Machine learning in trading: theory, models, practice and algo-trading - page 319
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
Profit charts on "pattern pattern vs neural network".
Both models were trained to trade eurusd on the plus side in October 2016; constant lot, no stops or takeaways; always in the trade long or short; trade on H1 at opening prices. Trading on the chart - last 5 years, including one month of training data.
Learning models without crossvaluation, they just squeezed the maximum profit they could out of the price.
There is a place on the charts where the server did not give normal ticks, there is some kind of drain, that place ignore.
Here is the neuron. You can clearly see the time interval at which it was trained, this is the only place with a stable profit.
And here is the model with pattern recognition. The result is negative, but still better than neuronics. And there are plenty of times when it was profitable for weeks. But then it was lost.
It's cool, but it is not clear what to do with it.
I know the answer. I'm not at home right now. I'll be at my computer a little later, and I'll share my thoughts on it with you.
Here are a couple more graphs.
1) Neuronics cannot be trained without crossvalidation, it retrains with a 100% guarantee.
I took the same data, divided it into two parts for training and crossvalidation. You can see on the graph that profits on the training data are not increasing sharply upwards now because the model training stopped as soon as the crossvalidation result began to deteriorate. It did not have time to memorize all of the training examples and in theory should now trade by logic rather than by memory.
But it did not really affect the overall result, the total losses over 5 years is only 80 euros less (~10% of the total losses).
2) Model with patterns. Since training the model on 1 month of data produces a very irregular sinusoid with a period of about 2 months (but not exactly), it's worth trying to train the model on two months instead of one.
I want it to be a grail after that. But it appears illogical and incomprehensible - intervals of gains and losses of years, sometimes with abrupt transition from one to another. Someone flicked a switch in early 2013 - and the specific patterns used by the model turned on, although the model had no access to prices before August 2016. And then in 2017 - someone flicked the switch again - and those patterns began to drain dramatically.
If you train the model on different months, on intrevals of different lengths - each time you get amazing and unique results. Sometimes they are profitable, but it is never clear how long will a particular pattern last and whether it will sell out fast. Forex is not a constant randomness, it is an aggressive environment which sometimes behaves against the established rules, just to drain more people.
Here are a couple more graphs.
1) Neuronka cannot be trained without crossvalidation, it retrains with a 100% guarantee.
I took the same data, divided them into 2 parts for training and crossvalidation. You can see on the graph that profits on the training data are not increasing sharply upwards now because the model training stopped as soon as the crossvalidation result began to deteriorate. It did not have time to memorize all of the training examples and in theory should now trade by logic rather than by memory.
But it did not really affect the overall result, the total losses over 5 years is only 80 euros less (~10% of the total losses).
2) Model with patterns. Since training the model on 1 month of data produces a very irregular sinusoid with a period of about 2 months (but not exactly), it's worth trying to train the model on two months instead of one.
I want it to be a grail after that. But it appears illogical and incomprehensible - intervals of gains and losses of years, sometimes with abrupt transition from one to another. Someone flicked a switch in early 2013 - and the specific patterns used by the model turned on, although the model had no access to prices before August 2016. And then in 2017 - someone flicked the switch again - and those patterns began to drain dramatically.
If you train the model on different months, on intrevals of different lengths - each time you get amazing and unique results. Sometimes they are profitable, but it is never clear how long will a particular pattern last and whether it will sell out fast. Forex is not the constant randomness, it is an aggressive environment, which sometimes behaves against the established rules, just to drain more people.
Yes yes, but my TS also has ups and downs. In general, it is losing, but there are periods when it takes and on this subject there is a thought...... later a little bit...
Try mirroring Neuronka. What will it do? Will it dial in?
Try mirroring Neuronka. What will it do? Can she score?
No, even worse.
I wanted to write a long post about eqvity and show how I'm leaking Reshetova NS optimizer, but she did not, so excuse me. The lines indicate the area of optimization, work from the first of January and to this day.
I noticed it a long time ago that it is enough to determine the equity balance of the trading system and then one may enter the market, preferably with a drawdown. In other words, let's say 10 TS are working, but we have to choose the one that has started to fill up. As a rule I do it with the help of support and resistance lines, yes yes you are right on the balance curve.
But here's the thing. Hmm, I don't even know how to put it. The thing is, we're optimizing for balance growth as a rule.
And what if the optimizer is set up so that it is looking for the beginning of balance growth like this.
After all, a good result of optimization is exactly the type after which equity starts to grow.
So here we go!!! bros karapuziki help me I together we will be happy. The thing is that the sequent has two parameters natroek, the first varies from 4 to 10, the second about the same. As a result we have about 36 combinations or options, and taking into account that the second parameter can not be less than the first is even less. So !!!! Who could make the optimizer (preferably МТ4) find such parameters that would make equity look like this?
I.e., the beginning of the set. There may be several options in one optimization field and I should keep an eye on the one which is the best. So on!!!!! What do you think about it?
parameters, at which equity will be of the following form
This is so against all rules and logic that it might even work.
One more parameter to be added to the Expert Advisor, i.e., the breaking date.
If the Expert Advisor is optimized from January until the end of March, then the Breaking Date can be set at the beginning of March. Then, when making a trading decision in the code during testing - trade in the opposite direction, if the breakpoint date has not been reached yet. As a result, the equity graph will stably go upwards after optimization but we know that the first part would actually drain.
You can also make a second additional parameter with int type, it will mean a shift of the date of the fracture. It is not known in advance at what date everything should start to go up. And we can optimize it using genetics along with other EA parameters.
The point is that we are not going to talk about NS, but about Sequent, so look at its equity with different input values at the same period of work. I didn't optimize it, I just went through it by hand.
5-5
6-6
7-7
This is a much more interesting case with the 4-8 parameter, the same period. And it works with a stop loss of 300 pips.
This is from January 1 to the present day. I confess that I have picked the last screen with the optimizer, well, there are only 49 passes in the optimizer. That is, in fact, I have to learn to choose correctly the right parameters not from the dimensionless set, but from the finite set, and with such a small number of variants. So..........