The Great Pumpkin | Page 2 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

What a great site! (Dailyfx.com rave) 4 replies

A Great Indicator needs a touch of a great coder 7 replies

Great Posts By Great Traders 79 replies

Average system + great discipline =great results? 44 replies

Attachments: The Great Pumpkin
Exit Attachments

The Great Pumpkin

  • #21
  • Edited 1:38am Oct 27, 2014 1:25am | Edited 1:38am
Quoting Winston Reed
I wish indicators were the answer to profitable trading. If they were everybody would be racking up pips. Sadly indicators only indicate the past and they do not predict the future. No matter how fancy your codes. In a strongly trending market you will do ok but in a ranging market you will suffer losses. Learn the market. Forget the indicators.
hi Winston,
to a large extent i agree with you. and i am learning the market. i just wrote a comment along very similar lines in another thread.
however, i think that there are instances where technical analysis can do a reasonably good job. one important area are statistically rare events. it is possible to identify very rare events and with a large probability bet that the situation will be resolved.
an example with an improved bollinger band indicator is, to look at the times in the uptrend where the price spikes down below a lower band. one can do some analysis to tune the bands width for a precise entry. then one can bet with a large degree of certainty that this is just a spike and not a trend change.
it is of course simpler when you approximately know what you're doing and can piggyback on the big players: set up an alarm when eurchf drups to/ below 1.2 (note that this is also a rare event) and enjoy your free money.
best wishes
k
  • #22
  • Oct 27, 2014 1:33am Oct 27, 2014 1:33am
Quoting kprsa
{quote} hi Winston, to a large extent i agree with you. and i am learning the market. i just wrote a comment along very similar lines in another thread. however, i think that there are instances where technical analysis can do a reasonably good job. one important area are statistically rare events. it is possible to identify very rare events and with a large probability bet that the situation will be resolved. an example with an improved bollinger band indicator is, to look at the times in the uptrend where the price spikes down below a lower band....
"Anything that came from numbers, we can re-calculate back with numbers" MTH
  • #23
  • Oct 28, 2014 1:16pm Oct 28, 2014 1:16pm
Quoting kprsa
Improved Bollinger Bands indicator. In this comment PipMeUp discusses problems with the standard implementation of the Bands indicator. His objections are: 1) The number of samples for the deviation should be typically much bigger than the one for the MA. This implementation allows this, as you can specify the numbers separately. 2) The lag of the simple moving average is half of its period. If we subtract the moving average at position from the price at candle ...

Hey,
thanks for sharing your costomized version!
Is it in any way possible to use it in an indicator window? I like to keep my main chart as clean as possible, so currently I put the Indi in another chart window. That's working quiete good when watching current price but having it in the same window would be a more elegant solution, especially when watching history prices.
Kind regards
PG

Attached Image (click to enlarge)
Click to Enlarge Name: bb.jpg Size: 120 KB
  • #24
  • Nov 9, 2014 2:06pm Nov 9, 2014 2:06pm
Quoting PipitoGrillo
{quote} Is it in any way possible to use it in an indicator window?
Try this version.
k

Attached File(s)
File Type: mq4 BB_improved_kprsa_separate_window.mq4   3 KB | 573 downloads


Attached Image (click to enlarge)
Click to Enlarge Name: Clipboard01.jpg Size: 92 KB
  • #25
  • Nov 9, 2014 2:48pm Nov 9, 2014 2:48pm
Quoting LITEchild
{quote}Is it possible to get a display for the min, max and average distance between the bands at any given point in time over a defined range of previous bars?
Check this version.
Attached File(s)
File Type: mq4 BB_improved_kprsa_LC.mq4   3 KB | 609 downloads
  • #26
  • Nov 9, 2014 3:37pm Nov 9, 2014 3:37pm
Quoting kprsa
{quote} Check this version. {file}
Wow! This is wonderful KP!!! Tnx a million man! I have a hunch about using those stats as a more precise volatility estimate for entering on pullbacks. Is the measurement displayed as pips? Also is it possible to have those levels subtracted from the lows and added to the highs at the close of the candle (as line levels) and updated as the chart progresses? I don't know if what I'm asking makes sense. Hope I'm not asking too much. hahahaha.
Attached Image (click to enlarge)
Click to Enlarge Name: USDJPYM30.png Size: 48 KB
  • #27
  • Nov 9, 2014 3:59pm Nov 9, 2014 3:59pm
Quoting LITEchild
I have a hunch about using those stats as a more precise volatility estimate for entering on pullbacks. Is the measurement displayed as pips? {image}
Hi LITE, just note that the statistics is based fully on the Close prices, not on the High and Low or their difference.
I guess it would make more sense for your case if they would be done for High and Low difference somehow?
The averages are in units of of your chart (price), it was simpler to program it this way. So currently the low band is at 114.59, upper is at 115.51, difference is Upper-Lower=115.51-114.59=0.92.
  • #28
  • Nov 9, 2014 4:17pm Nov 9, 2014 4:17pm
Quoting kprsa
{quote} Hi LITE, just note that the statistics is based fully on the Close prices, not on the High and Low or their difference. I guess it would make more sense for your case if they would be done for High and Low difference somehow? The averages are in units of of your chart (price), it was simpler to program it this way. So currently the low band is at 114.59, upper is at 115.51, difference is Upper-Lower=115.51-114.59=0.92.
It's fine for the min, max and averages to be compiled at candle closes. but the PRICE LEVELS I'm interested are to be computed by adding these numbers to the current(or previous?) candle high and low to get the most probable(possible?) estimates in terms of volatility for pullbacks, entries and targets.
So for the current CANDLE with low at 114.396 and high at 114.521, current stats read 0.491705, 1.17118 and 0.783414 for min, max and average volatility respectively. So the LEVELS I am interested in would be computed by ADDING these three stats to the HIGHS(which would yield 3 price levels), and also by SUBTRACTING them from the LOWS(yielding another 3 price levels). The result being displayed as line levels(updated at candle closes of course).
  • #29
  • Nov 10, 2014 1:48am Nov 10, 2014 1:48am
When I entered forex world 10 months ago, I never had imagined what I had started to learn.
Now, in the company of you guys, I think I have found the light.


Thanks KP, Kiads and all others. I am in.
Traders gain some knowledge, win, their emotions kick in, and they die.
  • #30
  • Nov 10, 2014 6:56am Nov 10, 2014 6:56am
Quoting kprsa
The usual tricks (combining two timeframes to trade pullbacks): {image}
KP

How to do this trick Sir?
Traders gain some knowledge, win, their emotions kick in, and they die.
  • #31
  • Nov 10, 2014 11:32am Nov 10, 2014 11:32am
Quoting kprsa
{quote}set up an alarm when eurchf drops to/ below 1.2 (note that this is also a rare event) and enjoy your free money.
  • #32
  • Nov 10, 2014 11:54am Nov 10, 2014 11:54am
Quoting mehtabkk
{quote} KP How to do this trick Sir?
Hi, I do not aspire to teach trading to anybody (I am also learning every day). There are many resources on the net for this.
Start with the moving average cross. The Bollinger Band image I was showing is a similar thing, but it gives you a bit more information on the chart. Note that this strategy works in a trend.

http://www.investopedia.com/universi...gaverages4.asp

Cheers,
k
  • #33
  • Nov 10, 2014 11:54am Nov 10, 2014 11:54am
Quoting kprsa
{quote}
  • #34
  • Nov 10, 2014 11:56am Nov 10, 2014 11:56am
Quoting LITEchild
{quote}
LITE, I started coding something for your request, but did not finish yesterday... Will get back to it this evening.

k
  • #35
  • Nov 10, 2014 12:01pm Nov 10, 2014 12:01pm
Quoting kprsa
{quote} LITE, I started coding something for your request, but did not finish yesterday... Will get back to it this evening. k
I trust you! Anything you code is awesome! Take your time buddy.
  • #36
  • Nov 11, 2014 2:45am Nov 11, 2014 2:45am
Hi all,

Quoting LITEchild
{quote}The maths is way over my head
You're not the only one.

http://www.mathsisfun.com/data/standard-deviation.html
http://www.mathsisfun.com/data/stand...-formulas.html

Calculating standard deviation involves some numbers and their mean (average). I.e. if you take 5 bars (close prices to be exact), it takes the mean of those 5 bars and then makes calculations based on the difference of THOSE bars from THEIR mean. The original BB indicator does just that. Let's say we're at current bar, having an index of 0. The previous bar has index 1, the bar before that has index 2 etc. (This is just for non-coders to understand the example.) So the indicator will use bars 4,3,2,1,0.

Now, let's say we use 6 as MA_Period (has to be even because the code divides it by 2 and adds it to the index variable, and that can only be an integer) and 10 as BandsPeriod in krpsa's indicator, and we're at the current bar again. According to his code, bars 5,4,3,2,1,0 are used to calculate the average (which is so far the same), and then, instead of subtracting values of bars 5,4,3,2,1,0 from this average (values the average is made of), it will subtract values of bars 12,11,10,9,8,7,6,5,4,3.

Now either I'm dumb, failing to see the point of subtracting the values of bars 12-3 from the average of bars 5-0 (the links provided in the thread didn't help either), or this is utter bullshit that just looks fancy on the chart.

I'll gladly stand corrected if I missed/misunderstood anything.
Even if you fall on your face, you're still moving forward.
  • #37
  • Nov 11, 2014 3:19am Nov 11, 2014 3:19am
Quoting Ralome
Hi all, {quote} You're not the only one. http://www.mathsisfun.com/data/standard-deviation.html http://www.mathsisfun.com/data/stand...-formulas.html Calculating standard deviation involves some numbers and their mean (average). I.e. if you take 5 bars (close prices to be exact), it takes the mean of those 5 bars and then makes calculations based on the difference of THOSE bars from THEIR mean. The original BB indicator does just that. Let's say we're at current bar, having an index of 0. The previous bar has index...

Hi, the idea for the next snippet of code is

sum=0.0;
k=i+BandsPeriod-1;
while(k>=i)
{
oldval=MovingBuffer[k];
newres=Close[k+MAPeriod/2]-oldval;
sum+=newres*newres;
k--;
}
deviation=BandsDeviations*MathSqrt(sum/BandsPeriod);
UpperBuffer[i]=oldval+deviation;
LowerBuffer[i]=oldval-deviation;
i--;

is that, to calculate the standard deviation, one needs to subtract the Close of a half a MA period ago from the current value of MA. The reason is, as already mentioned, the fact that the lag of the simple moving average is half its period. one should actually be drawing the simple moving average shifted backwards by half of its period if one wanted to compare things properly.

another change from the standard bands is that one can and should specify a different period for the calculation of standard deviation and the fast MA. the reason is that for the measurement of standard deviation one needs more samples than usually thought, so the standard Bands makes essentially a massive error in what it is supposed to do.

now, the code may be obscure, faulty, slow, or even completely useless or wrong, and i am happy that you criticise it on any of those grounds. but to call it "utter bullshit" is way over the top. i would warrant calling it bullshit if i would be trying to sell it and would be giving away just the ex4 file. but i am not, it is freely given away as source if anybody finds it useful (which i am not promising). (if you are looking for utter bullshit on ff to criticise, you really don't have to look far.) so, please, mind your manners.

k
  • #38
  • Nov 11, 2014 4:16am Nov 11, 2014 4:16am
Quoting kprsa
to call it "utter bullshit" is way over the top
Okay, I admit it did sound somewhat harsh, but I didn't want to offend anyone. Anyway, I apologize for that, and in the future I'll choose my words more carefully.

I understand what you're doing / the code does, I just don't understand why. In your reply you basically just repeated PipMeUp's comment, which offers no explanation to me either. Take the first one (quoted from his comment):

"The first issue with the BB is that it uses as many samples to estimate the variance as it uses to compute its moving average. (...) This is absolutely wrong!"

Why is this an issue? Since when is the formula for calculating standard deviation wrong? Any articles on this one (i.e. there's a new way of calculating standard deviation)? Or is it no longer called standard deviation then?

The second one:

PipMeUp: "The second issue is that the variance is estimated by taking the average of the squared difference of the current value of the SMA and the current price. This is also incorrect because the SMA lags by the half of its period. One shall measure the difference between the current value of the SMA and the price half the period ago."

Investopedia: "Variance: A measurement of the spread between numbers in a data set. The variance measures how far each number in the set is from the mean."

I see strong contradictions here. I also fail to see how the lag of the SMA plays a role in calculating standard deviation.
Even if you fall on your face, you're still moving forward.
  • #39
  • Nov 11, 2014 5:10am Nov 11, 2014 5:10am
Quoting Ralome
{quote} Since when is the formula for calculating standard deviation wrong? Any articles on this one (i.e. there's a new way of calculating standard deviation)? Or is it no longer called standard deviation then?
The formula for the standard deviation is not wrong, it is also used in this code. What is wrong is the number of samples required to reach a reasonably precise conclusion. The Bands take the same number of samples (candles/bars) for estimating the mean (moving average) and the standard deviation (bands). I think the default is 20 or so.
The 95% confidence interval for standard deviation based on 20 normally distributed samples with σ=1 is 0.7605 < σ < 1.4606. This means that the measurement of standard deviation done with 20 samples is going to result in a value within -24% and +46% of the real value (σ=1) with 95% chance. The prices are not normally distributed (their distribution has thicker tails than a Gaussian on the time-based chart), so the situation is even worse than this. This is the reason why Bands expand and contract with every microtrend. While this feature might be useful in trading, it is essentially an artifact from insufficient number of samples.
Now, the 95% confidence interval for standard deviation based on 200 normally distributed samples is 0.9107 < σ < 1.1089. This means that the measurement of standard deviation done with 200 samples is going to result in a value within +/-10% with 95% chance. Again, the prices are not normally distributed, but I am happy with the precision obtained by this number of samples. The end result is that the bands do not expand and contract so much.

Quoting Ralome
{quote} PipMeUp: "The second issue is that the variance is estimated by taking the average of the squared difference of the current value of the SMA and the current price. This is also incorrect because the SMA lags by the half of its period. One shall measure the difference between the current value of the SMA and the price half the period ago." Investopedia: "Variance: A measurement of the spread between numbers in a data set. The variance measures how far each number in the set is from the mean." I see strong...
There is no contradiction. It is just that the estimate of the mean, the simple moving average, lags in time by half the period behind the price. This is why there is that adjustment in the code.

Cheers,
k
  • #40
  • Nov 11, 2014 5:25am Nov 11, 2014 5:25am
Quoting kprsa
{quote} It is just that the estimate of the mean, the simple moving average, lags in time by half the period behind the price. This is why there is that adjustment in the code. Cheers, k
Example:

SMA(7)

Candle: 1 2 3 4 5 6 7
Prices: a b c d e f g

SMA calculates (a+b+c+d+e+f+g)/7, which is fine, and attributes it to price g, which is not correct. It should be attributed to the price d, since it is an estimate of mean around this price, particularly if one wants to calculate something based on this, like the standard deviation.

k
0 traders viewing now