Let's build together a system which makes 1000%/month | Page 13 | 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

  • #241
  • Aug 28, 2014 4:15pm Aug 28, 2014 4:15pm
Quoting Unabletrader
{quote}​ Huge mistake. Time is a fundamental concept of the auction process.
I agree only if we should treat every timeframe equally important, i mean MT4 is divided into basic TF like 30 min , 60 min, 4 hour ,etc

Why this one ? I can be the also 27 min or 13 min or 65 min charts, the 30 and 60 is a nice round number but in reality it doesnt mean jackass.
Heck it can be 2 hour 7 min 18 second timeframes aswell, it doesnt matter

Institutional platforms rarely have a chart, they only look at latest bid/ask and the already placed liquidity at nearby levels.So the big traders dont care about time as we do, because they put the trade and exit when they want, only we the retail herd need to follow that order and ride it until its done.

Trading on timeframes like discriminating 4 hour chart over 15 min is only a retail fantasy ,in reality they are all equal.

So to avoid this, all analysis must be done on the same chart timeframe, for the most correct analysis the M1 or Tick chart if available, and then the trader should switch to his "preferred" timeframe, but he will have the precise data from the "basic timeframe" already.
"There's a sucker born every minute" - P.T. Barnum
  • #242
  • Edited 8:04pm Aug 28, 2014 4:55pm | Edited 8:04pm
Quote
Institutional platforms rarely have a chart, they only look at latest bid/ask and the already placed liquidity at nearby levels.So the big traders dont care about time as we do, because they put the trade and exit when they want, only we the retail herd need to follow that order and ride it until its done.

It's not really true.

Microstructure traders don't really care about charts. But some instutionnals traders or fund managers use usual charts as well some retail traders don't need to use charts because they are microstructure traders too.

My point is that it's not about the type of investor you are (retail, institutionnal, well-capitalized, whatever) but more about your trading style.

If I understand well, you guys are using usual charts, so let me to tell you that timeframe is important in order to sequensize the auction. And you need to if you consider time, it's sth every trader should do imo.

Excuse-me for my poor english, I'd like to tell more about but it's quite difficult.
  • #243
  • Aug 29, 2014 4:56am Aug 29, 2014 4:56am
Quoting Proximus
a logical fallacy called: Recency Bias Since i learned this, i dont use anymore exponential or any other weighting only linear (Simple Moving Average).
From this text I understand that
- You can't walk straight forward if you stare at your feet: the ticks are the hairs on the elephant.
- The author can't count. If the market goes up 3% per month during 18 months then goes down 2% per month during 12 months the net result is +33.6% (1.03^18*0.98^12=1.336) not 12.28% as written in the text. (He says himself that I shall check the expectise of the expert ^^)
- Time scale is very important because being "right" on the direction is just not enough: they were all long in a market that rose 60% of the time while only one third of them made money. 20 out of the 30 are in the loss.
- If instead of faithfully following their beliefs that SPX must yield 12% per year they simply followed the market, guest 22 who is down 16.6% would have shorted the bear market during 9 months and made +19.4%; assuming 3 months of lag to recognize the down trend.
- The importance of the management of the positions is far beyond the importance of the entry. If Guest 1, who made 34%, recognized that after 3 fulls months of -2% he shall cut his long position he would have made +60.5% and would be now ready to re-invest!

What I don't understand is how a behavioral biais and an EMA can be linked together through this text.
No greed. No fear. Just maths.
  • #244
  • Aug 29, 2014 10:27am Aug 29, 2014 10:27am
Quoting PipMeUp
{quote} From this text I understand that - You can't walk straight forward if you stare at your feet: the ticks are the hairs on the elephant. - The author can't count. If the market goes up 3% per month during 18 months then goes down 2% per month during 12 months the net result is +33.6% (1.03^18*0.98^12=1.336) not 12.28% as written in the text. (He says himself that I shall check the expectise of the expert ^^) - Time scale is very important because being "right" on the direction is just not enough: they were all long in a market that rose 60%...
Wow i didnt read the whole text there, only the top row where it explains how behavioral bias destroys traders.It can have many types but i think the recency bias is the most common.

Now look in "news time" it might not be true, because after a news event, the impact decays probably in an exponential or atleast a non linear way, the farther you go away from the exact release time of the news, the less impact it will have on the markets, and the impact itself will decay non linearly.

But in "non-news time" when the orderflow is just quiet, the recency bias becomes valid, because this is the random part of the market, so not 1 tick will be different from other, so we should put equal emphasis on all of them.

Also if we look at longer timeframes, and view the market from a "bird-eye view" then suddenly all news become also equally important, so from a longer perspective the market becomes again linear in my opinion, because in normal cases you get the equal amount of good news than bad news, in the long run.

Just look at the GDP growth graph of the USA on the long term:

Attached Image (click to enlarge)
Click to Enlarge Name: united-states-gdp-growth.png Size: 34 KB


It adds up almost to 0 in the long term (with a slight bullish bias, which kept the GDP in growth for all those years)

So in my view only the magnification level determines the shape of the market, if you concentrate on 1 by 1 events then its non linear, but if you look at the longer picture it is linear.

So anyway if we want to catch long term trends, non-linear tools are useless.

Quote
What I don't understand is how a behavioral biais and an EMA can be linked together through this text.

And how does the EMA connect to this ? Well the EMA is the objectified tool of the recency bias itself. You emphasize the recent bars more than the non-recent bars, and the emphasis decays logarithmically, how is that not an objective recency bias?
"There's a sucker born every minute" - P.T. Barnum
  • #245
  • Edited 11:16am Aug 29, 2014 11:05am | Edited 11:16am
Quoting PipMeUp
What I don't understand is how a behavioral biais and an EMA can be linked together through this text.
Here is a visual example:

If we were to only use moving averages to spot trends, which i dont agree with entirely but just for example, then this is the difference.The AUD/JPY was in a very nice uptrend for 2 years, it had its fundamental reasons, but lets just focus on technical analysis for now. We either used an EMA to catch the 2 year long trend or a SMA.

Now lets say we look at the quarterly growth of price, that is: 4.34812141 * 5 (days) * 3 (months) , which is rounded 65 days.
So a 65 period moving average will smooth the price for a quarter year analysis, on the daily chart of course.

We plot the 65 period EMA and SMA on the chart, and you can immediately see the difference:

Attached Image (click to enlarge)
Click to Enlarge Name: AUDJPYDaily.png Size: 60 KB


The pink line is the EMA and the red is the SMA, so obviously the SMA filters the price better, and it shows us the retracements much better than the EMA, which mostly will whipsaw us.
All moving averages lagg, so the EMA cant really signal all retracements with good precision, so by the time the lagging EMA shows us a retracement, its already gone, and we will lose there.So why bother entering at retracements short, just stick with the trend.

And the SMA shows the trend much better, and the retracements are smoother, and perhaps we can do the buy & hold & close at retracement & buy back at low scheme much easier with a simple MA.

If the period is very long then the SMA and EMA will mostly be equivalent, but i think on smaller periods the SMA performs much better, dont you think ?

The only way the EMA would be better is if it would signal the move before it happened and then it would decay, but thats pretty impossible, so i vote for SMA in any case
"There's a sucker born every minute" - P.T. Barnum
  • #246
  • Aug 29, 2014 11:55am Aug 29, 2014 11:55am
The SMA is a very good tool to spot the trend, once its established, the problem that i see with moving averages, and thats why i think it should not be used alone, is that it doesnt signal the end of the trend.

Attached Image (click to enlarge)
Click to Enlarge Name: AUDJPYDaily.png Size: 67 KB


The quarterly 65 period SMA showed very well the uptrend, but once it was over, and the inevitable crash happened in 2007, by the time the SMA adjusted to the new market, we were in a 1700 pip loss in August.

Obviously the solution to this is not an EMA which "adapts quicker" to the market, but another tool which shows the velocity of the market.

Because the SMA is good as it is, however because its drawback, we need other tools aswell to help out its drawbacks.

So thats why we need a velocity indicator, so the moment the market starts to speed up, and goes against us, we can see that objectively.

Its as if we combine our 65 period SMA with a 65 period ATR

Attached Image (click to enlarge)
Click to Enlarge Name: AUDJPYDaily2.png Size: 52 KB


The increase in speed & market going against our trend, should be done by 1 indicator, that which shows us the velocity including the direction.
"There's a sucker born every minute" - P.T. Barnum
  • #247
  • Aug 29, 2014 12:31pm Aug 29, 2014 12:31pm
How does a thread with such a ridiculous title gets to have 13 pages ? It just shows that you rookies never learn.
The worst thing I can be is the same as everybody else. I hate that.
  • #248
  • Aug 29, 2014 1:19pm Aug 29, 2014 1:19pm
Quoting Proximus
{quote} Here is a visual example: If we were to only use moving averages to spot trends, which i dont agree with entirely but just for example, then this is the difference.The AUD/JPY was in a very nice uptrend for 2 years, it had its fundamental reasons, but lets just focus on technical analysis for now. We either used an EMA to catch the 2 year long trend or a SMA. Now lets say we look at the quarterly growth of price, that is: 4.34812141 * 5 (days)...
Sorry Prox' but the only thing I see obvious is that MT4 doesn't compute the moving averages correctly.
Attached Image (click to enlarge)
Click to Enlarge Name: AUDJPY-D1-SMA65-EMA65.png Size: 31 KB

RED = SMA65
BLUE = EMA65
No greed. No fear. Just maths.
  • #249
  • Aug 29, 2014 1:35pm Aug 29, 2014 1:35pm
Quoting GubyIQ
How does a thread with such a ridiculous title gets to have 13 pages ? It just shows that you rookies never learn.
Why do people never read the thread, OR use the search box, before asking a question to not ask the same question 9999999 times ?

Quoting Proximus
{quote} As i said before the 1000%/month is only to get the attention of traders.
Quoting Proximus
Nice post thanks, but the 1000%/month is more like to get the attention of traders, i cannot know how much % the system will make ,we will find out on the run.If it will only make 50%/ month its fine too, because 90% of traders cant even make quarter of that. But of course we wont stop dreaming, sky is the limit.
Quoting Proximus
{quote} Yeah maybe 1000% is just for getting the attention of traders, however that is not to say that i dont have that kind of expectation.We should do this and then we see how much we can get out of it. Anyhow, we also have a risk cap, so the risk will never be too much.I focus on risk minimization even if it cuts our profits.
I really dont like to answer to the same question 99999x times, so please use the search box before asking an obvious question again, thanks
"There's a sucker born every minute" - P.T. Barnum
  • #250
  • Aug 29, 2014 1:39pm Aug 29, 2014 1:39pm
Quoting PipMeUp
{quote} Sorry Prox' but the only thing I see obvious is that MT4 doesn't compute the moving averages correctly. {image} RED = SMA65 BLUE = EMA65
What do you mean by that, i dont understand ?

Attached Image (click to enlarge)
Click to Enlarge Name: AUDJPYDaily.png Size: 51 KB
"There's a sucker born every minute" - P.T. Barnum
  • #251
  • Aug 29, 2014 1:44pm Aug 29, 2014 1:44pm
Quoting Proximus
{quote} What do you mean by that, i dont understand ? {image}
Just count the number of crossovers on your screenshot (=18) and on mine (=9). Your EMA pink line oscillates much more than the SMA. That's not possible because the EMA reject frequencies better than the SMA for a given lag.
No greed. No fear. Just maths.
  • #252
  • Edited 1:59pm Aug 29, 2014 1:48pm | Edited 1:59pm
Quoting PipMeUp
{quote} Just count the number of crossovers on your screenshot (=18) and on mine (=9). Your EMA pink line oscillates much more than the SMA. That's not possible because the EMA reject frequencies better than the SMA for a given lag.


I think you made a mistake there Pip, your picture shows the trend from 30.8.2004 -> 23.12.2005 , while mine shows 15.12.2005 ->15.7.2007

Just a little misunderstanding there

Also there maybe differences between the price on your chart, here is my historic data:

Attached File(s)
File Type: csv AUDJPY1440.csv   155 KB | 318 downloads


So you can compare it with mine
"There's a sucker born every minute" - P.T. Barnum
  • #253
  • Aug 29, 2014 2:04pm Aug 29, 2014 2:04pm
I used the dates on your chart: 12th of September 2004 to 15th of December 2005
Attached Image (click to enlarge)
Click to Enlarge Name: AUDJPY-D1-SMA65-EMA65.png Size: 41 KB
No greed. No fear. Just maths.
  • #254
  • Aug 29, 2014 2:07pm Aug 29, 2014 2:07pm
Quoting PipMeUp
I used the dates on your chart: 12th of September 2004 to 15th of December 2005 {image}
I see, well then can you please download my historic data and compare it with yours to make sure they are the same.

Maybe the problem is with your charting, you should post the code of your MA here and see if there is a bug in the code.

I will use a custom moving average , not the built in one to check it.

If its true, then we may have a serious bug in MT4 which needs to be corrected ASAP
"There's a sucker born every minute" - P.T. Barnum
  • #255
  • Aug 29, 2014 2:20pm Aug 29, 2014 2:20pm
Open time is midnight UTC. Price is (bid+ask)/2
The chart was made from the data coming from FXCM.
Attached File(s)
File Type: csv daily_AUDJPY_dukascopy.csv   129 KB | 237 downloads
File Type: csv daily_AUDJPY_FXCM.csv   152 KB | 257 downloads
No greed. No fear. Just maths.
  • #256
  • Aug 29, 2014 2:53pm Aug 29, 2014 2:53pm
Quoting PipMeUp
Open time is midnight UTC. Price is (bid+ask)/2 The chart was made from the data coming from FXCM. {file} {file}
Yeah, the data is totally different, my broker has 3640 while ducascopy has 3363 days between 1 dec 2003 and 28 aug 2014.Also there is a 0.05 correlation between the data because the samples dont match in size, nor in the date.

This is because different LP have different day numbering, and or maybe some of them count in weekends, holidays ,etc.Also there is an offset becuse of the spread so its logical.

Also i`ve looked up the MT4 moving average , and this is the source code of it: http://www.mql5.com/en/code/7534
It matches perfectly with the MT4 built in indicator, no rounding errors, so if there is any problem with the SMA then you can see the bug in the code.


void sma()
{
double sum=0;
int i,pos=Bars-ExtCountedBars-1;
//---- initial accumulation
if(pos<MA_Period) pos=MA_Period;
for(i=1;i<MA_Period;i++,pos--)
sum+=Close[pos];
//---- main calculation loop
while(pos>=0)
{
sum+=Close[pos];
ExtMapBuffer[pos]=sum/MA_Period;
sum-=Close[pos+MA_Period-1];
pos--;
}
//---- zero initial bars
if(ExtCountedBars<1)
for(i=1;i<MA_Period;i++) ExtMapBuffer[Bars-i]=0;
}


I didnt analyzed it deeply but i think the highlighted line is useless, as the SMA should be an arithmetic average, so the vector of the SMA should always discard the elements 1 by 1, and calculate the new average as the while() moves one, there should be no subtraction.

I will code quickly a very simple arithmetic mean indicator and check wheather it is equivalent with MT4 SMA.Lets see if it matches
"There's a sucker born every minute" - P.T. Barnum
  • #257
  • Aug 29, 2014 3:18pm Aug 29, 2014 3:18pm
Ok its ready the PROXIMUS SIMPLE MOVING AVERAGE, CORRECTED.

It is a simple moving average, plotted on the CLOSE price, with MAXBAR settings, it is actually looking at the arithmetic mean of the price /PERIOD which can be set.

The MT4 basic one calculates it very well, except it has rounding errors, while mine has none, it is as it is, raw, without rounding.

Attached Image


The orange line is my SMA, and the pink is the basic MT4 Sma, so visually it looks the same, but mine is not rounded.

Attached Image (click to enlarge)
Click to Enlarge Name: AUDJPYDailyZ.png Size: 31 KB


Also this is the code of my SMA, very basic and simple

Inserted Code
double MOVING_AVERAGE(int k)
{
double AVG=0;
///--------
for(int i=k;i<k+PERIOD;i++)
AVG+=iClose(Symbol(),0,i);
///--------
AVG/=PERIOD;
return (AVG);
}

Here:

Attached File(s)
File Type: mq4 PROXIMUS CORRECTED SMA.mq4   1 KB | 440 downloads
"There's a sucker born every minute" - P.T. Barnum
  • #258
  • Aug 29, 2014 3:33pm Aug 29, 2014 3:33pm
No the highlighted line is correct. That's the way an online computation of a SMA is supposed to be done. They divide the sum by the period so there can't be any accumulation of rounding error. If I understand well yours is sub-optimal since you recompute the sum over the window at each step.

Anyway the problem comes from the EMA. It looks like the alpha parameter is not correctly calculated from the period. The usual formula is alpha = 2/(period+1) . Or they don't let the time to the filter to stabilize.
No greed. No fear. Just maths.
  • #259
  • Aug 29, 2014 3:42pm Aug 29, 2014 3:42pm
Quoting PipMeUp
No the highlighted line is correct. That's the way an online computation of a SMA is supposed to be done. They divide the sum by the period so there can't be any accumulation of rounding error. If I understand well yours is sub-optimal since you recompute the sum over the window at each step. Anyway the problem comes from the EMA. It looks like the alpha parameter is not correctly calculated from the period. The usual formula is alpha = 2/(period+1) . Or they don't let the time to the filter to stabilize.
Yes, mine makes extra work to calculate the sum at each bar, but mine is the more precise as i always divide the sum/period and dont round it.
Anyway, i made it in 5 minutes so what can you expect, i didnt considered speed to be an issue

That is exactly how they do it:

void ema()
{
double pr=2.0/(MA_Period+1);
int pos=Bars-2;
if(ExtCountedBars>2) pos=Bars-ExtCountedBars-1;
//---- main calculation loop
while(pos>=0)
{
if(pos==Bars-2) ExtMapBuffer[pos+1]=Close[pos+1];
ExtMapBuffer[pos]=Close[pos]*pr+ExtMapBuffer[pos+1]*(1-pr);
pos--;
}
}

I think the main problem is that the FOREX market is decentralized, so every LIQUIDITY PROVIDER will have different prices , open/high/low/close bars, different spread, and different trading days available, if the brokers or LP's country has different bank holidays, then by default you will get different price feed
"There's a sucker born every minute" - P.T. Barnum
  • #260
  • Aug 29, 2014 4:20pm Aug 29, 2014 4:20pm
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.
Inserted Code
ema <- ema + alpha * (value - ema)
The inclusion/exclusion of the week-ends and christmas/new year period perhaps explains the difference.
No greed. No fear. Just maths.
0 traders viewing now