Let's build together a system which makes 1000%/month | Page 14 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Bollinger Band: let's build a strategy together 35 replies

A Different Kind Of Challenge - From 1000 to >1000 0 replies

Let's build a trading system (EA) 8 replies

BELLA SYSTEM. Let's try it, let's improve it! 118 replies

Attachments: Let's build together a system which makes 1000%/month
Exit Attachments

Let's build together a system which makes 1000%/month

  • #261
  • Aug 30, 2014 12:58pm Aug 30, 2014 12:58pm
Quoting PipMeUp
There is a bug in this piece of code. If I get it ExtMapBuffer is the output. They immediately store the result after the 1st computation. An IIR needs some samples to stabilize. I discard the first period results before considering the filter stable. This is especially important with long term EMA. Yet only the first few samples should be wrong. At the time the filter reaches the first yellow vertical line it should be stable. Despite this I see no big difference with my code. I just rewrote the formula to have a single multiplication. ema <- ema...
Yes, well the fact that everyone has its own subjective charts, will not ruin anything, it is just that everyone will have different prices at the same time, but they can only trade their own prices, so its not really a difference.Besides the bigger LP you have the closer your bid/ask is to the average forex prices.

Also if your broker is on holiday while other brokers are active, its just a bad luck, so i guess subjectivity is again very underestimated in trading
"There's a sucker born every minute" - P.T. Barnum
  • #262
  • Aug 30, 2014 1:07pm Aug 30, 2014 1:07pm
I think you should be aiming for a system that uses many pairs over a longer time frame. Yes I know that by trading a basket you are actually only trading the currency exposure from that basket, but even so a basket will let you trade profitably over longer time periods. The reason I trade only using a basket is because you can't predict the news, and any significant unexpected news could cause a significant loss if it is against the trade direction.

The 2nd reason I suggest a basket is because with a basket it is much easier to generalise the data to predict probable outcomes for a given set of trades. If you are only trading a single pair, you basically have to develop a system that only works with that pair. Looking at some of your previous threads I see that you have experience with genetic algos etc. I imagine that with your GA experience you have successfully been able to make great backtests only to find forward testing failure, this is the biggest problem that a ga introduces. So, my question if you are planning to trade a single pair is, how do you find a strategy that is profitable over a long period of time? Any backtest you do could be either down to overfitting data or luck, or an actual profitable solution, but until you forward test you would not know.

It's not all bad news, I will give you a small insight into how I trade. I use a GA that runs constantly on one of my gpu machines. My GA is basically looking for a set of lot sizes that when applied to the 28 pairs I trade (I only trade 28 pairs at once) has produced a minimum target % profit (including spreads/costs) for every day over the past 100 days. I am not using daily prices, I use ticks that are subsampled at a daily rate for the past 100 days. This way I can close the basket once it is in profit and open the next basket based on the last 100 days from that second in time. Each basket has slightly different lot sizes based on the results of the GA. With a system like this you could reach your 10% per day target, and the benefit of this type of system is that you are reasonable covered against any type of news event. The system is not perfect, about 98% of baskets close in 12 hours in profit, and the losses are not significant.

Just food for thought...

Attached Image (click to enlarge)
Click to Enlarge Name: forwardTest2014.png Size: 29 KB


The above is a walk forward test of 140 days, each day trading a basket based on lot sizes calculated over the past 100 days.
  • #263
  • Aug 30, 2014 2:03pm Aug 30, 2014 2:03pm
Quoting beachloup
I think you should be aiming for a system that uses many pairs over a longer time frame. Yes I know that by trading a basket you are actually only trading the currency exposure from that basket, but even so a basket will let you trade profitably over longer time periods. The reason I trade only using a basket is because you can't predict the news, and any significant unexpected news could cause a significant loss if it is against the trade direction. The 2nd reason I suggest a basket is because with a basket it is much easier to generalise the data...
I want to trade a basket, thats not the issue here, by trading the basket you minimize the risk ,that is the "long tail" unexpected events,with diversification so in average you mostly get near median events and results on your equity curve.

Its like a stock index, which smoothens out the huge volatility and unexpected news from individual stocks, by combining them in a smooth upward going trendline.

The problem is that we still need individual analysis, i cant just add random pairs to the basket, because the total edge of the basket will be the sum of the individual edges of the basket.

So if either 1 currency pair is not profitable in the long term, why should we include that at all? Because it will only cause unnecessary DD.

So to build a basket, we still need to analyze each currency pair by itself, and then construct the basket.But then we are back at square 1, which is to build our trend following method.

So lets do that people
"There's a sucker born every minute" - P.T. Barnum
  • #264
  • Aug 30, 2014 3:29pm Aug 30, 2014 3:29pm
Quote
So to build a basket, we still need to analyze each currency pair by itself, and then construct the basket.But then we are back at square 1, which is to build our trend following method.
The idea behing basket trading is to find cointegrated instruments in such a way that the basket never trends. The problem in FX is that the currencies are strongly correlated but not necessarily cointegrated.
No greed. No fear. Just maths.
  • #265
  • Aug 30, 2014 3:50pm Aug 30, 2014 3:50pm
Quoting PipMeUp
{quote} The idea behing basket trading is to find cointegrated instruments in such a way that the basket never trends. The problem in FX is that the currencies are strongly correlated but not necessarily cointegrated.
Yes this is one approach. Finding cointergrated baskets is the first problem, but it solvable. You then need to know when to enter and exit the basket, this is normally based on some std dev or some other measure. This is not so easy to solve and whilst profitable can lead to large drawdowns whilst you wait for the basket to revert back towards the mean.

Most people agree that trend following solutions work. Indeed, in the past trend followers have worked, but have they worked over the past 2 years? Once upon a time I thought trend trading was the way forward. I decided that the way to determine a trend would be the based on the slope of moving average or filters filtered at different intervals. In effect I was trying to reproduce how people look at charts and determine trends. For example I look at a chart with a sma on it and guage the trend based on the slope of the sma. So my system was based on lots of averages and lots of slopes fed into a big neural network and run for days and days. I got to the point where I could make it profitable in backtesting but not in real life. I thought the profitability was based on the ann detecting the trend and trading it correctly, but actually is was just working on overfitted data and failed out of sample.

This brings me back to the same old question. If you develop a trend following system based on past data how do you know it will work moving forward?
  • #266
  • Aug 30, 2014 3:50pm Aug 30, 2014 3:50pm
Quoting PipMeUp
{quote} The idea behing basket trading is to find cointegrated instruments in such a way that the basket never trends. The problem in FX is that the currencies are strongly correlated but not necessarily cointegrated.
When I dipped into baskets I found this very same thing. The basket is not supposed to trend, but it does.

This leads to overexposure, which leads to account damage.
If you are good at something, never do it for free--Joker
  • #267
  • Aug 30, 2014 3:51pm Aug 30, 2014 3:51pm
Quoting Proximus
{quote} I want to trade a basket, thats not the issue here, by trading the basket you minimize the risk ,that is the "long tail" unexpected events,with diversification so in average you mostly get near median events and results on your equity curve. Its like a stock index, which smoothens out the huge volatility and unexpected news from individual stocks, by combining them in a smooth upward going trendline. The problem is that we still need individual analysis, i cant just add random pairs to the basket, because the total edge of the basket will...
So Prox, to build a system we must first add definitions, so lets define trend. What is a trend to you?
If you are good at something, never do it for free--Joker
  • #268
  • Aug 30, 2014 3:52pm Aug 30, 2014 3:52pm
An interesting take on the trend following turtle system. http://financialfreezeframe.wordpres...rs-dream-blog/

Looks like it was profitable for the original turtles but the system has lost it's edge.
  • #269
  • Aug 30, 2014 3:57pm Aug 30, 2014 3:57pm
Quoting PipMeUp
{quote} The idea behing basket trading is to find cointegrated instruments in such a way that the basket never trends. The problem in FX is that the currencies are strongly correlated but not necessarily cointegrated.
All markets in a way are correlated, the flow of money causes and is the cause of many things.

I`m pretty sure that the correlation between USD/RUB and NZD/CHF is small, so if a trend emerge on these currencies, will probably have different causes,so that itself can help us diversify.The smaller the correlation the better the diversification.

Also the basket doesnt have to be fixed, it should contain only the best pairs which can be traded at the moment once our edge is gone in 1 pair, then we should exclude that pair from our basket until our edge gets back.

Quoting beachloup
If you develop a trend following system based on past data how do you know it will work moving forward?
You dont, thats why we need to find the cause always, naked punting is for idiots as the saying says

If the cause is found, then until the cause persists, we can safely trade it, once that cause is gone, our position should be closed.
"There's a sucker born every minute" - P.T. Barnum
  • #270
  • Aug 30, 2014 3:59pm Aug 30, 2014 3:59pm
Quoting jmn5611
{quote} So Prox, to build a system we must first add definitions, so lets define trend. What is a trend to you?
I already defined it a few posts back, but its not the definition is the problem but the detection, or better yet, the forecast of the trend, now that would be a nice one.
"There's a sucker born every minute" - P.T. Barnum
  • #271
  • Aug 30, 2014 4:03pm Aug 30, 2014 4:03pm
Quoting Proximus
{quote} I already defined it a few posts back, but its not the definition is the problem but the detection, or better yet, the forecast of the trend, now that would be a nice one.
If the definition of trend is already defined, why not enter when that definition is met?
If you are good at something, never do it for free--Joker
  • #272
  • Aug 30, 2014 4:42pm Aug 30, 2014 4:42pm
Quoting jmn5611
{quote} If the definition of trend is already defined, why not enter when that definition is met?
Because you dont know when it will end.Just because it is trending now doesnt mean that it will be trending always, and most times a trend ends violently, crashing and you could lose all profits if you dont exit before the trend ends.

So the cause of the trend must be known, to check whether its still available or not.

You agree that naked punting is for idiots, so lets not punt naked
"There's a sucker born every minute" - P.T. Barnum
  • #273
  • Aug 30, 2014 6:35pm Aug 30, 2014 6:35pm
Quoting Proximus
{quote} Because you dont know when it will end.Just because it is trending now doesnt mean that it will be trending always, and most times a trend ends violently, crashing and you could lose all profits if you dont exit before the trend ends. So the cause of the trend must be known, to check whether its still available or not. You agree that naked punting is for idiots, so lets not punt naked
Two points.

1. Naked punting means directional trading without a hedge. That, is for idiots according to the video.

2. Most trends do not end violently. Most trends, if not all, will have an attempt the highest high of the run with subsequent failure. There is also a rollover of most momentum indicators attached to the chart.

Are you trying to say that you want to know precisely when a trend begins and ends?

You are not trying to say that, are you?
If you are good at something, never do it for free--Joker
  • #274
  • Aug 30, 2014 6:50pm Aug 30, 2014 6:50pm
Quoting jmn5611
{quote} 2. Most trends do not end violently. Most trends, if not all, will have an attempt the highest high of the run with subsequent failure.
I'm quite intrigued by some of the "core topics" that are being mentioned in this thread.

I can express what I see happen time and time again - currency pairs clarify their upside when moving down, and clarify their downside when moving up. (I can share data to support this...)

Just as an example we'll use a random number on the AUD/USD of .9366 established as what I would call a "long reversal target", or, probability is there for the pair to reverse short once price moves LONG to the level of .9366.

What happens more often than not is say, .9362 would be hit and what looks like a reversal would happen, but this isn't a reversal, it's just "steam" being let off so momentum can build for .9366 to be hit. A lot of times, this level appears as "hit" on a wick on near about every time frame.

Through this research, my stance has become:

- this market is just like a rubber band - the further in a direction it goes(rubber band being stretched), the more volatile the retrace will be(once the rubber band is released)

- the longer things stay the same (keep moving in one direction), the more they need to change (reverse).
"Holy Grail" exists - accepting where is the first step.
  • #275
  • Aug 31, 2014 1:26am Aug 31, 2014 1:26am
Quoting VaBikePacker
{quote}- this market is just like a rubber band - the further in a direction it goes(rubber band being stretched), the more volatile the retrace will be(once the rubber band is released) - the longer things stay the same (keep moving in one direction), the more they need to change (reverse).
I find this somewhat not true.
UJ went 77 to 104 2.7k pips without a correction. that's significant drive.
Same on Aud/NZD
Both had significant fundemetal shifts.There are more examples.I try to focus more on pairs that range more often.
regards
In trading, you have to be defensive and aggressive at the same time
  • #276
  • Aug 31, 2014 4:07am Aug 31, 2014 4:07am
Quote
When I dipped into baskets I found this very same thing. The basket is not supposed to trend, but it does. This leads to overexposure, which leads to account damage.
If the basket starts trending it shall be rebalanced. If despite the rebalancing it keeps trending, the root assumption "these instruments are cointegrated" doesn't hold any longer, temporarily or permanently. There is no reason for over-exposure you take the loss because your are proven wrong about the cointegration assumption which made you enter the market. Normally the loss is bounded to N standard deviations plus the cost of building and balancing the basket (spreads + commissions).

Quote
Most people agree that trend following solutions work. Indeed, in the past trend followers have worked, but have they worked over the past 2 years?
Quote
So Prox, to build a system we must first add definitions, so lets define trend. What is a trend to you?
THE key question IMO. For what I could define so far, it is simply the drift of the mean at the time scale (not timeframe) of your choice. So not 36 MAs, one MA. YOUR MA. To me trend following is NOT about staying long as long as the trend is up. It is about following what the drift of the mean of the price vs the position of the price with respect to this mean tells. Therefore, to me at least, range trading IS trend following: the trend is flat and you follow this fact. Given this assumption and until it is disproven, the trading is a mean reversion trading. Actually trend trading is also a mean reversion trading. The difference is simply that the trend element has a significantly non null value. But it is not traded exactly the same way because here you don't want be flat at the mean, you want be flat at the excess above the mean. What I'm missing for now is a formula of the lot size exposure which could link both.
Attached Image (click to enlarge)
Click to Enlarge Name: trend following.png Size: 43 KB


Quote
All markets in a way are correlated, the flow of money causes and is the cause of many things. I`m pretty sure that the correlation between USD/RUB and NZD/CHF is small, so if a trend emerge on these currencies, will probably have different causes,so that itself can help us diversify.The smaller the correlation the better the diversification. /.../ Also the basket doesnt have to be fixed, it should contain only the best pairs which can be traded at the moment once our edge is gone in 1 pair
You don't choose the instruments that form your basket based on your liking. They have to cointegrate. A well known method is "producers-consumers" relationship. For instance oranges producers and orange juices makers, oil vs airlines, steel vs automotive... But nowadays those companies hedge themselves!

Quote
If the cause is found, then until the cause persists, we can safely trade it, once that cause is gone, our position should be closed.
Read again (and completely!) your own link about Recency Bias. A retracement is not the end of the trend. The cause of the uptrend never stopped. Yet 20 out of 30 traders in the toy example were IN THE LOSS.

Quote
but its not the definition is the problem but the detection
Isn't it the same problem to solve?


I also tend to agree with Davit that the rubber band effect is not so true. At least the rubber band seems to break when strechted to much.
No greed. No fear. Just maths.
  • #277
  • Edited 5:17am Aug 31, 2014 4:54am | Edited 5:17am
Quoting Davit
{quote} I find this somewhat not true. UJ went 77 to 104 2.7k pips without a correction. that's significant drive. Same on Aud/NZD Both had significant fundemetal shifts.There are more examples.I try to focus more on pairs that range more often. regards
There was a dip previous (based on monthly) to that indicating the possibility of what you're describing. Depending on your willingness to take a large swing, get your best shorts in order as the pair approaches 110 and the possibility of it reaching as high as 112 is certainly "on the table". Note that this is based on a weekly chart, so, it definitely will take some time for this to materialize....it has taken about 3 years for this to materialize so far..

It's possible to give this a "higher resolution" by doing this on a smaller time frame. There may be a shorter time frame move indicating high probability short trades in the current area. Again, not a pair I focus on, so, you might take what I'm suggesting with a grain of salt unless you want to do the same analysis I do.....

I apply it to 30m.

data proving the concept on AUD/USD:

https://secure.trackvia.com/app/shar...e0f11904799198

data proving the concept on GBP/AUD: (just started applying the method to this pair recently, thus the short list - keep the link though - it'll grow..):

https://secure.trackvia.com/app/shar...e7fd76ccf0a987

You'll notice the dates on this list to be from 2009 and apply to the EUR/USD:

https://secure.trackvia.com/app/shar...73beb5ccf0a987

This is an idea I've worked on for quite a while...and came to the conclusion it was time to start trading it live to discover more...

After all this work, 100% of the predicted targets have been reached.

...the truth is in the numbers....you've just gotta look at the numbers the right way....then apply patience....
"Holy Grail" exists - accepting where is the first step.
  • #278
  • Aug 31, 2014 5:22am Aug 31, 2014 5:22am
Quoting VaBikePacker
{quote} After all this work, 100% of the predicted targets have been reached. ...the truth is in the numbers....you've just gotta look at the numbers the right way....then apply patience....
I'd really like to know how you do that VBP. You've posted these before and it's quite impressive.

By these numbers, I'm assuming each new entry is at the previous target hit?
Ghost Rider - WWTBMD?
  • #279
  • Aug 31, 2014 5:32am Aug 31, 2014 5:32am
Quoting albchr
{quote} I'd really like to know how you do that VBP. You've posted these before and it's quite impressive. By these numbers, I'm assuming each new entry is at the previous target hit?
yes. Correct. I don't place trades based on a chart/trendlines/elliott/fibo/stoch/MAs or anything like that.. I place trades based on research of patterns of the pair and the numbers I arrive at.

SO: you asked "how":

- track 30m data from 0000 GMT Monday 'til market close on Friday in a spreadsheet
- measure the distance from previous 30m open to current (scalptunity comes from this..)
- measure the distance from 0000 GMT of that day and 0000 GMT Monday (giving clear view of range of the pair over a weeks time)
- a combination of the distance from daily to lowest and/or highest open of the day develops potential reversal targets.
"Holy Grail" exists - accepting where is the first step.
  • #280
  • Aug 31, 2014 5:43am Aug 31, 2014 5:43am
Quoting VaBikePacker
{quote} yes. Correct. I don't place trades based on a chart/trendlines/elliott/fibo/stoch/MAs or anything like that.. I place trades based on research of patterns of the pair and the numbers I arrive at. SO: you asked "how": - track 30m data from 0000 GMT Monday 'til market close on Friday in a spreadsheet - measure the distance from previous 30m open to current (scalptunity comes from this..) - measure the distance from 0000 GMT of that day and 0000 GMT Monday (giving clear view of range of the pair over a weeks time) - a combination of the distance...
I shall try this. Thank you VBP.
Ghost Rider - WWTBMD?
2 traders viewing now