Topic: FST -> Dukascopy

Dear Mr. Popov

When you release the latest FST V2, will it be possible for me to develop some bridging code for it to trade dukascopy's java client? I'm guessing the bridge app will have to be in java - so a friend of mine said he will help me code it. If it works, we can offer it for free on your forum if you like - or do you already have plans to do this?

all the best

Re: FST -> Dukascopy

It'll be excellent if you do so. I'll assist you once I'm back in Europe. I'm planning FST v2 release for the middle of September. The source code will be available also.

Re: FST -> Dukascopy

Popov wrote:

It'll be excellent if you do so. I'll assist you once I'm back in Europe. I'm planning FST v2 release for the middle of September. The source code will be available also.

that would be great  smile

I look forward to it !

Re: FST -> Dukascopy

Mr.Popov

So far i've a couple of possible solutions to this.

java/c# interop bridge

http://hessian.caucho.com/

and

http://java-cs-bridge.sourceforge.net/


my feeling is the second one could be good - but have you had any experience with either of these yet?

If so, what is your opinion?

Re: FST -> Dukascopy

Soladod,
1. We'll probably start a new team work project for connecting FST to Dukascopy. We are now considering the idea. If you are interested, send a PM to fxwinner.

2. I'm not familiar with those Java bridges.

Re: FST -> Dukascopy

Popov wrote:

Soladod,
1. We'll probably start a new team work project for connecting FST to Dukascopy. We are now considering the idea. If you are interested, send a PM to fxwinner.

2. I'm not familiar with those Java bridges.


will do

Re: FST -> Dukascopy

In order to fully benefit from Dukas we need FSB trade tick data and utilize multicore as it could slow down FSB generator. Personally I think it will extremely benefit FSB and brings completely new area of strategies, and we don't need to use "Bar Closing" limitation, it would be real tick data backtester. Don't forget it is not API as for API you need $100k on your account. :-D

Re: FST -> Dukascopy

After I finish with the current FSB and next FST I'll make Dukas tick data backtester.

Multicore support will speed up FSB but will not bring any other advantage. Even now a user can start several copies of FSB and to imitate multicore usage.

Re: FST -> Dukascopy

Having 8 FSBs running on i7 even not close to performance of One FSB utilizing MuliCore,
1. shared memory / cpu access
2. Each instance only slow down processing waiting for each other
3. Slow exhaustion on strategies, and  generator runs on the same pair is useless by itself as it would go thru the same strategies over and over on each instance.

Re: FST -> Dukascopy

A multicore generator will calculate strategies using same settings and same data.
Several instances can generate strategies with differemt settings and different data.

Re: FST -> Dukascopy

I think that multicore calculations will increase the speed of the generator with 70-80%. I'll try it after finishing with tick data backtester.

Re: FST -> Dukascopy

Popov wrote:

I think that multicore calculations will increase the speed of the generator with 70-80%. I'll try it after finishing with tick data backtester.

I think that there are no great value in multicore feature because I have several times runned FSB up to one week simultaneously so that FSB made over 1,5 million calculations. After 150 000 thare are no real strategy improvements to get better combinations of the profitable strategy.

It seems that FSB is over-optimizing current indicators and strategy.

Re: FST -> Dukascopy

I myself  run the Generator for maximum 5 minutes. But multicore will speed up it additionally. If I find an easy and safe way to implement it, I'll do it.

Re: FST -> Dukascopy

well, after we get to Tick data you will understand that Multi-core is a Must feature, I am sorry to see all my 12 cores staying idle while only one loaded with 30%. Sure I can run multiple FSBs but this is not the point, multi FSBs brings confusion the same way as multi FSTs do. I personally prefer One instance FSB, and one instance FST running.

if we plan to add more complexity to FSB (more indicators, tick data etc) , then we have to do it scalable, one core is not scalable solution, you can stay this way if you like, but technology  moving fast and if we don't adopt it now, in a year we would be far behind. think this way.

Re: FST -> Dukascopy

fxwinner wrote:

I am sorry to see all my 12 cores staying idle while only one loaded with 30%.

What kind of CPU is that - 12 cores ?

I'm running FSB on AMD Phenom II x4