.

Of course it works fine for normal sin waves.

Krzysztof

Of course that here is a missing link Look for this signal for example. It has 2 cyclical components 20 and 100 bars. So S/N ratio will be (A20+A100)/others in simple words. But when you remove faster component and we will have only one sin S/N will fall but signal will be more tradable because of better curve !!!

So you can have congested non tradable market with a lot of cyclical components and high S/N ratio.

So i'm not 100% sure if S/N ratio defined like in this thread is much usefull as it can give mixed results.

Krzysztof

Hi Krzysztof,

Thanks for those observations. I haven't found anything better yet but am definitely interested in variations on the theme that might do a better job.

Thx
Ngawang

Hi...this TopCat of yours seems great...how to get them?

Hi Ngawang,

Do you have maybe printout or mat file for calculation the lag response of SNR ??

For example. We have cycle 100bars and window 200 bars than cycles are stopping and we are starting the trend. What would be the lag/delay till SNR will show 0 ??

Now i'm in process of coding SNR as Neuroshell dll so perhaps will know more about usability of this in a few days.

Krzysztof

Hi betrader,

It is very easy to get TopCat. You have two choices:

(1) Send me a banker's draft for $100million and when it has cleared my account I will email you the .exe file.

(2) Read this thread, learn about the methods, and write your own version.

CB