Systematic trading | Page 42 | 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

  • #821
  • May 20, 2010 4:15pm May 20, 2010 4:15pm
That's all very nice, but I would be wary of nice smooth equity curves...
I had a case of the 'too good to be trues' with my GA, it turned out that it was the order management options that I was feeding the GA. Anything that is 'tight' in terms of stops or take profits will be exploited by the GA to give erroneous results, only with tick data can one test these types of things. With 1 min data the back-test and forward test would diverge chaotically, one only really finds this stuff out by forward testing with small size.
The breaking of a wave cannot explain the whole sea.
  • #822
  • Edited 1:09am May 21, 2010 12:43am | Edited 1:09am
Quoting Craig
That's all very nice, but I would be wary of nice smooth equity curves...
I had a case of the 'too good to be trues' with my GA, it turned out that it was the order management options that I was feeding the GA. Anything that is 'tight' in terms of stops or take profits will be exploited by the GA to give erroneous results, only with tick data can one test these types of things. With 1 min data the back-test and forward test would diverge chaotically, one only really finds this stuff out by forward testing with small size.
I'm totally aware of this, however the trades are not as tight as they might seem to be - as I mentioned some time ago, I for example consider sl to be hit if same candle that the trade is opened with touches it and I still trade daily stuff even though some order management modes go into lower level data now. It's also quite important to note that the curve was a combined curve of 8 methods of over 1500 strategies that are within fitness limits I have specified as selected by the new combination algo I mentioned that does take correlation into consideration in a few different ways.

And I agree with you, I have no idea how different the real results are until I have been running these live for some time.

About the smoothness, it is mostly a phenomenon of quite good combination algo, individual strategy curves are not even nearly as smooth as total curve (note the winning percentage of 57.9% and see below for example - that is one of the strategies included in the combination (risk on the pic below is not the same as it would be for the strategy in the combined curve before)
Attached Image (click to enlarge)
Click to Enlarge Name: Screenshot-AlphaFactory-89.png Size: 22 KB
  • #823
  • May 21, 2010 1:44am May 21, 2010 1:44am
That's an interesting picture, you would have to have balls of steel to trade that by itself (only one winner in some years). Just goes to show the power of diversification in strategies. How long before you try to go live?

Kind of on the same subject, here is a link on some aspects of back-testing which are worth considering http://coreyhoffstein.com/2010/01/06...-back-testing/.
The breaking of a wave cannot explain the whole sea.
  • #824
  • May 21, 2010 4:58am May 21, 2010 4:58am
Quoting mikkom
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?
I have interfaced to the FIX api of both. For a bunch of reason, which I'm not gonna state in public, Duka is inferior in basically any way compared to IB...
  • #825
  • May 21, 2010 5:07am May 21, 2010 5:07am
Quoting BlackMage
I have interfaced to the FIX api of both. For a bunch of reason, which I'm not gonna state in public, Duka is inferior in basically any way compared to IB...
Thanks, so I'll probably use IB then.

Only thing I was really interested with duka was that they are currently offering trades of very small size (min of 1000 units) meaning I could run more systems in parallel with small risk sizes.
  • #826
  • May 21, 2010 5:37am May 21, 2010 5:37am
Quoting mikkom
Thanks, so I'll probably use IB then.

Only thing I was really interested with duka was that they are currently offering trades of very small size (min of 1000 units) meaning I could run more systems in parallel with small risk sizes.
Ok, I can see that in your special case it might make some sense to go with duka at least for a while. Just realize that small size == bad execution.

When its a non-HFT strategy and trading very small sizes is a concern, I would personally look into what Oanda offers as well.
  • #827
  • May 21, 2010 12:52pm May 21, 2010 12:52pm
I'd have to vouch for IB as well at this point, I'm using the CTI gateway, not FIX, but I have not had a single failure yet (touch wood). DTN & IB have been rock solid right though the past couple of weeks, which is good given the current levels of volatility. My server's CPU has been running hot!
The breaking of a wave cannot explain the whole sea.
  • #828
  • May 21, 2010 2:47pm May 21, 2010 2:47pm
Not a bad day, getting there slowly...
Attached Image (click to enlarge)
Click to Enlarge Name: Screenshot.png Size: 118 KB
The breaking of a wave cannot explain the whole sea.
  • #829
  • May 21, 2010 2:51pm May 21, 2010 2:51pm
Great work Craig

If I may friendly recommend you something: Find a way to cut a slice of those commissions, 20% of the profits go to the broker. Just think about how much that is on a yearly basis.

Quoting Craig
Not a bad day, getting there slowly...
  • #830
  • Edited 7:05pm May 21, 2010 2:55pm | Edited 7:05pm
Yeah, I know, it's also due to the small size I'm trading, baby steps...
The breaking of a wave cannot explain the whole sea.
  • #831
  • May 22, 2010 4:00am May 22, 2010 4:00am
Quoting Craig
Yeah, I know, it's also due to the small size I'm trading, baby steps...
So TWS gateway software has been working well for you? Does it require daily restarts?
  • #832
  • May 22, 2010 4:26am May 22, 2010 4:26am
So far, no live failures, so I'd have to say yes...

I've never actually tried to run it 24/7 so I don't really know, after the MB fiasco I had to resign myself to the fact that I was going to have to stop and start stuff each day. I did read on ET somewhere that some guy had been running it for a couple of weeks without a restart, maybe they would know on the Yahoo IB API group? (Maybe BlackMage knows?).
The breaking of a wave cannot explain the whole sea.
  • #833
  • May 22, 2010 5:25pm May 22, 2010 5:25pm
We normally reboot the server once a week for reasons which are not related to the gateway, but between reboots the gateway runs stable without restarts.

I have so far experienced two minor episodes:

- One time after the nightly FIX log out, the connection could not be reestablished until the gateway was restated.

- Another time the gateway was not responding when the trading session started and had to be restarted.

but one should have methods in place to alert about these types of issues anyway.

I never used IB's proprietary API so I can't comment on whether its is as reliable as their FIX connection.
  • #834
  • May 23, 2010 7:22pm May 23, 2010 7:22pm
Quoting Craig
That's an interesting picture, you would have to have balls of steel to trade that by itself (only one winner in some years). Just goes to show the power of diversification in strategies. How long before you try to go live?

Kind of on the same subject, here is a link on some aspects of back-testing which are worth considering http://coreyhoffstein.com/2010/01/06...-back-testing/.
Just imagine if better equitity curve systems were combined !
  • #835
  • May 24, 2010 3:39pm May 24, 2010 3:39pm
Any of you guys have had experience with LCG: http://www.londoncapitalgroup.co.uk/lcgfx_about.html

If so, how do they compare to IB.


Hi Craig,

Good to see IB is working for you
__Thanks__ MJ
  • #836
  • May 24, 2010 5:30pm May 24, 2010 5:30pm
Quoting M_j
Any of you guys have had experience with LCG: http://www.londoncapitalgroup.co.uk/lcgfx_about.html

If so, how do they compare to IB.


Hi Craig,

Good to see IB is working for you
Hey M_j,
How is your FIX engine going? Did you end up connecting to MB?
The breaking of a wave cannot explain the whole sea.
  • #837
  • May 25, 2010 5:27am May 25, 2010 5:27am
Hi Guys, good to see the progress and people having results live. I have been back and forth between using C#/Win and C/C++ Linux and have settled for the latter, I spent the last month resharpening my CC++ skills as well as multithreading (want the quote engine to scale up to 100s of symbols).

My concern at this stage is binding too much to MBs quote api given their lack of stability. I would love to go with IB I think their min is 10k, but I reckon 20-25k would be better to start out with them?

Do IB offer a quote feed api like MB?

Regards

Mark
Economists have forecast 9 out of the last 5 recessions
  • #838
  • May 25, 2010 6:39am May 25, 2010 6:39am
Quoting markmm
Hi Guys, good to see the progress and people having results live. I have been back and forth between using C#/Win and C/C++ Linux and have settled for the latter, I spent the last month resharpening my CC++ skills as well as multithreading (want the quote engine to scale up to 100s of symbols).

My concern at this stage is binding too much to MBs quote api given their lack of stability. I would love to go with IB I think their min is 10k, but I reckon 20-25k would be better to start out with them?
Depends on if you want to trade equities or plain FX/futures etc, IB minimum trade size (for FX/ idealpro) is 20k units but to daytrade stocks you have to have $25k in your account.

Yes they do give out quotes (and about a year of history data) through their custom api.
  • #839
  • May 25, 2010 7:23am May 25, 2010 7:23am
Quoting mikkom
Depends on if you want to trade equities or plain FX/futures etc, IB minimum trade size (for FX/ idealpro) is 20k units but to daytrade stocks you have to have $25k in your account.

Yes they do give out quotes (and about a year of history data) through their custom api.
Thanks Mikkom,

Ill need to look at other options for now.
Economists have forecast 9 out of the last 5 recessions
  • #840
  • May 25, 2010 1:46pm May 25, 2010 1:46pm
Quoting markmm
Hi Guys, good to see the progress and people having results live. I have been back and forth between using C#/Win and C/C++ Linux and have settled for the latter, I spent the last month resharpening my CC++ skills as well as multithreading (want the quote engine to scale up to 100s of symbols).
You can do C++ on both Win32 and Linux with very few of differences in the source code (e.g. where to log), certainly (for me) using VS to debug is a lot easier than gdb. If you use boost you should not really have to write any platform specific code, the only differences should be your configuration values.

Also, unless you really, really know what you are doing I would avoid multi-threading, I handle 100's of quote streams in a single thread no problem, if I need more, I scale out (extra processes).
The breaking of a wave cannot explain the whole sea.
0 traders viewing now