Has anyone wondered why and how much the trend building "walks"?
Here is the beginning -- https://www.mql5.com/ru/forum/139689.
I think the answer will be quicker here, especially since the same crap happens in terminal 5.
Just reproduced it :) . Frankly speaking, I thought trend-following was quite reliable, but I was brutally thrown to the ground.
What do you think, gentlemen developers?
_____________________________________
Just in case, I'll repeat myself. No history inserts, no timeframe switching, no bar skipping, both manual and automatic constructions done right here and now.
Just a change of scale.
I don't know how to play it?
I don't have any straying, what am I doing wrong?
Do you think the principle of constructing charts and everything belonging to him has changed at 5?
I think there's a quicker response here.
I do not understand how to playback?
I don't have any straying, what am I doing wrong?
Nikolai, I don't know how to reproduce. Look at the screenshots on the 4, maybe it will give you an idea.
Try reproducing the picture for #IBM on the clocks, they are on the demo at the metaquotes.
I think there will be a quicker response here.
Nikolai, I don't know how to reproduce. Look at the screenshots on 4ka, maybe it will give you an idea.
Try to reproduce the picture for #IBM on the clock, they are on the demo at metaquotes.
No way. It still works correctly.
And how far are the trendline reference points from the junction?
It is not very difficult to reproduce, let's take a EURUSD D1 chart (any other chart will do)
draw a line (using points as on the chart), fix the scale in the settings
look at the last candlestick which touches the line with its upper shadow
move the chart with the mouse so that one of the coordinates crosses the bottom border of the chart.
we observe that at the moment of crossing, the line starts to change its position6: the first point of the line starts to jump a small number of points
as a result, the last bar crosses the line with a small number of points.
Here is the previous figure as an example 
Yep, replicated that too, took the reference points to 6000 bars, and started changing the vertical scale,
as a result the trendline misses the crosshairs.
It's all about discreteness here, apparently just not enough accuracy to calculate the correct value to render.
It is not very difficult to reproduce, we take EURUSD D1 (you can use any other chart, just this one is more visible for me)
draw a line (on the points as on the chart), fix the scale in the settings
looking at the last candle that touches the line with its upper shadow
and try to put it on this last candlestick - the right key point of the trend line.
I don't think there will be any bounces.
So I agree with Nikolay's opinion that the angle accuracy is lacking because of the discreteness.
because of the discreteness of what lacks accuracy where?
the scale of the chart layer does not match the scale of the layer on which the lines are drawn or each layer has its own accuracy
And try to put it on this last candlestick - the right reference point of the trend line.
I don't think there will be any jumps.
Therefore I agree with Nikolay's opinion that the accuracy of the angle is lacking because of the discreteness.
The point is that when we build a line, we do not know where it will touch the chart in the future, and we build a line and wait for the price to touch the line, but it turns out to be visually shifting in the future.
Everything is OK programmatically, but visually +20 points)
ZS: the older the period, the larger the error may be
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
Here is the beginning -- https://www.mql5.com/ru/forum/139689.
I think the answer will be quicker here, especially since the same crap happens in terminal 5.
Just reproduced it :) . Frankly speaking, I thought trend-following was quite reliable, but I was brutally thrown to the ground.
What do you think, gentlemen developers?
_____________________________________
Just in case, I'll repeat myself. No history inserts, no timeframe switching, no bar skipping, both manual and automatic constructions done right here and now.
Just a change of scale.