Systematic trading | Page 44 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Trade another day through Systematic Trading 203 replies

Nonparametric Association Measures in Systematic Trading 57 replies

Anyone use Systematic from Alpari? 1 reply

A Systematic Approach to Markets 68 replies

Systematic Trading vs. Discretionary Trading 14 replies

Attachments: Systematic trading
Exit Attachments

Systematic trading

  • #861
  • May 28, 2010 2:01am May 28, 2010 2:01am
Quoting surfeur
Need a very good edge for find 10000 good trades/day

You need a big bank account to do 10000 trades a day. Just imagine the margin.
  • #862
  • May 28, 2010 2:07am May 28, 2010 2:07am
Quoting spekkiefx
You need a big bank account to do 10000 trades a day. Just imagine the margin.
Not really. Just do the trades high freq enough.
  • #863
  • May 28, 2010 3:40am May 28, 2010 3:40am
Quoting mikkom
Not really. Just do the trades high freq enough.
I have done much analysis with strategies on the tick data. The frequency per strategy is also limited when you want profitable systems with high expectancy. This means strategies need to run simultaneously having a direct impact on margin. (Although you can somewhat offset this by taking positions opposite a certain currency)
  • #864
  • May 28, 2010 4:03am May 28, 2010 4:03am
Quoting spekkiefx
I have done much analysis with strategies on the tick data. The frequency per strategy is also limited when you want profitable systems with high expectancy. This means strategies need to run simultaneously having a direct impact on margin. (Although you can somewhat offset this by taking positions opposite a certain currency)
That goes directly to correlation. If you have multiple correlated systems that don't close within really short period then this is indeed the case.

But this is not the kind of trading the paper was talking about.
  • #865
  • May 28, 2010 4:40am May 28, 2010 4:40am
I am just wondering ,how you guys trading?I mean is your system based on purely TA? If yes ,have you done some special study about digital signal processing or regression analysis(nth order polynomials etc..) for your indicators ?
  • #866
  • May 28, 2010 5:21am May 28, 2010 5:21am
Quoting purley
I am just wondering ,how you guys trading?I mean is your system based on purely TA? If yes ,have you done some special study about digital signal processing or regression analysis(nth order polynomials etc..) for your indicators ?
I am working with non-linear time frames for my higher frequency system. But it's not yet production ready.

I don't do any pattern type TA (H&S, rising wedge, triangle,...), nor use trendlines or resistance/support, since they are hard to define in an objective way, but I do use a combination of indicators. I'm also working on trying to extract the math behind popular indicators, for instance what does it mean in a mathematical way when RSI is overbought? I've searched for quant versions of popular indicators. I've found some papers about generalizations of EMA, but noting for RSI yet. I'm sure there is some quant version of RSI, but I've just not hit the right keyword yet. The same for MACD or stochastics.

On the DSP side I've read a lot about the Hilbert-Huang transform, but I'm not sure it's applicable to trading.
  • #867
  • May 28, 2010 5:42am May 28, 2010 5:42am
Quoting Adal
I don't do any pattern type TA (H&S, rising wedge, triangle,...), nor use trendlines or resistance/support, since they are hard to define in an objective way, [..]
Tom Demark has written some good stuff, not saying there is anything relevant there (or that there is not) but he has written a lot related to above.

btw, what are quant versions of indicators?
  • #868
  • May 28, 2010 5:44am May 28, 2010 5:44am
Quoting mikkom
This might be of interest to some of people here..

http://www.tradeworx.com/TWX-SEC-2010.pdf
Nice share - and quite informative. Some of the myths surrounding HFT needs to be debunked...

For those who are interested there is an increasingly large list of comments to the SEC proposals here:

http://sec.gov/comments/s7-02-10/s70210.shtml
  • #869
  • May 28, 2010 6:09am May 28, 2010 6:09am
Quoting spekkiefx
I have done much analysis with strategies on the tick data. The frequency per strategy is also limited when you want profitable systems with high expectancy. This means strategies need to run simultaneously having a direct impact on margin. (Although you can somewhat offset this by taking positions opposite a certain currency)
As Craig pointed out, the primary advantage of HFT is the ability to trade low-expectancy systems (and thus exploit smaller edges) combined with low variance in the return distribution due to the high frequency of trades and short holding periods (lower risk). This in turn enables a much more efficient use of margin - which is also one of the main reason why most HFT is done by proprietary traders and not funds.
  • #870
  • May 28, 2010 7:16am May 28, 2010 7:16am
Quoting mikkom
Tom Demark has written some good stuff, not saying there is anything relevant there (or that there is not) but he has written a lot related to above.

btw, what are quant versions of indicators?
Mikkom ,
Search for Don K. Mak ,
it is a heavy reading for my taste but still enjoyed it : http://www.amazon.com/Mathematical-T...5055814&sr=1-3
  • #871
  • May 28, 2010 7:36am May 28, 2010 7:36am
Quoting Adal
I am working with non-linear time frames for my higher frequency system. But it's not yet production ready.

I don't do any pattern type TA (H&S, rising wedge, triangle,...), nor use trendlines or resistance/support, since they are hard to define in an objective way, but I do use a combination of indicators. I'm also working on trying to extract the math behind popular indicators, for instance what does it mean in a mathematical way when RSI is overbought? I've searched for quant versions of popular indicators. I've found some papers about generalizations...
I cant' remember where but I read that Hilbert-Huang transform is too slow for trading.Regarding oscillators I read some interesting stuff about scaling.
  • #872
  • May 28, 2010 9:18am May 28, 2010 9:18am
Quoting Adal
I am working with non-linear time frames for my higher frequency system. But it's not yet production ready.


On the DSP side I've read a lot about the Hilbert-Huang transform, but I'm not sure it's applicable to trading.
I spent quite a bit of time researching the HH transform and EMD. There is code out there on the net for EMD mostly for matlab. There are three main drawbacks that I noted. First, to make it workable as a trading function an online version would have to be created whereas most of the code samples deal with static data - and applying a static EMD to dynamic data produces a jumpy effect - like a redrawing indicator. I never got around to creating the online version due to the next two drawbacks.

Second and perhaps most importantly, the premise of the HH transform is that it boils down noisy data into its main component trend - a single trend. Unfortunately market data, regardless of the timeframe doesn't always boil down to a single underlying trend. And I guess I should add that the third and somewhat related caveat is that using the identified underlying trend from the HH transform didn't appear to have forward looking properties when applied to the data. But I understand it works wonders for sorting underwater currents.
  • #873
  • May 28, 2010 11:22am May 28, 2010 11:22am
Quoting purley
I am just wondering ,how you guys trading?I mean is your system based on purely TA? If yes ,have you done some special study about digital signal processing or regression analysis(nth order polynomials etc..) for your indicators ?
As already pointed out, it depends on how you define 'TA', as it is defined in this forum (trend lines, ma's and oscillators), I don't use it. All my trading is based on Statistical/DSP analysis, but that's as far as I want to go on the subject.
The breaking of a wave cannot explain the whole sea.
  • #874
  • May 28, 2010 12:27pm May 28, 2010 12:27pm
Quoting BlackMage
For those who are interested there is an increasingly large list of comments to the SEC proposals here:

http://sec.gov/comments/s7-02-10/s70210.shtml
Thanks for that, BM - lots of interesting comments. I read randomly from the list - maybe 15 or so - and I did not find a single one defending HFT in equities. Rather, all accusing the latter of contributing to a "two-tiered" market system wherein collocated hedge funds systematically front run any "liquidity providing" orders. These accusing traders say they never provide, but only take, liquidity in their market orders any more as a result, and for that reason, HFT is responsible for removing liquidity from the market overall, with resulting calamities such as the May 6 "flash crash"...

Well, I don't know. I have not yet read the Tradeworx pdf & I am sure there will be arguments there that attempt to refute the latter. But it is so obviously and indefensibly unfair that the mere issue of "collocation" allows a given market participant some advantage over another that it is hard to believe that there will not be some structural remedy for that. It is pretty easy to imagine that such remedy would need to impose a minimum execution time s.t. there can be a level playing field between me sitting at my computer here in Podunk, 3,000 miles away from Wall Street and somebody upstairs from the NYSE....
"If The Fool persists in his Folly he will become wise." - William Blake
  • #875
  • May 28, 2010 12:40pm May 28, 2010 12:40pm
..typical comment...read for yourself....I am ambivalent....I would probably rather have a real economy, though, if forced to choose.

"Derivatives exposure is one of the main issues that need to be dealt with and regulated. High Frequency Trading(HFT) needs to be banned period. Computer algorithm trading is not investing. The markets have not been an investment medium for retirement, for the past 20 years and never will be again, until these trading activities are curbed. Circuit breakers only allow these individuals to continue to push markets to unhealthy levels that fundamentals do not support.

If you want to end volatility and market abuses, get rid of day trading and high frequency trading period. Otherwise, expect more loopholes to be found, and more ways fabricated by these large institutions to continue the casino operations. Yes, the US markets are nothing more than one big casino, where the house, Goldman Sachs and JPM Morgan Chase, will always benefit at the expense of others and their own clients."
"If The Fool persists in his Folly he will become wise." - William Blake
  • #876
  • Edited 1:06pm May 28, 2010 12:55pm | Edited 1:06pm
I think the problem with all of this (and politics and ideology in general) is where do you draw the line? If you enforce a minimum hold time to get rid of HFT, why stop there? Why not extend it out to get rid of day-trading...where does one stop? The unintended by-product of any such rule would be of course less liquidity and more volatility, but it seems people (mostly university professors and left wing malcontents) still think that they can control things as chaotic as markets even after the grand moral and intellectual failure of socialism and command/control economies.

In the end of course, the whole thing is a straw man, HFT was not responsible for the current economic woes, it was bad policy on the part of governments and greed on the part of the banks, HFT is just an easy target. Governments reserve the right to spend like drunken sailors then blame the markets when the money runs out (Germany for example). Basically the whole thing is about as logical as a medieval witch-hunt, but that's politics.
The breaking of a wave cannot explain the whole sea.
  • #877
  • May 28, 2010 1:11pm May 28, 2010 1:11pm
Quoting Craig
I think the problem with all of this (and politics and ideology in general) is where do you draw the line? If you enforce a minimum hold time to get rid of HFT, why stop there? Why not extend it out to get rid of day-trading...where does one stop?
What are these financial markets for? Why do we need day-trading? Why do we need retail forex trading? These are real questions. I walk around every day and I look at crumbling bridges, water systems, & a host of other public infrastructure. I talk to 18 year-olds who can barely read because their public schools are abysmal. Yet I hear that there is no money for these things. On the other hand, there is a $650 trillion market in CDS's. WTF?

Don't even get me started on the so-called "third world"... do it's unhappy denizens aspire to get a piece of the HFT pie? Fuck no, they would like to know where their next meal is coming from.

At the same time, I made more trading the E/U last night, in an hour of "work", than they do in a year. Classic ambivalence. Anyway, I don't want to pollute the thread and detract from the discussion of algorithmic trading & whatnot, so I won't spew on...
"If The Fool persists in his Folly he will become wise." - William Blake
  • #878
  • May 28, 2010 3:16pm May 28, 2010 3:16pm
Quoting purley
I cant' remember where but I read that Hilbert-Huang transform is too slow for trading.
Slow to compute? Maybe for milisecond level trading. Besides, I bet a GPU can do wonders here, since HHT is just the computation of zillions of cubic splines, and I'm sure there are superfast algorithms for splines for GPU.
Quote
Regarding oscillators I read some interesting stuff about scaling.
What do you mean here by scaling? As in scaling in/out or as in scaling law (fractal, ...).
  • #879
  • May 28, 2010 3:19pm May 28, 2010 3:19pm
Quoting FXEZ
Second and perhaps most importantly, the premise of the HH transform is that it boils down noisy data into its main component trend - a single trend. Unfortunately market data, regardless of the timeframe doesn't always boil down to a single underlying trend. And I guess I should add that the third and somewhat related caveat is that using the identified underlying trend from the HH transform didn't appear to have forward looking properties when applied to the data. But I understand it works wonders for sorting underwater currents.
Hmm, what do you mean by a single trend? HHT is a frequency domain analysis, it will output multiple trends (cycles). So you can have a big trend down, and a smaller trend up.
  • #880
  • May 28, 2010 3:49pm May 28, 2010 3:49pm
Quoting mikkom
Tom Demark has written some good stuff, not saying there is anything relevant there (or that there is not) but he has written a lot related to above.
I've read a paper which defined in mathematical form basic TA patterns, and then proceeded to backtest them on daily equity data. The only one they found to have a significant statistical significance was the H&S pattern, but it was not very profitable.

Quote
btw, what are quant versions of indicators?
Like I said in another thread, I've not seen a single candlestick chart in not even one of the quant papers I have. All use continuous data (line chart in MetaTrader). So for example if you have a system with an EMA on the high, it will not work here, since you have no high. This is part of what I mean by quant version - something that works on continuous data, not on candles.

I also mean taking your basic TA indicator, making a math model of it and then seeing how you can improve it. See the example below, where they started from your basic EMA and then modify it a few times until they arrived at the last form, which they use in the trading system presented in the paper.

http://dl.dropbox.com/u/190212/ff/quant_ema_1.png

http://dl.dropbox.com/u/190212/ff/quant_ema_2.png

http://dl.dropbox.com/u/190212/ff/quant_ema_3.png

http://dl.dropbox.com/u/190212/ff/quant_ema_4.png

http://dl.dropbox.com/u/190212/ff/quant_ema_5.png
1 trader viewing now