Menu

Options

Bookmark Thread

First Page First Unread Last Page Last Post

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

Quoting Dezil
Hi, mikkom
Can you give some example of splitting the risk between 2 systems. For ex. one system generate 120 trades and another one generates 190 trades. Both sistems are profitable.
Let's call 120 trade system system A
190 trade system system B

This is far oversimplified but for if you want the systems to be weighted by trade count for example if you usually use 3% per trade for both systems, you could use 120/(120+190)*1.2% for A and 190/(120+190)*3% = 1.8% per trade for system B.

Or you could use 1.5% or weight it by some other way.

I'm very interested if someone has some thoughts about how to mathematically determine the optimal weighing for systems, I have personally used a brute force approach to go through different pairs and see how the systems work with different weight combinations.

I did an excel (well openoffice spreadsheet to be exact) to calculate this with random numbers with overall profit factor known, I'll see if I can find it somewhere.
Quoting aidans
Is there any correlation between Monday Bar Directon and week bar close ?

and How about little sunday bar (IBFX server) direction and Monday D1 Bar Close.
Very good question. I think I'll take a look at this. I'm quite sure there is a correlation but I'll do a chart.
I would suggest like this:

make simulation (or use real data if you have over some long time) for the Profit/Loss of each system. Draw this as statistical distribution (histogram) - this will be your guide for the weights choice.

Then create a formula (this needs a little research) but the idea is to use smaller weights to those systems with the bigger losses in their statistical distribution of PnL, and bigger weights to those with smaller losses. This way you minimise risk (variance of PnL) and produce smoother total PnL curve.

One idea is to create a linear combination of all PnL distributions and optimise the weights of this combination with optimisation target minimal loses (either minimal maximum loss or minimal total loss).

You will need to redetermine the weights of this formula every now and then as you take more trades, to keep up with the market changes.

PS some people use covariance matrix for this but I dont think its necessary or gives enough reward for the effort put in.

Quoting mikkom
...I'm very interested if someone has some thoughts about how to mathematically determine the optimal weighing for systems...
example:

for simplicity say you have 2 systems

PnL histogram, system 1 is values a1, a2, ..., a10 (if you have 10 bins in histogram)
for system 2 histogram values: b1, b2, ..., b10

a(i) and b(i), i = 1.. 10 are known values, for i = 1..5 negative values, i = 6..10 positive values.

weights you want to find: w1 for system 1, w2 for sys 2.

case 1: minimise max loss: max loss is a1 for sys 1, b1 for sys 2

so minimise over w1, w2 the quantity maxLoss = w1 * a1 * fa1 + w2*b1*fb1

case 2: minimise total loss

so minimise over w1, w2 the quantity totalLoss = w1 * (a1 * fa1 + a2 * fa2 + ... + a5 * fa5) + w2 * (b1 * fb1 + ... + b5 * fb5)

where fa(i) are frequencies of a(i) in histogram, same for fb(i), and you need to take absolute values of a(i) and b(i).

If you need to google more search for portfolio optimization. This is a vast and very specialised area, there are people who do just this as full time job, even teams of people doing this as full time job. You can get really sophisticated if you want and you have the resources. What I suggested here was extremely simple that can be implemented really fast and it's better than nothing. Don't take it as the best possible solution.

I hope this helps.
Hello everybody,

I began trading Forex since March this year. I use MT4 as a platform. I have demo account there. The system opened me 5000 dollars demo account. First month I completely ruined 2 accounts, so I lost 10000 dollars. Next month I decided do not pay attention to market news and to sophisticated technical analysis. I decided to make one or two trades a day only if I see a strong trend. And for every trade I made I immediately place stop and take profit orderes, so i could loose 250 dollars or win. By doing this simple thing I managed to increase my balance from 5000 up to 8600 dolars in two months. The ratio of my stops to profits are 1:2.
Also I trade between 12:00-13:00 New York time and do just one trade or two but in diffrerent currency pairs.
After placinf all my orders I quit Forex and come back next day to see what happened to my trades and make another trade if I see some good "lines" on the chart.
I am a novice in this Forex thing and would like you proffies to give me advice. Is it just luck or some sort of strategy? Should I keep trading this way or it is going to be some sort of time wasting?

Thanks,
Quoting milib
By doing this simple thing I managed to increase my balance from 5000 up to 8600 dolars in two months. The ratio of my stops to profits are 1:2.
Also I trade between 12:00-13:00 New York time and do just one trade or two but in diffrerent currency pairs.
After placinf all my orders I quit Forex and come back next day to see what happened to my trades and make another trade if I see some good "lines" on the chart.
I am a novice in this Forex thing and would like you proffies to give me advice. Is it just luck or some sort of strategy? Should I...
As this thread is mostly about quantifiable results I have to ask you have you backtested this method in non-biased way? If not, why not?

IMHO only way to be sure is to test the method yourself for a long period of time - if you don't know how to code your strategy, you can always use tools like forextester (google it). Most important thing is that you yourself are comfortable with the method and trust it.

Risk/reward ratios and sessions etc are pointless unless you yourself understand where the edge lies - there is no single yes or no answer to the question you ask.

I personally try to not interfere with my systems and only open/close/modify trades once a day (at the begin of new daily bar on my broker). In other times the computer could be switched off.
Quoting challenger78
If you need to google more search for portfolio optimization. This is a vast and very specialised area, there are people who do just this as full time job, even teams of people doing this as full time job. You can get really sophisticated if you want and you have the resources. What I suggested here was extremely simple that can be implemented really fast and it's better than nothing. Don't take it as the best possible solution.

I hope this helps.
Sure, I have quite a bit about portfolio optimization. Covariance matrix is not that hard to implement really but IMHO it's far too simplistic. If you have some fresh pointers to some relevant articles or studies please feel free to post some links.

Some people have argued that because the problem is so complex - correlations and system performances do change in fact quite dramatically over time - it's best to use brute force (genetic algos or such) to find the optimal combinations of algos and markets, do you have any thoughts about this?

(I personally decided to go with genetic algos in my weighting, basically calculating optimal sharpe ratio and used the fitness landscape to eliminate curve fitting as well as I could but this still requires myself to estimate where the fitness alndscape looks well and where it does not)
Quoting mikkom
If you know trade frequency and risk for each trade you know the maximum risk too.
First of all thanks a lot for starting this extremely interesting thread and project. I am myself in the project of implementing automated trading but with much less experience than you in trading and systems.

I'm wondering, in terms of risks, what you think of the black swan risk? One specific example I can think of is when you get a serious runaway market against you (say a central bank decides to suddenly act on its currency), so that your slippage is enormous. Doesn't it ruin your whole risk model and suddenly make your risk incredibly high and unpredictable?

In a completely distinct subject, I'd be interested to know how you get your 10 years of data? Did you gather them in an on-going way or do you have a way to collect them now for cheap?
What is the granularity of your data? Daily or lower?
I'm experimenting with daily indicators at the moment (trying to get into daytrading as well to diversify my strategies)

Has anyone experimented with this kind of method? Any success?

Below is an example of bollinger bands applied to only todays hours, I think this might be starting point for better pattern recognition than just plain candles.

Indicator included if anyone is interested
Attached Image (click to enlarge)
Click to Enlarge Name: intraday_bb.gif Size: 28 KB
Attached File(s)
File Type: ex4 mikko_bb_period.ex4   3 KB | 439 downloads
Here is my latest experiments with price action and ATR during each day.
Each lines shows certain fibo level of daily (5 days) ATR. When price touch the line it means that price make for ex 0.618 of 5 day ATR.
Attached Image (click to enlarge)
Click to Enlarge Name: ex1.gif Size: 29 KB
Here is an interesting article about high frequency trading

Traders Profit With Computers Set at High Speed

http://www.nytimes.com/2009/07/24/bu...pagewanted=all
Another interesting article about probably the most famous quant fund on planet

Jim Simons: Method of Operation

http://stableboyselections.com/2009/...-of-operation/

Full testimony [video] here

http://video.nytimes.com/video/2008/...es-simons.html
Quoting mikkom
Has anyone experimented with this kind of method? Any success?

Below is an example of bollinger bands applied to only todays hours, I think this might be starting point for better pattern recognition than just plain candles.
This reminds me of an old post from Bemac on a similar principle but using MA's... may be something helpful in there.

http://www.forexfactory.com/showpost...29&postcount=7
FUBAR...
There has been a lot of talk about "high frequency trading" and flash orders in the press, especially because goldman codes were stolen or something. Here is a very interesting chart on one possible way of how high frequency trading could work. Interesingly, this is almost riskless way to make money.

I don't blame people who want to get rid of this and claim this is illegal (front running).

[I have no idea where the chart is from, I found it plain like this]
Attached Image (click to enlarge)
Click to Enlarge Name: YeWxA.png Size: 140 KB
Quoting ak4x
This reminds me of an old post from Bemac on a similar principle but using MA's... may be something helpful in there.

http://www.forexfactory.com/showpost...29&postcount=7
thanks ak, I'll take a look at that.
Quoting mikkom
Here is an interesting article about high frequency trading

Traders Profit With Computers Set at High Speed

http://www.nytimes.com/2009/07/24/bu...pagewanted=all
It is discussed/dissected here:

http://www.quantnet.org/forum/showthread.php?t=4916
Quoting mikkom
There has been a lot of talk about "high frequency trading" and flash orders in the press, especially because goldman codes were stolen or something. Here is a very interesting chart on one possible way of how high frequency trading could work. Interesingly, this is almost riskless way to make money.

I don't blame people who want to get rid of this and claim this is illegal (front running).

[I have no idea where the chart is from, I found it plain like this]
The article is further deconstructed here:

http://www.reddit.com/r/business/com...actually_work/

It also seems to explain the origin of the diagram you linked to.
Quoting BlackMage
The article is further deconstructed here:

http://www.reddit.com/r/business/com...actually_work/

It also seems to explain the origin of the diagram you linked to.
Yup, it's where I found it [the link to image that is]. - Didn't read the comments though - thanks for noting that, there is some very good discussion there (usually reddit discussion about anything related to trading is quite bad quality)

ps- I'm a total reddit addict
Quoting mikkom
Yup, it's where I found it [the link to image that is]. - Didn't read the comments though - thanks for noting that, there is some very good discussion there (usually reddit discussion about anything related to trading is quite bad quality)

ps- I'm a total reddit addict
I'm a reddit addict too. Yeah, on reddit I haven't seen any quality discussions related to currency trading either. There is a forex subreddit, but it is mostly empty.
I don't know what time it is at your place, but right here it's 4:20.
Some more stuff about high frequency

http://arstechnica.com/tech-policy/n...ies-abound.ars
1 trader viewing now