I made one of these things once ... - page 10
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
I used to think so too, until some time ago. Indeed, where do beginners start? The more indicators they have, the better. And then they do not know what to do with a lot of parameters. Eventually, you start to realize that any parameter is an arbitrary decision of the builder. We have a legitimate desire to get rid of them altogether. But is it the right thing to do?
If we say that the market has a fractal structure, we should(must!) admit that this structure consists of levels. This means that at least the values that parameterize this level structure are immanent to the market and therefore should be present in the TS. And by the way, they are all well-known to us: trading horizon, the minimum price range for a deal - each trader determines these values for himself before the TS construction begins. And since these two values are related, then at least one parameter should be present in the TS. And I suggest it. Especially, it is determined based on external considerations, as you requested.
I agree. We cannot get away from this parameter. But here we tried to get down to business and immediately encountered a question - how to define "level escape". And I don't yet understand how to do without another parameter here.
There is another important criterion. Since we are going to collect statistics on history, the algorithm should not be too "heavy". And since we are going to work with levels, it would be natural to tie the criterion to the distance between levels. For example, we divide this interval by a power of two, so this power could be the parameter. Importantly, it has a very small range of reasonable values, 2, 3, well 4.
For this problem, though, it would be natural enough to divide by 10, but somehow it seems inevitable to me that the question about other ways of dividing into levels will arise :).
By the way, I remember I once forced the algorithm of calculation of Murray levels exactly with the purpose of possibility to work with them on a long history :).
...I meant that you cannot know how many times price has crossed a level within a bar. Consequently, if you want to react to every real crossing, you cannot get statistics for such an algorithm on history. If you have decided to count not more than one crossover per bar, it means that you have already entered one parameter - delay and set its value: "till the end of the calendar minute" (so to speak). The question is then inevitably raised as to whether such a delay is appropriate.
I was thinking about it, and about the option of +- 2, 3, ... from the level. We want to determine the significance of a round level compared to another, so the algorithm should be simple, have a minimum of parameters, preferably none at all, but it doesn't work.
what we need.
1. the criterion of effectiveness, is not money, but points.
2. the criterion of a level breakout
3. some fixed time interval to give the price time to move away from the level...
the criterion may not be Boulishev, the essence is to have these 4 numbers, entry, exit, max and min. then let's calculate everything else. entry is a level to be examined.
2. really, we don't know how many times the bar will be crossed, so 1 entry per 1 bar, only min. we need the direction of penetration. that's why the simplest one. see the 3rd transaction in the picture. we don't wait for the end of the bar. if the open was below and the price was even 1 pip higher, it will break through upwards...gather statistics
3. let us assume 1 hour
the algorithm is simple, has minimum parameters, and allows you to compare it with, say, the 00+7 level. any other algorithm has more parameters, which complicates comparison and analysis
the criterion is not bullshit. the point is to have these 4 figures, entry, exit, max and min. then we'll calculate everything else we need. entry is the level under study. the other three figures need statistics
for point 2. really, how many times in a bar we do not know, so 1 entry per 1 bar, only minutes. we definitely need the direction of penetration. so the easiest. look at the 3rd trade of the figure. we do not wait for the end of the bar. if the open bar was below and the price was even 1 point above the level, all this is a breakdown up...
If I don't know what desert it is, let's assume it's Kara Kum :) . I agree, a perfectly sensible approach to get into action.
The problem is that the number of entries (i.e. the number of crossings) will determine the contribution of that level to the overall total. If the criterion for limiting the number of crossings is chosen arbitrarily, the statistic will only reflect this particular case. So this statistic won't convince anyone, there will be a "more correct" way to set it. This is why I want to have some logic in the choice of methodology from the start.
I see. the comparison criterion needs to be written down first, until we know how to compare, there is no point in doing a bad job))
I think we should not limit the number of crossovers. We'll act as a statistician drawing distributions of these 3 values and finding the mean, variance. max and min for the sample, median. but how to choose the best one we need to think.
The simplest way is to compare separately the best input, the best output and the best profit.
But an aggregate criterion, which takes into account all at once, is something to think about.
I can process everything myself in Matcad, i.e. build histograms, estimate distribution law, get distribution parameters, compare. the main thing is to give me the data for processing
.
I suggest the general criterion is as follows, since we have no output, we don't check it in principle, we have no output logic.
We check the entry - subtract the drawdown from the maximum, if the distribution law is normal, then we take the average value, though most likely we will have to take the median.
The best level is the one with the biggest value, the ideal level - the price goes in the entry direction, without the drawdown...
How is that a criterion ?
Maybe you can do it this way?
x=DoubleNormalize(Close[i],3) ;
line=DoubleNormalize(Close[i],2);
if(x==line) flag = true;
if(flag && x>line) { up=1; flag=false; }
I agree. There is no getting away from this parameter. But here we have tried to get down to business and are immediately faced with the question of how to define 'level departure'. And I still do not understand how we can do without one more parameter.
I think we did not understand the interpretation of the word "level". I wrote about the levels of the market fractal structure. And you, as it seems to me, are speaking about the levels of price, i.e. its absolute values.
Suppose we want to trade intraday but don't want to do pips. Suppose that the daily price movement on our selected currency pair is 70-80 points. Then we can set a target profit for the trade of 50 pips. This number can be taken as a basic parameter because it can be said to unambiguously determine the zigzag that will represent, on the one hand, the ideal pattern for our trade, and, on the other hand, the fractal structure of the market at the selected level.
In this context, there cannot even be any question of moving away from this level, as the TS should be built to best follow price movements of 50 pips or more, rather than any price movement around some fixed level.
I realise that I got into your discussion of a particular TS that trades from absolute levels. Of course, this is a completely different approach to trading, which incidentally still requires its justification, at least statistical. Otherwise we cannot avoid arbitrary choice of these levels. That is why I meant in general terms, in the sense of the best approach to TS creation. IMHO
In general, there is one more important criterion in this question. Since we're going to gather statistics on history, the algorithm should not be too "heavy". And since we are going to work with levels, it would be natural to apply a criterion to the distance between the levels. For example, we divide this interval by a power of two, so this power could be the parameter. Importantly, it has a very small range of reasonable values, 2, 3, well 4.
For this problem, though, it would be natural enough to divide by 10, but somehow it seems inevitable to me that the question about other ways of dividing into levels will arise :).
By the way, I remember I once forced the algorithm of calculation of Murray levels exactly to make it possible to work with them on a long history :).
Degree of two, and it is the first one, seems to me the most acceptable value. In my opinion, it most corresponds to the fractal market structure. Several years ago, when Vladislav put forward the indicator of Murray's levels, I estimated its adequate behavior to the price fluctuations. The only thing that I did not like was the basis it was bound to. The basis for its calculation was zero price value. I think it was wrong.
In general, if we trade by levels, we obtain at least three parameters: 1) basis, i.e. at least one absolute value that sets the grid support; 2) interval between levels - it fully corresponds to the parameter of target profit value I wrote about; 3) parameter that defines departure from the level. If we construct finer grid levels by dividing the interval in two, the level escape is determined naturally.
But dividing by 10 seems to me to be a tribute to our habit to the decimal scale. This can hardly be justified.
I think that's where we disagree on the interpretation of the word "level". I wrote about levels in the fractal structure of the market. And you, as it seems to me, are talking about price levels, that is, about some absolute values of price.
Suppose we want to trade intraday but don't want to do pips. Suppose that the daily price movement on our selected currency pair is 70-80 points. Then we can set a target profit for the trade of 50 pips. This number can be taken as a basic parameter because it can be said to unambiguously determine the zigzag that will represent, on the one hand, the ideal pattern for our trade and, on the other hand, the fractal structure of the market at the selected level.
In this context, there cannot even be any question of moving away from this level, as the TS should be built to best follow price movements of 50 pips or more, rather than any price movement around some fixed level.
I realise that I got into your discussion of a particular TS that trades from absolute levels. This, of course, is a completely different approach to trading, which, by the way, still requires its own justification, at least statistical.
Maybe you can do it this way?
x=DoubleNormalize(Close[i],3) ;
line=DoubleNormalize(Close[i],2);
if(x==line) flag = true;
if(flag && x>line) { up=1; flag=false; }
Yes, like this, only if(flag && x[i]>=line && x[i+1]<line) ....
However:
1. We need to clarify which is faster, through NormalizeDouble or through a transition to integers. as I did above. NormalizeDouble always seemed like a slow function to me, but I need to clarify.
2. We should not work with Close, but with High and Low, otherwise we won't have a chance to reproduce the algorithm of inputs in real time, imho.
I suggest the general criterion is as follows, since we have no output, we don't check it in principle, we have no output logic.
We check the entry - subtract the drawdown from the maximum, if the distribution law is normal, then we take the average value, though most likely we will have to take the median.
The best level is the one with the biggest value, the ideal level - the price goes in the entry direction, without the drawdown...
How is that a criterion ?