What is the optimum depth of history for identifying a useful signal? - page 17
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 did not expect such a response from the respected one ... OK ... say the price is a person and he stands somewhere ... so I ask the question where he stands and why ... answering them paints a picture of the market
What's there to be surprised about, okay, the question of where the price is, it seems to be on the screen. But the question of why it is there is practically absolute demagogy. And besides the answer to it-because so-and-so ago it was there and now it is here-is an example of this demagoguery. And this is not an answer, but a statement of fact that follows from the first question.
It would be interesting to hear an answer on why.
The impression is deceptive, I tasted it.
Nevertheless, can we get back to where we started?
So I was asking, if we trade intraday on minutes, and the deal is supposed to hold on average half a day, what would be your optimal correct depth of history in bars (i.e. time)?
All depends on the TS and its conditions. The deal is usually closed when TP or SL is reached, as well as on the reverse signal of the indicator. I did not try to close trades after a certain period of time. For example, the TS I am using, gives the best performance on H1 data from 01 01 2014 to date, with an optimal number of bars of history = 240 bars, which is 2 weeks of trading. Why exactly 2 weeks - I do not know.
There are 6014 bars in the history.
11027 ticks simulated
Modeling quality n/a
Chart mismatch errors 0
Initial deposit 1000.00
Spread Current (51)
Net profit 10864.18
Total profit 21090.43
Total loss -10226.25
Profitability 2.06
Expected payoff 4.32
Absolute drawdown 941.41
Maximum drawdown 1789.48 (20.23%)
Relative drawdown is 94.14% (941.41)
Total trades 2516
Short positions (% win) 1873 (87.72%)
Long positions (% win) 643 (51.01%)
Profitable trades (% of all) 1971 (78.34%)
Loss trades (% of all) 545 (21.66%)
Largest
profitable trade 11.00
losing trade -21.43
Average
profitable trade 10.70
losing trade -18.76
Maximum number
continuous wins (profit) 335 (3663.71)
Continuous losses (loss) 48 (-965.65)
Maximum
Continuous Profit (number of wins) 3663.71 (335)
Continuous loss (number of losses) -965.65 (48)
Average
continuous winnings 23
Continuous loss 6
yosuf:
Why exactly 2 weeks is unknown to me.
Short positions (% of winners) 1873 (87.72%)
Long positions (% of winners) 643 (51.01%)
You look like a father working hard)))) Sometimes look at the chart
It would be interesting to hear from you as to why.
I must be in the wrong forum ... The answer to where the price is, is technical analysis (trend, channel, resistance levels, etc.) I study the price chart.
To answer the question why I look at the fundamental factors that the market participants take into account (i.e. I see the market mood and its interests).
So I was asking, if we trade intraday on minutes, and the deal is supposed to hold on average half a day, what would be your optimum correct depth of history in bars (that is time)?
I must be in the wrong forum ... The answer to the question where the price is is technical analysis (trend, channel, resistance levels, etc.) I mean I study the price chart.
To answer the question why I examine the fundamentals that are taken into account by market participants (i.e. I see the market mood and its interests).
But the fundamental will not give a clear answer to why the price is there.
If it does, it will not answer within the time frame you were talking about.
Actually there is nothing to discuss, you did not want to answer, that is, you either give an answer about the depth of history,
Half a day is 720 minutes. I cannot understand why a one-minute chart is needed for such a long trade. At such a sampling rate using MT tools no signal processing is possible, except for the most primitive ones, like standard indicators, the uselessness of which has long been obvious to all.
The question was to a specific person, regarding his predispositions in the analysis.
As for you, you've done the analysis on minutes and weeks, and you didn't ask such questions before https://forum.mql4.com/ru/42459/page2#503868.
And now it's impossible.
The question was to a specific person, regarding his predispositions in the analysis.
As for you, you've done the analysis on minutes and weeks, and you didn't ask such questions before https://forum.mql4.com/ru/42459/page2#503868.
In that post it literally says the following: Changing the timeframe means changing the sampling rate and nothing else. I wrote the same thing here as well. Those pictures just show decomposition of price into several components by filters. It's not the most complex thing but even that requires significant computing resources. And now I'm making an extrapolator. Calculation of 32 bars on a medium lousy computer takes 50 ms, while calculation of 64 bars takes 500 ms. Just thinking about 720 bars makes me sick...
In principle, you're probably right that we should use a considerably longer history than the duration of the trade. But I understand it in such a way that I should take several tens of bars and use them to calculate the next 1-2-3 bars. At the next bar we can repeat it.
That post literally says the following: Changing the timeframe means changing the sampling rate and nothing else. I wrote the same thing here as well. In those pictures it's just a filter decomposition of the price into several components. It's not the most complicated thing but even that requires significant computing resources. And now I'm making an extrapolator. Calculation of 32 bars on a medium lousy computer takes 50 ms, while calculation of 64 bars takes 500 ms. Just thinking about 720 bars makes me sick...
In principle, you're probably right that we should use a considerably longer history than the duration of the trade. But I understand it in such a way that I should take several tens of bars and use them to calculate the next 1-2-3 bars. On the next bar we can repeat it.
I don't know how else to argue... I will send you the link, maybe it will tell you better.
https://forum.mql4.com/ru/4368/page13#251394
where the picture is.
I stick to the same opinion so far, only whether the depth of 11500 bars is necessary in that case (what is in the link), that's what I mean.
Or about these pictures
https://forum.mql4.com/ru/24013/page52
or about your own https://forum.mql4.com/ru/42459/page2#503868 for example in this picture for m1 what kind of history you take, 64 bars if the slowest blue line has a period far beyond hundreds of bars... that's what I mean. And I don't believe in prediction based on 32 or 64 bars...
I don't know how else to argue... I'll send you a link, maybe she can tell you better.
https://forum.mql4.com/ru/4368/page13#251394
where the picture is.
I stick to the same opinion so far, only whether you need the depth of 11500 bars in that case (which is in the link), that's what I mean.
Or about these pictures
https://forum.mql4.com/ru/24013/page52
or about your own https://forum.mql4.com/ru/42459/page2#503868 for example in this picture for m1 what kind of history you take, 64 bars if the slowest blue line has a period far beyond hundreds of bars... that's what I mean. And I don't believe in predictions based on 32 or 64 bars...
It's easier for me to comment on my own pictures than other people's pictures, especially since those other people's pictures look creepy. I took the long story to paint the story. I never even thought about using all that data for prediction.
Conversely, I don't believe in prediction based on thousands of bars. I mean I do, but it's a meaningless waste of computing power.
I can show you the spectrum of the 1st derivative of the price.
And here is what the forecast looks like:
The white line is the forecast and the green line is the actual result. The forecast was made on 64 bars to the left of the dotted line. Only the nearest few bars can be trusted. Low frequencies are not extrapolated very well, high frequencies perfectly. The accuracy of the high frequencies extrapolation is very slightly dependent on the sampling rate.