Machine learning in trading: theory, models, practice and algo-trading - page 1007
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
Now I thought in addition, you can collect a collection of history and then check each pattern, it will be long, but on the hour markers will quickly count.
candlestick pattern analysis you mean?
Here's the trick...
Suppose the pair jumped up by 100 points within 4 hours, and at the beginning of the second 4-hours period it rolled back by the same 100 points
We have a "rail" pattern
let's assume the difference in server time between two brokerage companies is 1 hour
draw these two candlesticks in one and 2 DTs .........
What do you mean by candlestick pattern analysis?
here's the funny thing...
Suppose the pair jumped up by 100 pips in 4 hours and at the beginning of the second 4-hour period it rolled back by the same 100 pips
Suppose there is a 1-hour difference in server time between the two brokerage companies.
draw these two candlesticks in one and two DTs .........
I do not know what to call it probably a graphical figure. For example, I look for such a V-shaped trough:
I find such patterns in history:
But some of them are not exactly alike. A few, but it happens.
The rails are the simplest thing. I don't know why everyone admires them so much... I think it's just more candlelight bullshit.I do not know what to call it probably graphical shapes. For example, I'm looking for a V-shaped trough like this:
I find such patterns in history:
But some of them are not quite the same, but they are.
you can look up those who've spent a lot of time on these patterns.
there's tons of candlestick analysis indulgences on the net.
I personally - pass, because at the beginning of the path I already ran into something I wrote a little bit higher
Misha, I'm not arguing - I could be wrong about something. But, I see a real crisis of the genre in this branch.
I have a cool thing in my hands - NeuralNet package for VisSim - and I am afraid to even touch it, because I see, that quite smart people are not able to do anything here.
If in my branch, in diffusion processes, I know, then here I still have to read and learn. What is there to learn? That "it's all gone, neural networks don't work..."? You need at least 1 person with a positive signal, even if it's +1% per month. It would really inspire many people, me included.
You're getting some kind of decadent attitude from the maximum. It's not so bad.
I have a signal in my profile. A neuron with one hidden layer takes returns {open[i]-open[i-1]; open[i-1]-open[i-2]; open[i-2]-open[i-3]; etc.] on a new H1 bar. , uses them to make a forecast of the next returnee and then the EA opens a deal long or short depending on whether the forecasted returnee is positive or not.
The Expert Advisor is simple, the strategy is simple, everything is minimal, but still enough to beat the spread. It's been 5 weeks of trading and my balance is still not depleted.
As it has been said in the thread dozens of times, such a signal can be created in a couple of evenings using neuronics or the forest. For a real profit you need to spend a couple of weeks (months) to create a good target and predictors (even indicators from MT5 may be useful), and to train the neuronics on them.
p.s. Don't believe the whiners. There are a lot of smart tips in this thread, which you should check yourself, and not rely on someone's expressed "phi". Traditionally, the smartest advice here gets the most unfounded criticism (paid dealers-critics sixes?).
p.p.s a little math. This is a picture of trades from the signal. If the time series open[] were Markov, then any attempt to guess the color of the next candle would be impossible. The 309 trades would bring in an average of -0.04 cents each (spread), or -12 euros. And 11 profitable deals in a row would be possible with 0.5^11 chance.
The signal contains both profits and long series of profitable deals. It is worth thinking - is it really a Markov process?
You got some decadent attitude from the maximum. It's not that bad.
I have a signal in my profile. The neuron with one hidden layer takes returnees {open[i]-open[i-1]; open[i-1]-open[i-2]; open[i-2]-open[i-3]; etc.] on a new H1 bar. and then the EA will make a forecast of the next returnee and open a long or short position depending on whether the forecasted returnee is positive or not.
The Expert Advisor is simple, the strategy is simple, everything is minimal, but still enough to beat the spread. It's been 5 weeks of trading and my balance is still not depleted.
As it has been said in the thread dozens of times, such a signal can be created in a couple of evenings using neuronics or the forest. For a real profit you need to spend a couple of weeks (months) to create a good target and predictors (even indicators from MT5 may be useful), and to train the neuronics on them.
p.s. Don't believe the whiners. There are a lot of smart tips in this thread, which you should check yourself, rather than trusting someone's "phi". Traditionally, the smartest advice here gets the most unfounded criticism (paid dealers-critic pipsqueaks?).
p.p.s a little math. This is a picture of the trades from the signal. If the time series open[] were Markovian, then any attempt to guess the color of the next candle would be impossible. The 309 trades would bring in an average of -0.04 cents each (spread), or -12 euros. And 11 profitable deals in a row would be possible with 0.5^11 chance.
The signal contains both profits and long series of profitable deals. One should realize if it is a Markovian process indeed.
Doc, good for you, buddy.
Bravo, I would even say! And the algorithm you use is correct.
The process... That's what it's all about here. For forecasting purposes, you need "memory"-the dependence of the current value on the previous value. It is for my strategy that I need a Markov process, similar to the Ornstein-Uhlenbeck process with a return to the expectation.
Actually, both in this branch and in mine we need a key of "Markovianness/non-markovianness" ("no memory"/"with memory"; stationary/non-stationary).
I seem to have found this key. But, not sure yet - this month, based on bidding results, will be clear.
Here the coconut has fitted its own grail, following Sanych's martingale on unreal ticks and Aleshenka's negative error
Where are the normal strategies?
p.p.s. A little math. This is a picture of trades from the signal. If the open[] time series were Markovian, then any attempt to guess the color of the next candle would be impossible. The 309 trades would bring in an average of -0.04 cents each (spread), or -12 euros. And 11 profitable deals in a row would be possible with 0.5^11 chance.
The signal contains both profits and long series of profitable deals. It's worth considering whether it is a Markovian process.
I don't really understand your definition of Markovism, but I think it doesn't quite coincide with the usual one. For example, a trend (like in your picture) and a Markov is quite compatible.
The probability of such a series in such a sequence (even in the case of symmetric wandering) is higher (a problem from the field of elementary combinatorics).
As has been said in the thread dozens of times - such a signal can be created in a couple of evenings using neuronics or a forest. For a real profit you need to spend a couple of weeks (months) to create a good target and predictors (even indicators from MT5 may be useful), and train the neuronics on them.
Doc, it's not an example to follow.
Show me 2 weeks of work
Here the coconut has fitted its own grail, following Sanych's martingale on unreal ticks and Aleshenka's negative error
Where are the normal strategies?
You're putting up labels again. Maybe it's time to stop?
You're putting up labels again. Maybe it's time to stop.
I can't, take it as a half-joke, I'm not angry)