Systematic trading | Page 41 | 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

  • #801
  • May 3, 2010 11:53pm May 3, 2010 11:53pm
Quoting aidans
A lot of discussion is go on about the speed of the engine, where is the bottle neck?...
You still need to run the 2000 strategies to compute all the positions they will open. But it's an interesting idea, of saving only the positions. I'm not sure if it's an obvious idea, since I'm not doing data-mining.
  • #802
  • May 4, 2010 12:10am May 4, 2010 12:10am
Quoting aidans
A lot of discussion is go on about the speed of the engine, where is the bottle neck?...
That's basically what I do, I thought storing trades instead of running strategies over and over is quite obvious [?]
  • #803
  • May 6, 2010 10:49am May 6, 2010 10:49am
I found this video to be worth 8 minutes and 51 seconds (plus a few milliseconds) of my life:

http://www.technologyreview.com/video/?vid=504
  • #804
  • May 6, 2010 11:23am May 6, 2010 11:23am
Quoting BlackMage
I found this video to be worth 8 minutes and 51 seconds (plus a few milliseconds) of my life:

http://www.technologyreview.com/video/?vid=504
Good video thanks, he didn't mention the transaction costs, for a million trades at $1 vs 1 trade at $1 million it must be a problem.
Economists have forecast 9 out of the last 5 recessions
  • #805
  • May 6, 2010 1:00pm May 6, 2010 1:00pm
Quoting BlackMage
I found this video to be worth 8 minutes and 51 seconds (plus a few milliseconds) of my life:

http://www.technologyreview.com/video/?vid=504
Cool video, I hope that guy draws his fibs correctly
The breaking of a wave cannot explain the whole sea.
  • #806
  • May 6, 2010 2:59pm May 6, 2010 2:59pm
Quoting markmm
he didn't mention the transaction costs, for a million trades at $1 vs 1 trade at $1 million
Well, if he trades retail forex then his commisions are included in the spread cost which is a percentage of total position, aggregate for all entries assuming single direction. Seems there's mighty wisdom in these methods (I imagine vaguely related to gridding) regardless of trading frequency.
  • #807
  • May 6, 2010 3:56pm May 6, 2010 3:56pm
Quoting mikkom
That's basically what I do, I thought storing trades instead of running strategies over and over is quite obvious [?]
This is what I do also. I currently maintain over 100000 strategies using tickData. This with good old "slow" java and some db technology.
  • #808
  • May 6, 2010 8:48pm May 6, 2010 8:48pm
Hi spekkiefx

Can you give us some hint on the DB technology.

Either Tick Data or strategies are timing series, which DB is the most handy for the job?


Hi mikkom,
At current stage, I might load the text tick data into Java.
you mentioned in here that the cached version of arrays on the disk. What is the technology name? I may check it out.

Thank you.


Quoting spekkiefx
This is what I do also. I currently maintain over 100000 strategies using tickData. This with good old "slow" java and some db technology.
  • #809
  • May 6, 2010 11:47pm May 6, 2010 11:47pm
Quoting aidans
At current stage, I might load the text tick data into Java.
you mentioned in here that the cached version of arrays on the disk. What is the technology name? I may check it out.
I use raw java IO to read/write values directly to the HD, nothing fancy.
  • #810
  • May 7, 2010 12:20am May 7, 2010 12:20am
Gentlemen, this thread has spurred me to register (been lurking for a while). So now I've got my ten years of 1m for the majors loaded up on my trusty mySQL server. Wanted to share the results of a simple query with you, does it pass the sniff test - wouldn't be the first time I goofed in SQL:
Average volatility in pips
EURUSD UTC
Attached Image (click to enlarge)
Click to Enlarge Name: volitbyyear.png Size: 17 KB


Attached Image (click to enlarge)
Click to Enlarge Name: heat.png Size: 14 KB
  • #811
  • May 7, 2010 12:21am May 7, 2010 12:21am
Quoting aidans
Hi spekkiefx

Can you give us some hint on the DB technology.

Either Tick Data or strategies are timing series, which DB is the most handy for the job?


Hi mikkom,
At current stage, I might load the text tick data into Java.
you mentioned in here that the cached version of arrays on the disk. What is the technology name? I may check it out.

Thank you.
I personally use oracle but I see no reason why you couldn't use mysql or something. When performance is an issue I write raw java I/O from the db data. Best of both worlds this way. Good old sql can save some time when analyzing things.
  • #812
  • May 7, 2010 12:22am May 7, 2010 12:22am
Quoting aidans
Hi spekkiefx

Can you give us some hint on the DB technology.

Either Tick Data or strategies are timing series, which DB is the most handy for the job?


Hi mikkom,
At current stage, I might load the text tick data into Java.
you mentioned in here that the cached version of arrays on the disk. What is the technology name? I may check it out.

Thank you.
I personally use oracle(you download a free developers version from oracle) but I see no reason why you couldn't use mysql or something. When performance is an issue I write raw java I/O from the db data. Best of both worlds this way. Good old sql can save some time when analyzing things.
  • #813
  • May 17, 2010 11:50am May 17, 2010 11:50am
Not exactly related to systematic trading but worth watching anyway

http://www.babelgum.com/browser.php#...URED/0,5003507
  • #814
  • May 17, 2010 1:55pm May 17, 2010 1:55pm
.. In other news, I got lots of 1 min stock + futures data from pi trading today (couple of DVDs), now I'll really see how fast I can tune my testing framework

My plan is to build a small cluster of generic, cheap computers and run a parallelized system in top of that.

Related question: what kind of bad tick/bad bar checks do you use?
  • #815
  • May 17, 2010 3:07pm May 17, 2010 3:07pm
Open, Low or Close above the High.
Open, High or Close below the Low.
Prices = 0.
Spike filter.
Date Gaps.
The breaking of a wave cannot explain the whole sea.
  • #816
  • May 17, 2010 3:40pm May 17, 2010 3:40pm
Quoting mikkom
Related question: what kind of bad tick/bad bar checks do you use?
That's a very hard problem. Here's a "professional" solution: http://books.google.com/books?id=16xhmuRY-XwC&lpg=PP1&dq=high%20frequency%20trading&pg=PA82#v=onepage&q=high%20frequency%20trading&f=false

Given the complexity of a full solution, I've decided to not solve it I just compare the data with another source for big errors (20 pips+), and ignore smaller errors (by requiring at least 20 pips profit per trade for my systems).

I've also looked at OpenCL a bit. I think I know how to write efficient kernels for converting tick data to bar data, and also how to compute simple indicators like EMA and RSI on it. So I could precompute in one pass all the indicators I need for the simulation run, and then run on these precomputed values, instead of building them on the go. Even more, I could start running the strategy even before the indicators are fully computed, with the CPU running the trading system and the GPU building the indicators. However I don't need this kind of speed at the moment.
  • #817
  • May 17, 2010 3:51pm May 17, 2010 3:51pm
I have tried to read that chapter in that book about 5 times now, the non-standard notation drives me around the twist. Their filter has so many degrees of freedom I'm not sure how they manage to do anything useful with it. Most people just use a if it's X% away from the last quote drop it type scheme, I just copied what is done in NinjaTrader for quote filtering, it's fairly straight forward.
The breaking of a wave cannot explain the whole sea.
  • #818
  • May 17, 2010 11:40pm May 17, 2010 11:40pm
Quoting mikkom
Not exactly related to systematic trading but worth watching anyway

http://www.babelgum.com/browser.php#...URED/0,5003507
nice find mikkom, a good relaxing overview and some history
  • #819
  • May 19, 2010 11:19pm May 19, 2010 11:19pm
I am learning a thing or two from this thread - thanks to mikkom, Craig, et al. I won't have much to say, though, just a rank beginner.

Best Regards,

Rich
"If The Fool persists in his Folly he will become wise." - William Blake
  • #820
  • Edited 1:36pm May 20, 2010 12:55pm | Edited 1:36pm
This is one of my latest runs (fx, not equities) - after some light bulb moments (as silly as it might sound, after about 10 years of trading - with less trading in the middle of the mentioned years - I think I have finally made myself clear what entry and exit actually mean/are and how to get best results from them) I rethought and coded quite a much of the selection/combination algos, changed fitness algo and added some more order management possibilities to the mix.

I have been trying to verify that this is correct for the past 3 days or something because if this is -- well looks like I'm (or my framework is to be exact) onto something.

Pretty pic attached (last 2 years are out-sample, data is to the end of march).

I haven't tested this live yet so I can't verify if this actually works like this in reality - this would require some major changes with my live EA, and I'm actually thinking of moving back to IB or Dukas so it looks like I have some api coding ahead..

Note the MAR of 14.5 I'm still very unsure if these results are real but I have been going through code, data and generated trades for the last 3 nights and I just can't find anything wrong..

Has anyone here done any automated trading with Dukas thought their api? If yes, how is the api working and how are the slippage/fills compared to IB?
Attached Image (click to enlarge)
Click to Enlarge Name: Screenshot-AlphaFactory-88.png Size: 24 KB
2 traders viewing now