Offering Help

[zdo 20]

flyingdutchmen,

 

Hi. What is the most efficient (and ubiquitous across scales) way in EL to measure when a line (any line) has stopped curving and has straightened out for n bars? For purposes of example, let's look for 3 bars with the line ‘straight’ Note it’s not required that it be perfectly straight, just within tolerance.

Many thanks.

 

zdo

[flyingdutchmen 10]

  zdo said:

flyingdutchmen,

 

Hi. What is the most efficient (and ubiquitous across scales) way in EL to measure when a line (any line) has stopped curving and has straightened out for n bars? For purposes of example, let's look for 3 bars with the line ‘straight’ Note it’s not required that it be perfectly straight, just within tolerance.

Many thanks.

 

zdo

 

i surpose you are refering to straight as in not "choppy" and the line can be any value-line from price data to some kind of indicator line ?

[zdo 20]

  flyingdutchmen said:

i surpose you are refering to straight as in not "choppy" and the line can be any value-line from price data to some kind of indicator line ?

 

flyingdutchmen,

 

re: “… and the line can be any value-line from price data to some kind of indicator line ? “

Yes, any ‘chartable’ line. A SimpleMovingAvg is the simplest example. Example

attached.

 

re: “i surpose you are refering to straight as in not "choppy" “ For this question, “choppy” is not quite the ‘other’ of straight. For this, it is straight (within tolerance) as in not curving. See attached – the grey rectangles show (very roughly) the curving areas and the none marked areas are (roughly) within the tolerance of straight.

 

I use the word ‘roughly’ mainly because the eyes are easily deceived - especially to see straightness when it’s really curving…(maybe it's just my eyes, btw. but in any event I need and appreciate help with this)

 

Many thanks.

[flyingdutchmen 10]

one way which comes to mind would be to calcuate the procentual difference between

the line and its liniear regression line over the n amount of bars, another way to calcuate

the angle of the slope and the difference between the slope of current bar and 1 bar back over the n amount of bars.i am in the midle of some other things at this very moment but i will try to give an easy language example soon

[flyingdutchmen 10]

. .

Edited November 13, 2009 by flyingdutchmen

[flyingdutchmen 10]

  zdo said:

flyingdutchmen,

 

re: “… and the line can be any value-line from price data to some kind of indicator line ? “

Yes, any ‘chartable’ line. A SimpleMovingAvg is the simplest example. Example

attached.

 

re: “i surpose you are refering to straight as in not "choppy" “ For this question, “choppy” is not quite the ‘other’ of straight. For this, it is straight (within tolerance) as in not curving. See attached – the grey rectangles show (very roughly) the curving areas and the none marked areas are (roughly) within the tolerance of straight.

 

I use the word ‘roughly’ mainly because the eyes are easily deceived - especially to see straightness when it’s really curving…(maybe it's just my eyes, btw. but in any event I need and appreciate help with this)

 

Many thanks.

 

i still owe you an anser on this one, sorry it took me a while to reply.

you where asking for an easy and fast way to check if your slope is within tollerance,

here is a small script which you could save as a false/true function if it is what you are searching for.

read the examples i have wrote in the brackets.this skript may be a bit hard to

understand at first but it is pretty logical and simple.

it uses a 50-period sma to be your line and calculates the difference between the value of the current bar and n bars back

dividet by n, this gives the value each bar should go up or down to meet your creteria and will be used to check all other

bars within a loop

 

 

inputs: n(5), toll(.5);{tollerance in percent of difference}

variables: result(0), diff(0), index(0), counter(0);

result = average(close,50);{something we will use as a line}
diff = absvalue(result-result[n]) / n;{absolute difference between current value and value n bars back devidet by n}



counter = 0;{set counter to zero}

if result > result[n] then begin{first loopblock in case of current value being higher same value n bars back, second block visa versa}

for index = n-1 downto 0 begin
if result[index] >= result[n] + ((n-index)*diff) - (toll*diff) and{make sure each bar coming in after n is higher but not higher then the tollerance}
 result[index] <= result[n] + ((n-index)*diff) + (toll*diff){make sure each bar coming in after n is lower but not lower then the tollerance}
        then counter = counter + 1;{counting in case of succes}
end;
 end else
for index = n-1 downto 0 begin
if result[index] <= result[n] - ((n-index)*diff) + (toll*diff) and
 result[index] >= result[n] - ((n-index)*diff) - (toll*diff)	
        then counter = counter + 1;
end;

condition1 = counter = n;{if succesfull counter should be equal to n after having all bars calculated}

if condition1 then plot1( result );{show if succesfull}

 

 

 

in case of a sideways market the statement (toll*diff) could be a bit to narrow

because the difference between current value and the value of n bars back is nihil, and therefore

give less signals. this you could change by (toll*value[index+1] to give

more signals in case of a "straight" slope in a sideways market.

it will take a percentage of the value 1 bar ago within the calcuation of the loop, adapt the input to it.

Edited November 16, 2009 by flyingdutchmen

[zdo 20]

Many Thanks, flydutchmen. Looks good at first glance. Will take a look at it later today.

 

PS Something in the pictures and frames prevented me from getting over to the Thanks button

[flyingdutchmen 10]

the idea of this litle script is that on each bar it will take

the current value and the value of n bars back and by the difference over those n bars

it will create a "virtual" and "straight" line from your MA n bars back to the MA of currentbar and then messure

the difference from the MA to that virtual line. if the difference is greather that the virtual line

incl the tollerance on of those bars the counter will never be equal to n, therefore condition

will stay set to false.all bars must stay within this tollerance in order to set the counter equal to n

[zdo 20]

flyingdutchmen,

 

Is your offer for helping still open? I got another one ... (that may or may not be helpful to the many, btw…)

 

 

Many thanks.

[BlowFish 10]

Is there way to tell if you are processing a historical bar (backfilling data) or a live tick? I could have swore that I had done such a think once but Im boogered if I can remember how?

[Tams 20]

LastBarOnChart

this function will only process the current tick.

 

 

 

 

note: this function is a bit on the CPU heavy side. Only a wee bit, you shouldn't notice the effect... but if you use it too liberally, you might know it.

[BlowFish 10]

Ahh of course thanks. <Doh>. Though I think once you reach lastbar on chart you should wait for the next bar before switching to real time processing to prevent having a bar half processed live whilst being processed as a historical bar.

[accendotraders 10]

Thanks dutch for your kindness in helping the non coders / statistic folks.

 

I am looking for something that plots a dot or other indicator on chart when the following criteria happens:

 

The slow stochastic must be over sold (below the 20 band), sloping & crossing upwards, with higher avg volume in ticks of the last 10 bars, a volume flushout and we must have a formation like a hammer, doji or bullish engulfing candle.

 

thanks in advance,

[andypap 10]

hi

 

i dont know if you still offer help.

 

i am intresed for a trendline indicator

 

to explain better the code should not have Tl type functions and will plot values on everybar.it should resemble a trendline even if its not as exact as a manually drown trenline...

 

my own best shot at this was to take a long term average that resembels a trendline

and then

if the close > than the average i would store the highest for barcounter above the average,

take the difference (highest(h,barcounterabov e) - average ) and add that number to all bars that h < highest(h,barcounterabove ) to the average value..

 

the result is a channel like projection at the high that keeps the average slope untill a new high is found..

 

 

 

the problem with this code is that the average is not conecting the low swings like a trendline would and at the first x bars after a new up trend is develloping the high is breaking every other bar almost

 

 

regards

[cooper59 10]

Hi Dutch

If you r offering help still... I was looking for a TS indicator (ELD) for a Vwma (Volume weighted moving average). In the indicator list on TS, it has a weighted moving average but unless I am missing it, I cannot find a Vwma.

Should you decide to take on this mission, good luck (and I will be looking forward to the results). This coded request will self destruct in 60 seconds. :helloooo:

 

Bond, James Bond (cooper) lol

[BlowFish 10]

  cooper59 said:

Hi Dutch

If you r offering help still... I was looking for a TS indicator (ELD) for a Vwma (Volume weighted moving average). In the indicator list on TS, it has a weighted moving average but unless I am missing it, I cannot find a Vwma.

Should you decide to take on this mission, good luck (and I will be looking forward to the results). This coded request will self destruct in 60 seconds. :helloooo:

 

Bond, James Bond (cooper) lol

 

Google it. I found several on the first page of results.

[cooper59 10]

  BlowFish said:

Google it. I found several on the first page of results.

 

I had tried google before I came here as well as the TS forum (which is where I expected to find it but to no avail. The first 3 listings I found on google, were for people with programing experience which I have little of (Although I have learned how to copy and paste for easy language). I saw nothing on google that could be copied over to TS Easy Language. Searching further into google was again for a person with programing experience or downloading from unknown sites that may very well contain virus's. Again, I do apologize for posting in the wrong room originally. With that being said, if anyone does have the programing for a Vwma for TS or if dutch is still looking for work, I would appreciate any help. Thanks