From theory to practice - page 1176

 
Maxim Dmitrievsky:

overfitting is common

No. This is not a fitting.
Uladzimir Izerski:

Since about year 16 all brokerage houses started using the overnight spread widening. Perhaps that is where the problem lies.

Probably saw that a lot of people earn at night)
 
multiplicator:
and on alpari the same thing.
no broker has a good history before 2016?
or maybe it really worked like that before 2016).

On small frames, the histrionics are somehow distorted. I did a similar thing in 2010, and it's been the same twist for the last 2 years))) Small is small. Now with 16, in 2 years it will be with 18)))))))))

 
multiplicator:
how temples break down.


it worked perfectly until the beginning of 2016. then it broke.

or is it possible that the broker in mt5 only started to account spreads normally since 2016....

Has anyone encountered such a picture in the tester?
I have the opposite, since 2016 the testers are perfect and before 2016 it feels like the quotes are holey.
 
vladevgeniy:

On the smaller frames, the histrionics are somehow distorted.

so m1 in mt5 should be normal.
as they say: real bids and aski, real spreads.
 
multiplicator:
so m1 in mt5 should be fine.
I have no idea what to do with them, but I'm not sure what to do with them.

When I tried it, there wasn't even an mt5 yet, and the same thing. Something is distorted. Maybe it's the old quotes from the metrics. Maybe they are whitened there I don't know. In any case, such a system is not viable in real life, most likely.

 
Those who approach the grail search professionally, using econometrics and mathematical statistics should not forget that the ark was built by amateurs and the Titanic by professionals. Here the solution should be as simple but unexpected as the "Columbus egg").
 
khorosh:
Those who approach the grail search professionally, using econometrics and mathematical statistics should not forget that the ark was built by amateurs and the Titanic by professionals. Here the solution should be as simple but unexpected as the "columbo egg").

I agree with you )

 


On a graph of velocity increments = (difference between current and past price)/interval

 
Evgeniy Chumakov:


In the velocity increment graph = (difference between current and past price)/interval

Zhenya, lost important information

Let's say price rose 100 pips in 100 bars and fell 100 in one bar.

The vectors have a different slope, but the increment is the same.

If you could just rotate 100 and -100 instead of 1 and -100, that would be OK.

The solution, as Krosh says, is as simpleas a columbo egg, and it exists.

I'm sure A_K will snag such a movement here and there, 100%.
 
Evgeniy Chumakov:


Just take a step of n pips and that's the solution or not?

A_K found it. (I have a different one).

It's just that he's got a trend left to figure out.

---

mmm.....

If the trend goes in the red, then we should tighten possibility to enter unprofitable deal, i.e. no-backward counter-trend.

At least, we should divide the signal line which wobbles between two levels and affects these levels by the slope angle of the trend line drawn on the middle between the levels.

I haven't done it myself, but that's probably what I would do ... probably... try and test

it turns out that 4 curves need to be drawn and one trend