Algorithmic Trading Group Forum
https://www.algotradinggroup.com/cgi-bin/yabb2/YaBB.pl
Trading Strategies >> Development >> Ultra high frequency FX
https://www.algotradinggroup.com/cgi-bin/yabb2/YaBB.pl?num=1237777321

Message started by Algo Designer on 03/23/09 at 04:02:00

Title: Ultra high frequency FX
Post by Algo Designer on 03/23/09 at 04:02:00

Does any of our forum members specialise on ultra high frequency FX algorithmic development/trading?

Title: Re: Ultra high frequency FX
Post by esmondo on 04/06/09 at 13:03:54

If you are talking about FX Futures, then i might eb able to assist you.

Title: Re: Ultra high frequency FX
Post by Cooper on 04/06/09 at 13:46:12

It seems to me that forex spreads are so wide as to make ultra-high frequency very difficult. I stick to medium-high frequency.

Title: Re: Ultra high frequency FX
Post by BlackMage on 04/06/09 at 13:58:37

What is your definition of "ultra high frequency"?. This often means different things to different people.

Spreads in spot forex is generally lower than in e.g. CME currency futures, so this should not necessarily be prohibitive for (ultra) high frequency trading.

Title: Re: Ultra high frequency FX
Post by Alexander Temerev on 04/06/09 at 14:36:08


Algo Designer wrote on 03/23/09 at 04:02:00:
Does any of our forum members specialise on ultra high frequency FX algorithmic development/trading?


I do. Any questions? :)

Title: Re: Ultra high frequency FX
Post by Algo Designer on 04/06/09 at 14:47:43


esmondo wrote on 04/06/09 at 13:03:54:
If you are talking about FX Futures, then i might eb able to assist you.


I am experimenting with a number of high-frequency trading algorithms generating quant signals and making market "around them" as well as crossing the spread. FX Futures is one of the instruments the models can employ.

Title: Re: Ultra high frequency FX
Post by Algo Designer on 04/06/09 at 14:59:42


Cooper wrote on 04/06/09 at 13:46:12:
It seems to me that forex spreads are so wide as to make ultra-high frequency very difficult. I stick to medium-high frequency.


Very true, especially in the current market conditions. What it means from my perspective, is that even if the market local / global microstructures are identified and the theoretical signal are statistically sound, most of them cannot be traded upon. This is where a robust execution strategy would come in handy to work the signal while it is valid and avoid crossing the spread to avoid significant transaction costs that kill the Sharpe ratio and the overall risk profile of the strategy.


Title: Re: Ultra high frequency FX
Post by Algo Designer on 04/06/09 at 15:15:14


BlackMage wrote on 04/06/09 at 13:58:37:
What is your definition of "ultra high frequency"?. This often means different things to different people.

Spreads in spot forex is generally lower than in e.g. CME currency futures, so this should not necessarily be prohibitive for (ultra) high frequency trading.


In my definition, anything that averages at least one signal a minute (more or less). The number of real trades can be lower or higher depending on the market liquidity and the rate of change.

Yes, FX ultra high frequency trading is easier to implement on FX spot due to its liquidity.

Title: Re: Ultra high frequency FX
Post by Algo Designer on 04/06/09 at 15:30:19


Alexander Temerev wrote on 04/06/09 at 14:36:08:

Algo Designer wrote on 03/23/09 at 04:02:00:
Does any of our forum members specialise on ultra high frequency FX algorithmic development/trading?


I do. Any questions? :)


Great! Lots of them. FX algorithmic space is a fascinating new area compared to relatively well established cash equities for example.

I would like to engage a group of like-minded individuals to discuss topics varying from signal generation to optimisation of executions strategies that deliver them to the real market, including but not limited to:

  • Research tools
  • Quant libraries
  • Data management tools and systems
  • Market data vendors
  • Market makers and liquidity providers
  • Strategy design and persistent patterns
  • Testing methodologies and techniques
  • Crash analysis
  • Order optimisation and execution strategies
  • Risk management
  • Performance Monitoring
  • Active strategy selection

    The list goes on and on... :)




Title: Re: Ultra high frequency FX
Post by Alexander Temerev on 04/06/09 at 23:56:42


Algo Designer wrote on 04/06/09 at 15:30:19:
I would like to engage a group of like-minded individuals to discuss topics varying from signal generation to optimisation of executions strategies that deliver them to the real market, including but not limited to:

  • Research tools
  • Quant libraries
  • Data management tools and systems
  • Market data vendors
  • Market makers and liquidity providers
  • Strategy design and persistent patterns
  • Testing methodologies and techniques
  • Crash analysis
  • Order optimisation and execution strategies
  • Risk management
  • Performance Monitoring
  • Active strategy selection

    The list goes on and on... :)



My own little business is taking a trading idea from quants and turning it into fully automatic strategy deployed in production and running 24x5. I develop connectivity links to data providers / liquidity providers, smart order routers / executors etc. Also I solve problems like temporarily thin / absent liquidity, dealing with reconnections / network failures, optimizing execution for lowest slippage, and other grease monkey tasks no one else wants to take. :)

Title: Re: Ultra high frequency FX
Post by Nicolas Macherey on 04/07/09 at 10:29:08

Hi,

I am developping automatic forex trading strategies in a middle High Frequency mode. (60 min to daily charts).
My business is oriented toward the developpement of fully automated trading strategies starting from historical price data and going to order generation. (Market neutral, Absolute returns...)
If you need any info...

Title: Re: Ultra high frequency FX
Post by Algo Designer on 04/08/09 at 13:05:24

I am experimenting with a set of fully automated engines in 10 - 15s range (spot FX) taking care of all the essential layers, market data, pre-processing, prediction engine, optimisation, back-testing, walk-forward testing, trade simulation, order execution strategy, OMS, market links, etc.

Title: Re: Ultra high frequency FX
Post by ManInMoon on 04/15/09 at 11:24:27

Hi guys,

Is anyone actually running Forex HF models right now?

Moon

Title: Re: Ultra high frequency FX
Post by Nicolas Macherey on 04/15/09 at 11:44:35

It depends what you mean by High Frequency, I am running 60 min and Daily Forex models on EURUSD and USDCHF


Title: Re: Ultra high frequency FX
Post by ManInMoon on 04/15/09 at 12:24:00

Hi Nicolas,

60 mins sounds interesting. Can you give me an idea of your daily tradecount for those? i.e. do you trade just 1 or twice a day on average, or much higher?

Moon

Title: Re: Ultra high frequency FX
Post by Nicolas Macherey on 04/16/09 at 09:06:57

We trade at most 2 times a day and at least 1 time in 3 days... It depends on the market volatility. However we have really much better results for our daily charts than for our 60 min charts...

Last year we recorded about 30% TNP for USDCHF and 25% TNP for EURUSD with an initial leverage of 1 and no reinvestment strategy.

good day !

Title: Re: Ultra high frequency FX
Post by Algo Designer on 04/16/09 at 12:38:32

From what you have described, you are doing remarkably well! 25 - 30% TNP with 1 : 1 levarage. With this trading frequency, slippage would not be an issue at all. How long have you been running your strategies for? Are you managing a fund or a prop trading firm?


Title: Re: Ultra high frequency FX
Post by Nicolas Macherey on 04/16/09 at 12:45:46

In fact our models are running in real-time in TradeStation since september 2007 for FOREX and since January 2007 for US FUTURES, we are a young company who want to do Technical Advising... You can have a look at our presentation on www.graymat.fr (we do not show our results) ! For the moment we only put some of our money on a trading account and are looking for clients...

regards

Title: Re: Ultra high frequency FX
Post by Algo Designer on 04/16/09 at 13:06:23

I am impressed with your website and the quality of presentation, simple and straight to the point, as long as you can back it up by numbers. If you the returns you have mentioned are sustainable, what is the point in looking for clients? To scale the business up?

How frequenly are your models re-trained? Is this externally or internally (presumably by the models themselves) initiated?

Title: Re: Ultra high frequency FX
Post by Nicolas Macherey on 04/16/09 at 14:00:38

Thanks,

The problem to grow the business up is that our strategies are Absolute returns uncorrelated. And in France it's very difficult to sell it.
We are all Engineers in our team and we do not have a CTA or a Fund Manager... And it's more difficult...
However we are working with professionnals since 4 years to validate our models in the markets but they cannot help us to sell it.

Our model are re-trained monthly in real trading conditions but for reasons of tests at the moment, we just let them trade untill they do not work anymore.

Here you have a link to one of our product sheet on USDCHF. The trading strategy is Long/Short 100% of time in the market. Spreads and slippages are taken into account. The model has not been re-trained since september 2007.

regards

Title: Re: Ultra high frequency FX
Post by skyrocket on 05/19/09 at 01:20:03

I wonder what kind of strategy you guys have been using?
Is it a bit like trational technical analysis or like quite maths based strategy, like ( co-integration, PCA) or fancy maths (like GA, NN)?
I'm quite new to this area.

Many thanks for any comments.

regards,

Title: Re: Ultra high frequency FX
Post by Algo Designer on 05/22/09 at 02:09:49

In my case, no traditional TA. My models use ANN, GA, Swarm Intelligence (SI), Cellular Automata, SVM and various non-linear regression techniques. I am also experimenting with the Minority Game.

Title: Re: Ultra high frequency FX
Post by Nicolas Macherey on 05/22/09 at 06:10:31

Hi,

This is the same for me... I am using GA, Swarm Intelligence, ANN, Fuzzy Logic Systems...

Title: Re: Ultra high frequency FX
Post by Cooper on 06/18/09 at 06:47:33

I trade medium-high frequency, using 1-minute bars, but typically averaging one trade per day, 15 currency pairs. Typical trade might last a few hours.  Biggest problem is liquidity. If you want to do ultra-high frequency, you will be limited to a few million per trade before you start to get bigger slippage.

Title: Re: Ultra high frequency FX
Post by Algo Designer on 07/01/09 at 06:26:02

I fully agree with you there on the liquidity limitations for ultra-high frequency. Slippage does become a major problem.

1-minute bars appear to be a sweet spot IMHO. Is your trading semi or fully automated?

Title: Re: Ultra high frequency FX
Post by Cooper on 07/01/09 at 16:43:59


Algo Designer wrote on 07/01/09 at 06:26:02:
Is your trading semi or fully automated?

Fully automated.

Title: Re: Ultra high frequency FX
Post by Algo Designer on 07/02/09 at 02:28:13

That's the spirit! Same here. In your strategies, do you use a stationary set of rules or "fancy" adaptive techniques?

Title: Re: Ultra high frequency FX
Post by Cooper on 07/02/09 at 02:52:09


Algo Designer wrote on 07/02/09 at 02:28:13:
That's the spirit! Same here. In your strategies, do you use a stationary set of rules or "fancy" adaptive techniques?

Somewhat adaptive, but needs to be much more adaptive. I need to improve the feedback loop surrounding regime drift.

Title: Re: Ultra high frequency FX
Post by Algo Designer on 07/02/09 at 04:47:16

That makes sense. Where are you located? I am in Sydney.

Title: Re: Ultra high frequency FX
Post by Cooper on 07/02/09 at 04:54:52

SF Bay Area.

Title: Re: Ultra high frequency FX
Post by Nicolas Macherey on 07/04/09 at 06:00:47

Hi,

What kind of paradigm are you using for trading with ultra high frequency ?

I have implemented many tests that shows that it's not possible to trade forex with a frequency less than one hour, cause of slippages and commissions. It the same for many other stocks... More over liquidities are very difficult to manage in a UHF manner...

On the other side, I'd like to start a web site specialized for systematic traders (fully automatized) who wants to show their results and track records. The aim is to put in the same place many informations for traders who are looking for partnerships with money managers and CTAs.
Are you interested on it ?

regards
Nicolas

Title: Re: Ultra high frequency FX
Post by floortrader on 07/07/09 at 13:24:22

yes, I would be very interested in participating and supplying the start-up capital for the site. Think its a great idea to have a home base for all of the next gen algo trading minds.

Title: Re: Ultra high frequency FX
Post by Nicolas Macherey on 07/08/09 at 06:43:55

Hi ,

Great, I started a new topic for this discussion in Professionnal Algo Trading Web Site in Software Infrastructure ...

Thanks for your reply!

you can mail me by my web site if you want to discuss about it (www.graymat.fr)

Title: Re: Ultra high frequency FX
Post by BlackMage on 07/21/09 at 14:19:49


Nicolas Macherey wrote on 07/04/09 at 06:00:47:
What kind of paradigm are you using for trading with ultra high frequency ?

I have implemented many tests that shows that it's not possible to trade forex with a frequency less than one hour, cause of slippages and commissions.


My take on this is, that for trading on the highest frequencies there is absolutely no way around looking very deeply into the particular microstructure of the market(s) in question.  

This article (and its references) could be a starting point:

http://www.cims.nyu.edu/~stoikov/LimitOrderBook.pdf

Knowledge of the particular market microstructure will amongst many other things be valuable in developing the order execution algorithms which are crucial for making HF work.


Since discussions on HF trading are few and far between (for obvious reasons) lets see if we can spark some renewed interest in this thread...

Title: Re: Ultra high frequency FX
Post by Algo Designer on 07/21/09 at 15:23:29


Nicolas Macherey wrote on 07/04/09 at 06:00:47:
Hi,

What kind of paradigm are you using for trading with ultra high frequency ?

I have implemented many tests that shows that it's not possible to trade forex with a frequency less than one hour, cause of slippages and commissions. It the same for many other stocks... More over liquidities are very difficult to manage in a UHF manner...


My experience so far (Nicolas, as I mentioned before, I use regression machinery similar to yours):

  • Sub 1m: Successfully picking up structures that are tradable, but slippage and commissions eat up 70 to 80% of returns. Very risky.
  • 1m: More structures emerge, the proportion of slippage and commissions reduces, but remains very volatile. Some of the patterns can be traded during the Asian hours.
  • 15m - 1m: This is my favourite frequency range! I like lots of small low-risk trades. They allow me to keep an eye on stats and timely detect regime shifts to either pull out or retrain.


Title: Re: Ultra high frequency FX
Post by Algo Designer on 07/21/09 at 15:36:26


BlackMage wrote on 07/21/09 at 14:19:49:

Nicolas Macherey wrote on 07/04/09 at 06:00:47:
What kind of paradigm are you using for trading with ultra high frequency ?

I have implemented many tests that shows that it's not possible to trade forex with a frequency less than one hour, cause of slippages and commissions.


My take on this is, that for trading on the highest frequencies there is absolutely no way around looking very deeply into the particular microstructure of the market(s) in question.  


There are funds out there consistently making money over the last few years whose trading algorithms can be scribbled in a few minutes on the back of a napkin. (They are probably smiling at me now ;-) )

Title: Re: Ultra high frequency FX
Post by lp on 07/27/09 at 16:18:44

Hi,

Can anyone recommend some good books or papers explaining how to use GA, Swarm Intelligence, ANN, Fuzzy Logic Systems and the likes to design a systematic trading strategy.

Also what kind of software is best suited for these applications, Matlab or similar?

Thanks for the help, I am quite new to this area.

lp

Title: Re: Ultra high frequency FX
Post by DerikB on 07/27/09 at 18:18:12

Hi,

Try

Neural Networks for Pattern Recognition (Paperback)
by Christopher M. Bishop

There is also another book with Matlab code. I couldn't find it but search amazon for neural network pattern recognition.

Regards
Derik

Title: Re: Ultra high frequency FX
Post by Nicolas Macherey on 08/03/09 at 15:20:49

Hi,

You can try :

+ Rocket Science for Traders and MESA from John F. Ehlers
+ Electronic and Algorithmic Trading Technology, The Complete Guide
from Kendall Kim
+ Analysis of Financial Time Series from RUEY S. TSAY
+ Fuzzy Sets and Fuzzy Logic Theory and Application from George J. Klir and Bo Yuan
+ Support Vector Machines from Ingo Steinwart Andreas Christmann
+ Bayesian Statistics is also interesting
+ ...

There are many books that can be read... And many papers can be found in the Electrical Engineering group from IEEE.

regards

Title: Re: Ultra high frequency FX
Post by Nicolas Macherey on 08/03/09 at 15:50:06


Algo Designer wrote on 07/21/09 at 15:23:29:
My experience so far (Nicolas, as I mentioned before, I use regression machinery similar to yours):

  • Sub 1m: Successfully picking up structures that are tradable, but slippage and commissions eat up 70 to 80% of returns. Very risky.
  • 1m: More structures emerge, the proportion of slippage and commissions reduces, but remains very volatile. Some of the patterns can be traded during the Asian hours.
  • 15m - 1m: This is my favourite frequency range! I like lots of small low-risk trades. They allow me to keep an eye on stats and timely detect regime shifts to either pull out or retrain.


Hi,
I can explain a little more what I mean by paradigm...
For example we can model the market using a predefined time frame and perform different algorithm analysis in order to take profitable (or trying to) decisions on a Close to Close manner without looking intrabar. The goal there could be to stay 100% of time in the market either Long or Short.

We can also model the market using a shorter time frame, to identify turn-key trading points. The goal is to take Long/Short positions to capture short term trends and to get out on predefined SL or TP targets. This can be used for example during market openning hours...

However, this can be achieve by identifying market partterns in technical indicators or in the market itself. Now, the problem is to manage the models stability... The paradigm used in your trading model can be very important. More over, even if your trading stratgey is modelled by a very small number of parameters you can have strong divergences between a backtest use and a Real Time use.  I know by experience that even if large unseen data sets are used for the validation process, this will not necessarily be significant for the future. (kind of cheat...)

regards

Nicolas

Title: Re: Ultra high frequency FX
Post by Pankaj Kumar on 08/28/09 at 08:24:18


Nicolas Macherey wrote on 04/07/09 at 10:29:08:
Hi,

I am developping automatic forex trading strategies in a middle High Frequency mode. (60 min to daily charts).
My business is oriented toward the developpement of fully automated trading strategies starting from historical price data and going to order generation. (Market neutral, Absolute returns...)
If you need any info...

I am working on my phd thesis which is model one such strategies.I would be obliged you you can share the details..my id is kumar.x.pankaj@gmail.com

Title: Re: Ultra high frequency FX
Post by lp on 08/29/09 at 15:41:03

Thanks guys for the books recommendations.

Title: Re: Ultra high frequency FX
Post by Cat on 09/07/09 at 11:08:03

I was recently invited to the following seminar - Dr. Aly Kassam — Implementing High Frequency Trading Algorithms Ten Times Faster.

Abstract

This presentation will discuss some of my investigations into the use of a set of visual programming tools to develop and deploy high frequency program trading systems. The tools are well developed and used in other engineering disciplines, but as yet relatively new for finance. In particular we will look at a new design paradigm — Model Based Trading — together with automatic HDL and C-code generation from models and deployment to specific hardware including DSP boards.

Speaker

Aly Kassam has had a long working relationship with The MathWorks Ltd — the creators of MATLAB — where he is currently a Principal Application Engineer for Computational Finance. Prior to this he worked for a couple of years as a program trader at Barclays Capital in London. He has a degree in Astronomy and Physics along with a Masters in Nonlinear Dynamics and Chaos from UCL in London, and a Doctorate in Numerical Methods for PDEs from the Computing Laboratory at Oxford University.

Should anyone have specific questions, I can put you in contact with the Thalesians who organised this event and I'm sure they would be happy to help.

thanks

Cat ;D

Title: Re: Ultra high frequency FX
Post by Yury R on 01/19/10 at 22:27:12

I just wanted to revive the thread. I am working for a high-freq shop myself at the moment but considering starting something of my own in some not so distant future.

I am very interested in this discussion if there is still anybody around.

Also, is anybody here who is doing quantitative trading/research in London area?

Title: Re: Ultra high frequency FX
Post by Algo Designer on 01/27/10 at 12:30:06

Hi Yuri,

I would like to revive this thread as well. I truly hope I will have a little bit more time for the forum this year! :-)


Title: Re: Ultra high frequency FX
Post by FXJunkie on 01/28/10 at 16:41:44

Very interesting site and discussion.... im looking for a FX savvy quants / programmer who can develop / build my vision but also add their own insightful input... I have access to some of the best fx pricing/execution in the world (stable .2 for EURUSD up $5)....but lack the math / programming sophistication....

Let me know if your up for the challange and adventure. For the right person £££ can be involved...

FXJUNKIE

ps- im hyperparanoid about hurting my currently position so bare with me 8-)

Title: Re: Ultra high frequency FX
Post by BlackMage on 01/31/10 at 16:23:06


FXJunkie wrote on 01/28/10 at 16:41:44:
I have access to some of the best fx pricing/execution in the world (stable .2 for EURUSD up $5)


I assume you mean up to 5 million $? What commission are we talking about?

I would generally be interested in knowing more about what's the best obtainable pricing. Currently the best I am offered is approx. 0.3 average on EURUSD  (ranging from 0-1.0) with a commission of 20$ pr. million (with potential volume rebates). This is for sizes in a couple of millions. While liquidity and execution looks good, 20$ a million seems a bit expensive (same commission is obtainable with retail brokers...).

Anyone got comments on this?

Another thing to look for is how badly the banks are able to screw you with "dirty tricks" on a particular platform ;)


FXJunkie wrote on 01/28/10 at 16:41:44:
ps- im hyperparanoid about hurting my currently position so bare with me 8-)


Understandable. The field of HF/auto trading is about as secret as it gets...









Title: Re: Ultra high frequency FX
Post by Nishant Ranajn on 02/03/10 at 06:15:30


FXJunkie wrote on 01/28/10 at 16:41:44:
Very interesting site and discussion.... im looking for a FX savvy quants / programmer who can develop / build my vision but also add their own insightful input...


... and I am a programmer looking to program some strategies :) ... i am currently developing very very basic strategies ... i am keenly looking for opportunities on learning business.

Title: Re: Ultra high frequency FX
Post by Algo Designer on 02/25/10 at 14:15:20

Nicolas, how are your AI strategies doing? Have you tried training and running them on bars shorter than 60m?


Title: Re: Ultra high frequency FX
Post by ManInMoon on 05/14/10 at 10:32:44

qroach,

I am in London area.

I am keen to discuss HF.


qroach wrote on 01/19/10 at 22:27:12:
I just wanted to revive the thread. I am working for a high-freq shop myself at the moment but considering starting something of my own in some not so distant future.

I am very interested in this discussion if there is still anybody around.

Also, is anybody here who is doing quantitative trading/research in London area?


Title: Re: Ultra high frequency FX
Post by AlgoHFT on 08/13/10 at 14:19:25

Hi,

New to this forum.

Is this thread still active?

J

Title: Re: Ultra high frequency FX
Post by Algo Designer on 08/13/10 at 22:57:24

It has been inactive for a while, but I would like to revive it.

Title: Re: Ultra high frequency FX
Post by Jeff S on 09/28/11 at 14:20:06

I'd also be very interested to see this thread become active again.

I'm very much a beginner when it comes to quantitative trading, but it seems to me that the profitability of HF strategies deployed in the FX market would be dependent upon superior execution algorithms. I backtested a very simple but very robust trading rule with 1-min currency data that showed excellent results initially, but proved to be an unmitigated disaster after slippage and transaction costs were accounted for.

The trading rule I used modelled the mean-reversion of the time series and, because the trading frequency was so high, the variance of the returns asymptotically approached zero. I couldn't (and still don't) understand how anybody could make money trading at such high frequencies when even this extremely profitable strategy wasn't able to stand up to the effects of slippage.

Title: Re: Ultra high frequency FX
Post by kamensk on 02/19/12 at 16:33:17

ultra high frequency  ? you talking about 10b/s data prosesing ?
how math money do i need to start in ultra HF ?
i think CME link form $600- $12000 ,
data enter server just rack space $125,
so so server SSD drives $5000,
CME lease seat chapest one idem-4 $70 just for DOW
or any good !! not scam for $ 1000 , currenex account starts from $50000 and up
. low latency data $1000 p/mo
Do you think possible to start Ultra HFT for litle less then $200000 ?
;D

Title: Re: Ultra high frequency FX
Post by 90delta on 02/29/12 at 18:25:47

Most HF FX strategies are liquidity providers as opposed to crossing the spread.  Triangular arbitrage opportunities occur so infrequently that it's near impossible to trade and even harder to build out an entire desk/strategy based on triarb opportunities.

Most firms are making markets on multiple legs and instruments and trading the bid/ask spread against each other (spot cash vs. equity vs. future).  

Title: Re: Ultra high frequency FX
Post by Algo Designer on 02/29/12 at 22:11:31

Absolutely agree with your comments.

kamensk, setting up an HFT desk will cost a lot more than USD200K.

Title: Re: Ultra high frequency FX
Post by FXHFT on 04/27/12 at 16:30:55


Cooper wrote on 04/06/09 at 13:46:12:
It seems to me that forex spreads are so wide as to make ultra-high frequency very difficult. I stick to medium-high frequency.


If you execute UHFT with low volume (range $1-$3 million), I can recommend you Tier2 with spreads 0.1-0.5 at the fifth decimal - only for institutional players of course). Send me PM.

Title: Re: Ultra high frequency FX
Post by FXHFT on 04/27/12 at 16:34:26


Algo Designer wrote on 02/29/12 at 22:11:31:
Absolutely agree with your comments.

kamensk, setting up an HFT desk will cost a lot more than USD200K.


Hum, read about http://www.streambase.com/products/streambasecep/streambase-studio/#axzz1t9LB4Qi3. A couple of large Tier1 are using it, some decent HFT FX hedge funds and some market makers - as far as I am informed.

The good thing is, you can change HFT strateges "on the fly" - means you don´t need a developer "department"...

Title: Re: Ultra high frequency FX
Post by DTD on 06/20/12 at 05:01:02

I was wondering why everyone only uses  bar data to generate signals here. Obviously there is the risk of over-trading when you sample at a much higher rate (say seconds) than your desired trading frequency but surely this can be accommodated by using EMAs?

For example, I am attempting to train a ANN to forecast the sign of the difference of an EMA which naturally doesn't change nearly as much as the underlying high frequency data. Yet I hopefully am not throwing out as much signal as simply sampling at 60min?  :-/

Algorithmic Trading Group Forum » Powered by YaBB 2.2.2!
YaBB © 2000-2008. All Rights Reserved.