History Centre updated - free history of minute quotations since 1999 - page 6

New comment
 
Renat писал (а):
 You follow the standard beginner's path, trying to get down as low as possible and burrow into the tick stream. To put it crudely - you're doing pips, constantly depending on a couple of pips of divergence and always on the lookout for the perfect quote.

All this has been discussed many times. Read posts related to this topic and click "similar" at the bottom right of interesting topics. You'll find plenty of similar discussions. For example: Standard misconceptions in trying to trade in noise
I don't know. It is clear that being forced by duty to constantly communicate with users, you have long explicitly or implicitly classified them and attributing them to this or that group is also automatic. I think in my case the "identification" procedure has failed, but in general it doesn't matter much. At least I've written twice that I would have been interested in a filtered story. You can't tell in advance at which level of detail the useful information starts. In any case, a player of the first "non-pips" level needs information from the last "pips" level by definition :).
Another question is whether this filter exists explicitly? Very possibly, it does not. That is, the task of "updating" the history may be very difficult, if not unsolvable.
 
By the way, a question that may seem unexpected :) . Is the genetic algorithm in the tester an independent development or a by-product of the "intellectualisation" of the server?
 
The genetic algorithm has nothing to do with the server, it is embedded in the user (client) terminal.
 

Well, thanks for the reply. Surely the server has many tasks to optimise, but "parliament is no place for discussion" :)

 
lna01:
Renat wrote (a):
 You follow the standard beginner's path, trying to get as low as possible and burrow into the tick stream. To put it crudely - you're doing pips, constantly depending on a couple of pips of divergence and always on the lookout for the perfect quote.

All this has been discussed many times. Read posts related to this topic and click "similar" at the bottom right of interesting topics. You'll find plenty of similar discussions. For example: Standard misconceptions in trying to trade in noise
I don't know. It is clear that being forced by duty to constantly communicate with users, you have long explicitly or implicitly classified them and attributing them to this or that group is also automatic. I think in my case the "identification" procedure has failed, but in general it doesn't matter much. At least I've written twice that I would have been interested in a filtered story. You can't tell in advance at which level of detail the useful information starts. In any case, a player of the first "non-pips" level needs information from the last "pips" level by definition :).
Another question is whether this filter exists explicitly? Very possibly, it does not. That is the task of "rejuvenating" the history may be very difficult, if not impossible.
Apparently I'm not wrong. You continue to insist on a "filtered/ideal" history, which clearly points to attempts to burrow into tick flow and pips.

I'm not accusing, just reiterating that this is a standard mistake of novice traders. You should move exactly in the opposite direction and write robust Expert Advisors, that are not sensitive to tick noise and easily swallow the difference in quotes, at least in spreads.

Read all about History Center in the forum - a lot has been discussed in detail.
 
Renat:
You keep insisting on a "filtered/ideal" story, which clearly indicates attempts to burrow into tick-flow and pipsqueak.

OK, let's try again.
Filtering in the broad sense can be any transformation of data. A quotes server receives some (I don't know what) data as input and outputs the quotes, thus filtering the input data. It doesn't matter if it has a function called "SuperFilter" or not. Thus, my first thesis is (and was initially) the statement of the following fact: a quote stream is a result of filtering.
The second thesis is that I agree with the fact that this filtering was performed differently in 1999 and 2007. Actually I wasn't arguing it, you just pointed this fact out.
Finally, the third point was this: data for 1999 would be much more useful if they were filtered the way they are in 2007. If it is possible, of course.
I don't see where the requirement of a perfect history can be found here. Maybe, from the developer's point of view, they are ideal, I cannot judge. A stereotype of perception is a more probable explanation.
Now to tick flows and pipsing. The simplest way to fight the noise is averaging. I suppose that when calculating a value by 1440 minute bars, I will have a better estimation than when calculating the same value by one daily bar. Or even 24 hour bars. Note that we are talking about the same playing horizon. In other words, interest in minute bars does not necessarily mean that we intend to engage in pipsing in noise (though in fact any movement can be called noise - if one looks at it from an appropriate horizon). It is another matter, that with the growth of statistics the result improves much slower, than the calculation time increases, but I would prefer to determine the optimum by myself, depending on what is calculated.
Ticks were indeed mentioned once, but it was a question of what only MQ itself can do (if, of course, it can).

 
lna01:
Renat:
You keep insisting on a "filtered/ideal" story, which clearly indicates attempts to burrow into tick flow and pips.

OK, let's try again.
Finally, the third thesis was this: the data from 1999 would be much more useful if it was re-filtered the way it was done in 2007. If that is possible, of course.

If you read the links to which I point (search rules), there would be fewer theses. For it has all been written about and explained a long time ago.
 
lna01:
Now for the tick flows and pips. The simplest way to deal with noise is by averaging. I assume that if I calculate some value on 1440 minute bars, I will have a better estimate than if I calculate the same value on one daily bar. Or even 24 hour bars. Note that we are talking about the same game horizon.
Without entering into a general discussion, I will allow myself to object to this thesis.
Theoretically, the bigger the sample, the higher the accuracy. In practice, however, an extra thousand bars will give such a negligible increase in the accuracy of the calculated value that it is not worth paying attention to. Starting with some size of the sample, its increase loses its practical value. And this "some size" is much closer to 24 than to 1440.

 
timbo:
lna01:
Now for the tick flows and pips. The simplest way to deal with noise is by averaging. I assume that if I calculate some value on 1440 minute bars, I will have a better estimate than if I calculate the same value on one daily bar. Or even 24 hour bars. Note that we are talking about the same playing horizon.
Without getting into general discussion, I would like to take the liberty of objecting to this thesis.
Theoretically it's true - the bigger the sampler, the higher the accuracy. However, in practice, an extra thousand bars will give such a negligible increase in accuracy of the calculated value that it is not worth paying attention to. Starting with some size of the sample, its increase loses its practical value. And this "some size" is much closer to 24 than to 1440.


There further it is said: Another thing that with growth of statistics the result improves much more slowly, than growth of time of calculation, but I would prefer to define an optimum by myself, depending on what exactly it is calculated. I have nothing to add.
 
It's not a question of time, it's a question of meaning.
Calculation time is not critical - buy a better computer and get on with it.
It just doesn't make sense. The result does not improve slower, but only by hundredths, thousandths or even ten thousandths of a percent. And if your expert is sensitive to hundredths of a percent, that is the lack of robustness with all that it implies...
12345678910111213...21
New comment