Systematic trading | Page 30 | 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

  • #581
  • Mar 2, 2010 7:15am Mar 2, 2010 7:15am
Quoting takef586
I do not want to repeat myself too much, but the only ECN retail traders can access to trade forex are currency futures. In this case your orders are actually being matched against other orders or against streams in a central matching engine with no chance of undoing the trade. In all other cases, the so called ECN's are just smoke in your eyes, even if they can resemble them. I traded on Interactive Brokers FX, not IBFX, sorry.
Do you (or someone else) know about specifics on how matched orders are cancelled afterwards in IDEALPRO? I'm quite interested in this because I have previously assumed it would not be possible (I'm not doubting you, just interested in this because I did trade with IB about a year ago and have thought of moving back).
  • #582
  • Mar 2, 2010 7:25am Mar 2, 2010 7:25am
This is simple. When you have a limit order out, or you "go at the market" in an "apparent" market, what really happens, is that you ping the best or the matching price for a confirmation. In some cases, a price has a "lifespan" and this automatically generates a trade if you "hit" the price within e.g. 50 or 100 msec since it got emitted. In most cases though, the deal confirmation is subject to "approval" from the bank that generated the price. It is easy not to grant the approval, if your refusal window is, say 300msec. You simply verify what the next price inside your engine is, and if it is worse for you, the liquidity provider, then nothing done, if it is better, then you confirm the trade. In practical terms, this means that a trader's stops get more slippage and his profit taking orders at times do not get executed.
  • #583
  • Edited 8:44am Mar 2, 2010 7:35am | Edited 8:44am
Quoting takef586
This is simple. When you have a limit order out, or you "go at the market" in an "apparent" market, what really happens, is that you ping the best or the matching...
Do you know if this possibility is just for banks or other institutional (/retail? - assumign the same latency which is of course ridicilous) clients as well or do you mean that all IDEALPRO orders are always routed first to banks?

I haven't seen any document on how IDEALPRO has been implemented so I really have no idea on how they operate (I thought I did) so I'm very interested.

edit: Okay found it.

http://www.interactivebrokers.com/en...F-FXTrader.php
Quote
IdealPro is actually a SMART routed venue where we consider each liquidity provider as an exchange.)
Now I totally understand what takef means, many thanks about this. I have always thought IDEALPRO is the central exchange and providers just quote prices as part of the other traders but it seems that is not the case.
  • #584
  • Mar 2, 2010 7:50am Mar 2, 2010 7:50am
On platforms like this, only the end users are actually being matched against each other, while the liquidity providers typically have the right to reject and typically they inhibit trading among themselves. It can be different e.g. on Hot Spot Pro, where you actually can decide if you want to trade against people who have the right to reject or not, and some banks go there to trade on purpose, as the platform is anonymous and they can get some extra liquidity.
  • #585
  • Mar 2, 2010 8:29am Mar 2, 2010 8:29am
Quoting takef586
On platforms like this, only the end users are actually being matched against each other, while the liquidity providers typically have the right to reject and typically they inhibit trading among themselves. It can be different e.g. on Hot Spot Pro, where you actually can decide if you want to trade against people who have the right to reject or not, and some banks go there to trade on purpose, as the platform is anonymous and they can get some extra liquidity.
On some platforms you can explicitly flag your order not to be filled by a main liquidity provider (bank) but only by a "real" market participant trading on that platform. However you have no real idea about the actual liquidity then - you more or less have to view it as a dark pool (or that's at least the the way I see it).
  • #586
  • Mar 2, 2010 3:28pm Mar 2, 2010 3:28pm
What about something like this, where the broker is your agent (and obligated by contract to give you best execution possible):

Quote
http://www.fxcmpro.com/product-services
Act as a price taker or price maker.
Access Executable Streaming Prices (ESP) or Request For Quote (RFQ).
Robust order book with zero latency and no last look, allowing clients to cross with other client interest
How will they scam you in your opinion (because you said that CME is the only non-scam posibility for a retailer).
  • #587
  • Mar 2, 2010 3:59pm Mar 2, 2010 3:59pm
Simple. Anything which is not going to be matched against another guy plugging in an order, i.e. 99.99% of your trading volume, will still be going to be subject to the "approval" of the market maker. FXCM is a known bucket shop BTW, so there is no point in going into depth on this. They are famous for "market mining" with invented spikes to trigger your stops. This is a convenient way to send you to the cleaners fast - if you had a stop and the spike catches you, then you are done, if you had a limit order, then nothing done - sorry it was a mistaken quote...
  • #588
  • Mar 2, 2010 4:59pm Mar 2, 2010 4:59pm
Quoting takef586
Simple. Anything which is not going to be matched against another guy plugging in an order, i.e. 99.99% of your trading volume, will still be going to be subject to the "approval" of the market maker. FXCM is a known bucket shop BTW, so there is no point in going into depth on this. They are famous for "market mining" with invented spikes to trigger your stops. This is a convenient way to send you to the cleaners fast - if you had a stop and the spike catches you, then you are done, if you had a limit order, then nothing done - sorry it was...
This was about FXCM Pro, not FXCM. Executable Streaming Prices are by definition "not subject to approval".

You say that 99.99% of the market is made by the market makers, and only 0.01% of the market are clients, right?

And stop hunting is the name of the game in forex, you don't need spikes for that. If it's a spike, you can audit and revert your order. If it's a real stop hunting expedition by the market, then tough luck. Personally, I love it when USD/JPY goes for the big stops once a month. Risk free money for me.
  • #589
  • Mar 3, 2010 2:15am Mar 3, 2010 2:15am
I think that's enough of broker talk, all in all, takef is totally correct on this one - and bucketshops are a lot worse than ECNs like idealpro and it seems even idealpro seems to allow frontrunning in some cases. Read all about the recent flash order fiasco to understand why this is so important (there was interesting article somewhere, I don't remember where how flash order ban instantly affected mean reversion) - and in forex there is no regulation so this is even worse.

So to other things.. I have been thinking of doing a little statistical study on breakouts for a long time. I will post the results here if I get good ideas on what should I study so post your thoughts

I trade a lots of different kinds of breakouts so I would like to quantify situations and patterns pre/after breakout to see if there are some market conditions where breakout is successful (good definition might be that the days close is above/below breakout) and what conditions are more likely to cause false breakout. I have a quite a lot of ideas on what I'm going to research but I'm looking for more so post your thoughts if you have any.
  • #590
  • Mar 3, 2010 4:08am Mar 3, 2010 4:08am
As you probably know I am focusing on intraday. I started out with breakout systems. Since then I have developed a number of other strategies. Surprisingly (to me anyway) the breakout system I currently have is actually quite disappointing compared to the others. Maybe it is because so many people use breakout systems as they are probably the easiest to program.
  • #591
  • Mar 4, 2010 1:49am Mar 4, 2010 1:49am
GA Update, finding the aforementioned bug has forced me to completely reevaluate some of my optimization criteria. As it turns out, some of my criteria to prevent over-fitting where far too strict and actually having the reverse effect. I now have a much better appreciation of the subtly of the GA fitness measure. Hopefully I can post some results to back up this work soon.
The breaking of a wave cannot explain the whole sea.
  • #592
  • Mar 4, 2010 6:39am Mar 4, 2010 6:39am
Quoting Craig
GA Update, finding the aforementioned bug has forced me to completely reevaluate some of my optimization criteria. As it turns out, some of my criteria to prevent over-fitting where far too strict and actually having the reverse effect. I now have a much better appreciation of the subtly of the GA fitness measure. Hopefully I can post some results to back up this work soon.
i am looking forward to those results
I'm not insane; my mother had me tested.
  • #593
  • Mar 5, 2010 6:07am Mar 5, 2010 6:07am
This book turned up on my radar:

http://www.algo-dma.com/

Worth noting is that it got a good review on Amazon by Aaron who amongst other things is a highly regarded member of NuclearPhynance.
  • #594
  • Edited 1:01pm Mar 5, 2010 11:24am | Edited 1:01pm
That books looks good, I'll be checking that out.
Anyway, here is another GA walk-forward for about the past 3 months.
There is promise, but the system needs more refining, which I shall continue to do.
Attached Image (click to enlarge)
Click to Enlarge Name: Screenshot-1.png Size: 9 KB
The breaking of a wave cannot explain the whole sea.
  • #595
  • Edited 12:59pm Mar 5, 2010 12:23pm | Edited 12:59pm
Here is another one which completely bombs with exactly the same approach, why? I have no idea as of yet.
Attached Image (click to enlarge)
Click to Enlarge Name: Screenshot.png Size: 9 KB
The breaking of a wave cannot explain the whole sea.
  • #596
  • Mar 6, 2010 4:36am Mar 6, 2010 4:36am
Quoting Craig
Here is another one which completely bombs with exactly the same approach, why? I have no idea as of yet.
For some reason EURUSD is (in my testing) by far the easiest pair to optimize to. That's one of the reasons why I find it very important to optimize to multiple pairs simultaneously (I usually use 4-6 pairs).

That also helps a lot with curve fitting because it's not easy to curve fit to multiple pairs simultaneously, there are too many variables. That's also one of the reason why I always want to use very long optimization periods - it helps to have multiple types of markets in the data. That way you will have a combination of systems that will work okay or lose little in bad markets and do well in good markets, combined curve will be good.

I don't think how far I want to go to this but I also found out that it's not good idea to calculate direct correlations between systems but check the result curve of combined systems, that in itself explicitly includes the correlation check in almost optimal way.
  • #597
  • Mar 7, 2010 9:10pm Mar 7, 2010 9:10pm
Quoting mikkom
For some reason EURUSD is (in my testing) by far the easiest pair to optimize to. That's one of the reasons why I find it very important to optimize to multiple pairs simultaneously (I usually use 4-6 pairs).
I'm also finding this as well, strange...I wonder what this implies? Is the EUR/USD more amenable to TA, on the face of it this would seem like a strange conclusion given that it is the most traded market and hence (according to common wisdom) should be the most efficient market. I notice that EUR/USD optimizations tend to generate more trades in the training sample than other markets, I wonder if this is some kind of 'law of large numbers' thing happening. I have attached a composite EUR/USD & GBP/USD walk-forward result, I'm going to do some real money forward testing this week with small size, should be interesting

Edit: Given how well this approach seem to work with EUR/USD, I wonder how it would go on the more liquid futures markets.
Attached Image (click to enlarge)
Click to Enlarge Name: Screenshot.png Size: 8 KB
The breaking of a wave cannot explain the whole sea.
  • #598
  • Edited 1:52am Mar 8, 2010 1:25am | Edited 1:52am
I think this has very much to do with stop (including tp) placements which is at least for me very important aspect. EURUSD cumulates more stops (some people might call it order flow) to certain levels so randomity on those levels is less than it would be on less traded pairs.

edit: This might be of interest http://www.newyorkfed.org/research/s...orts/sr150.pdf

Quoting Craig
Given how well this approach seem to work with EUR/USD, I wonder how it would go on the more liquid futures markets.
That's a very good question, I have also been pondering about testing on very large set of stocks + futures. Because curve fitting to that size of sample is almost impossible the results should be quite reliable. Problem being of course that the larger the set, the slower the process. Also the correlation is not that obvious as it is between currency pairs so that should be a big plus.
  • #599
  • Edited 2:53pm Mar 8, 2010 2:40pm | Edited 2:53pm
Here are the results of todays live forward test, with the EUR/USD & GBP/USD rules generated yesterday.
Edit: ...of course MBs awesome FIX gateway sucked the big one at the end of the session when I tried to close the last trade, so you'll just have to imagine a ($1.20) loser on the end.
Attached Image (click to enlarge)
Click to Enlarge Name: Screenshot.png Size: 8 KB
The breaking of a wave cannot explain the whole sea.
  • #600
  • Mar 9, 2010 5:49am Mar 9, 2010 5:49am
Craig, you need to ditch MB
0 traders viewing now