Systematic trading | 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

  • #1
  • First Post: Edited May 7, 2010 2:11am May 7, 2009 1:19pm | Edited May 7, 2010 2:11am
Hello all,

I decided to open this thread to discuss about systematic trading and development of strategies in "quantifiable" way.

At this moment I have decided to move away from metatrader that I have primarily been using for my trading strategy testing and development [silly, yes I know]. I have been reviewing all kinds of platforms and all of them seem quite not what I'm looking so I'm going to build my own testing platform during the thread and discuss the process [more about it during the thread].

I'm going to go a bit different route than most of the people [well most of the retail folk at least] and use a database to store all the data. At the moment I already have a solid schema with 1 minute data since 2001 stored for all majors so if anyone wants to have some insight into daily ranges etc, feel free to ask It's going to take about 5 seconds to see a statistical table of london opening ranges for example. I'm also going to use database a bit differently than most people but again, more about this later.

I'm also going to discuss how to separate elements from a system and what really makes a system [hint: there are three elements that can be divided to smaller elements] and hope to get some ideas on how to clarify this even further.

I also hope we can find some quantifiable edges during the thread but that's still something that's totally open and I hope to get some ideas from the people following the thread.

I plan to add automation to automatically discover additions to systems I plan to test but this is kind of advanced stuff and maybe I should talk about this after the groundwork is done [?].

I want to warn people, I have background in tech and I have been doing programming for 15 years [and trading a bit less than 10 more or less actively, on some years I didn't trade at all] so sometimes I totally underestimate how hard the tech stuff is for some people, feel free to ask anything.

That's about it.. Let's see where I should go first..

If you want to get some statistical data about majors, feel free to ask, I might post some details without asking.. I have only basic stuff right now, I have to complete the framework to get to more advanced stuff

Some links to other sites related to systematic trading

I opened this for everyone to post interesting links or browse past ones
Systematic trading subreddit

http://marketsci.wordpress.com/
http://www.nuclearphynance.com/
http://www.wilmott.com/index.cfm?NoC...=Yes&forumid=1
http://www.quantnet.org/forum/
http://www.tradingblox.com/forum/
http://www.elitetrader.com/vb/forumd...?s=&forumid=32
http://www.maxdama.com/
http://intelligenttradingtech.blogspot.com/

Material related to systematic trading

http://www.cis.upenn.edu/~mkearns/projects/plat.html

Open source software related to systematic trading

http://www.marketcetera.com/site/
http://quantlib.org/index.shtml
http://groups.google.com/group/jsystemtrader
  • #2
  • May 7, 2009 1:37pm May 7, 2009 1:37pm
First, how do you define a Forex day for a daily range? Since its a 24 hour market what times do you use?

Otherwise, I would be interested in some USD/JPY daily ranges, and possibly a little explanation about how you determine such numbers. What software do you have, where did you get your data, how do you test it, etc.

That would be great, thanks!
All I ask for is a chance to prove that money cannot make me happy.
  • #3
  • May 7, 2009 1:40pm May 7, 2009 1:40pm
Hi Mikko,

I look forward to reading your thoughts on a systematic approach as this might help me quantify my discretionary methods and help provide me with some consistency.

I have also done some work with ranges and hope your thread will help me enhance my results.

On the subject of platforms, I was fortunate to demo the Topgun platform recently and its a fantastic tool, but as you need to open an account with their preferred brokers or pay a couple of thousand dollars to use a different broker, I abandoned it. Might be worth you looking at it as it has some really great features that I haven't seen in other platforms (at least not without having to program an indicator or EA or such).

Bring it on fella, I'm waiting...
  • #4
  • May 7, 2009 1:45pm May 7, 2009 1:45pm
You can break the day into the relevant sessions (Asian, Eurozone, US), use your broker midnight as a start and end point, Use London or US or Japan midnight as start and end... you get the picture.

Quoting Marsh
First, how do you define a Forex day for a daily range? Since its a 24 hour market what times do you use?
  • #5
  • May 7, 2009 1:47pm May 7, 2009 1:47pm
I hope we can get a good discussion going here, mikkom, it sounds like we do similar things, I do purely systematic trading I just had finished hosting my ATS on a Linux node in CA to be close to the MB gateways, I am using C++ and a Postgres database.
The breaking of a wave cannot explain the whole sea.
  • #6
  • May 7, 2009 1:49pm May 7, 2009 1:49pm
Here is a simple example of what can be done in few seconds - this is EURUSD london session range as a curve (2001 to today).
Attached Image (click to enlarge)
Click to Enlarge Name: london.png Size: 9 KB
  • #7
  • May 7, 2009 2:27pm May 7, 2009 2:27pm
Quoting Craig
I hope we can get a good discussion going here, mikkom, it sounds like we do similar things, I do purely systematic trading I just had finished hosting my ATS on a Linux node in CA to be close to the MB gateways, I am using C++ and a Postgres database.
I went with MySQL (tested a bit berkeley DB first but surprisingly Mysql is faster) and java. c++ is faster but java is easier to distribute if I decide to get a cluster at some point.

Mysql is not exactly a top notch database but MyISAM is about the fastest engine there is when doing lots of qeries [and trading research is all about queries] and I'm not too interested in transactional features with this framework.
  • #8
  • May 7, 2009 2:29pm May 7, 2009 2:29pm
Quoting Marsh
First, how do you define a Forex day for a daily range? Since its a 24 hour market what times do you use?

Otherwise, I would be interested in some USD/JPY daily ranges, and possibly a little explanation about how you determine such numbers. What software do you have, where did you get your data, how do you test it, etc.

That would be great, thanks!
You are right there is no exact "daily range", there are just session ranges. When I have my stuff ready I'll do some kind of analysis on session ranges on all majors - it's going to be interesting chart.
  • #9
  • Edited 2:55pm May 7, 2009 2:44pm | Edited 2:55pm
Quoting mikkom
I went with MySQL (tested a bit berkeley DB first but surprisingly Mysql is faster) and java. c++ is faster but java is easier to distribute if I decide to get a cluster at some point.

Mysql is not exactly a top notch database but MyISAM is about the fastest engine there is when doing lots of qeries [and trading research is all about queries] and I'm not too interested in transactional features with this framework.
I had a look at MySQL, it is faster and more popular, but it seems they had kind of a disaster with a recent release not being up to scratch, apparently the management is now focusing on 'features'. I don't run a lot of big historical queries, I just want the data to be safe. I only keep the prevoius months worth of data for reasons I think we previously discussed in some other thread.

Edit: Now I think back there were some seemly arbitrary limitations with the MySQL timestamps, I timestamp the order send time and fill time down to the nearest millisecond, if I remember correctly MySQL only had second resolution. I do this timestamping in order to measure the effectiveness of simulated trading vs. real.
The breaking of a wave cannot explain the whole sea.
  • #10
  • Edited 3:14pm May 7, 2009 3:02pm | Edited 3:14pm
Quoting Craig
I had a look at MySQL, it is faster and more popular, but it seems they had kind of a disaster with a recent release not being up to scratch, apparently the management is now focusing on 'features'. I don't run a lot of big historical queries, I just want the data to be safe. I only keep the prevoius months worth of data for reasons I think we previously discussed in some other thread.

Edit: Now I think back there were some seemly arbitrary limitations with the MySQL timestamps, I timestamp the order send time and fill time down to the nearest...
I think I'm not going to get into database discussion here [it's a bit irrelevant] but basically both mysql and postgres are mature databases. Postgres is in the past been known for corruption so just get your daily backups.. [I think new versions are better]

I think you are correct about micro/millisecond thing, I'm not planning to do high frequency stuff [at least yet] so it's not an issue for me.
  • #11
  • May 7, 2009 3:20pm May 7, 2009 3:20pm
Hi Mikkom

I'm glad to see a thread on systematic trading here. Looking forward to hearing you guys talk about the secrets I haven't discovered yet. Ready to participate in interesting research if anybody needs a bit of help (I'm a programmer working for one of UK's biggest hedge funds and after hours looking for that edge that'll allow me to quit my job ).

I'd be quite interested to find out what the foundations of your strategies are and what areas you find most interesting to explore. keep up the good work!
  • #12
  • Edited May 8, 2009 7:38am May 7, 2009 3:32pm | Edited May 8, 2009 7:38am
A trading system is a combination of following things

Opening a trade

Position sizing

Closing a trade or what I call position management

That's it. Simple eh Well not really. That's why we have to define different sections of each of the above. I try to overline my thoughts below.

Entering a trade
- Mechanical rules for triggering an entry position. This is a trigger for entry method to start.
- Entry method - how to enter into position (single entry at some point, fade in, enter at market when certain second rules match - lots of stuff can be added here)
- Select a position management method for entered trades. Position management can be split so parts of position (for example divided by percentage or some other method) are managed in a different way.

Position sizing
- I would always go with r% [for total position size]
- Rules to modify risk level if needed
- (entry method defines how the risk is divided if entry is done with multiple positions)

Position management
- When a position or part of a position is closed
- adjusting of "sl" and "tp" (also fractional sl and tp)
- (possibility of triggering of other trades based on closed trade)

I would be very happy to hear what I'm missing if I am missing something. I'm trying to gather all the main elements of systematical trading to these simple sections.
  • #13
  • May 7, 2009 3:37pm May 7, 2009 3:37pm
Quoting mikkom
You are right there is no exact "daily range", there are just session ranges. When I have my stuff ready I'll do some kind of analysis on session ranges on all majors - it's going to be interesting chart.

I appreciate it mikkom. Thanks for doing all this.
All I ask for is a chance to prove that money cannot make me happy.
  • #14
  • May 7, 2009 3:39pm May 7, 2009 3:39pm
Look toward 1.3285
Reasons may follow.
  • #15
  • May 7, 2009 3:44pm May 7, 2009 3:44pm
Quoting mikkom
A trading system is a combination of following things

Opening a trade

Position sizing

Closing a trade or what I call position management

That's it. Simple eh Well not really. That's why we have to define different sections of each of the above. I try to overline my thoughts below.

Entering a trade
- Mechanical rules for triggering an entry position. This is a trigger for entry method to start.
- Entry method - how to enter into position (single entry at some point, fade in, enter at market when certain second rules match - lots of stuff...
Do you also want to include the question 'when should I trade (or not trade) system 'x'?
The breaking of a wave cannot explain the whole sea.
  • #16
  • May 7, 2009 3:46pm May 7, 2009 3:46pm
Quoting Craig
Do you also want to include the question 'when should I trade (or not trade) system 'x'?
I think that's covered by position sizing rule and entry trigger rule. I personally don't ever shut down system completely, just adjust the risk lower.

Or are you talking about system rot ie disappearing edge?
  • #17
  • May 7, 2009 3:50pm May 7, 2009 3:50pm
[quote=mikkom;2719980]I think that's covered by position sizing rule and entry trigger rule.

Agreed. A position size of 0 (zero) lots is a just a special case of trade. A one with 0 risk and 0 profit (ror) expectancy. A perfectly valid case in a systematic approach.
  • #18
  • May 7, 2009 3:51pm May 7, 2009 3:51pm
Quoting mikkom
I think that's covered by position sizing rule and entry trigger rule. I personally don't ever shut down system completely, just adjust the risk to very low.

Or are you talking about system rot?
I've never heard of the term 'system rot', but I think we are talking about the same thing, systems become more or less profitable over time. Instead of adjusting risk I switch systems on or off, but it all boils down to the same question, which is 'how do I react to changes in profitability?' or more specifically 'how do I assign a metric to profitability in order to adjust my risk?'.
The breaking of a wave cannot explain the whole sea.
  • #19
  • May 8, 2009 1:17am May 8, 2009 1:17am
One thing I'm interesting in measuring is news impact. Does anyone know where I could find a news calendar with numbers in machine readable form (xml, csv..) that goes back to about 10 years (it doesn't have to be free)?
  • #20
  • Edited 1:51am May 8, 2009 1:27am | Edited 1:51am
Quoting Craig
I've never heard of the term 'system rot', but I think we are talking about the same thing, systems become more or less profitable over time. Instead of adjusting risk I switch systems on or off, but it all boils down to the same question, which is 'how do I react to changes in profitability?' or more specifically 'how do I assign a metric to profitability in order to adjust my risk?'.
I have never been able to do a solid metric for system efficiency (usually you never know when for example range expansion hits after long whipsaw period) but my thoughts are that the metric results to be used should be mapped to risk, not necessarily 1:1 though.

edit: just to clarify yes there are some aspects that give better expectancy, for example for breakout systems large breakouts tend to happen when volatility is going down but I have not been able to quantify this to so much reliability as I would want to, therefore I hope my framework will help in that. I don't want to guess things and thats' why I moved back to fixed r% in my live trading and not variable r%. It's going to be really interesting to see how sharpes and such ar affected when extra risk management is added (I tend to do it both ways so I'll also increase risk on positive expectancy).

Acrary (on elitetrader) had quite an interesting idea for testing if edge was still effetictive, I hope I don't misquote him but the idea was to run many random trades and compare your systems trades against them to see if the edge was still valid. Efficiency of your system versus random trades is one metric that I have been pondering but it's more of a system rot measurement than actual measurement of when to use a system.

I'm going to implement an automatic discovery algorithm that goes through your trade history and looks for periods where the trades were most unsuccesful and finds common features of the curve at those points but there still has to be some human factory there because I understand quite well how bad curve fitting can be [yes I have tested evolutionary methods in the past ].
8 traders viewing now