Evgeniy Scherbina / Profil
- Information
10+ Jahre
Erfahrung
|
27
Produkte
|
622
Demoversionen
|
0
Jobs
|
1
Signale
|
0
Abonnenten
|
What it means is that USD should let it go. AUDUSD, EURUSD, GBPUSD may instantly go upwards, while USDCAD, USDCHF, USDJPY will head downwards.
The questions is: Will this move in currencies take place after the conference on July 31st or later in September?
My answer is: It will happen this Wednesday. I am going to remove all positions favoring USD. And I recommend doing so to you.
My major symbols are: AUDCAD, AUD, EUR, GBPCHF, GBP, CAD, CHF, JPY, and SEK. While the supporting symbols are: Oil, Dixie, Dax, Ftse, and Gold. Yes, you got it right! I have finally incorporated stock indices into Forex trading. All of which is available with any broker, because the rare and valuable staff is downloaded from Yahoo Finance.
What we see in this chart, is that Oil, Dixie, Dax and Gold are mostly winning. The one loser is Ftse. Oil lost 2 major symbols in the test period Jan-May 2024, while Dixie lost only one major symbol. But Dixie has the highest drawdown of 10% on SEK, together with Ftse. The drawdowns are not shown in this chart, but surely it is not only about how tall the columns are.
I think I will put it all together in one strategy, including the losing Ftse, so that everyone knows what an amazing and extensive testing has been done to achieve these results. While Oil should be the preferred choice, the other options will add confidence to it.
It uses the American Dollar Index, or "Dixie", as a supporting symbol.
LuminaFX check it out here: https://www.mql5.com/en/market/product/114681
LuminaFX is a fully automated advisor which trades 9 symbols from one chart: AUDCAD, AUDUSD, EURUSD, GBPCHF, GBPUSD, NZDUSD, USDCAD, USDCHF, USDJPY, and USDSEK. The advisor implements a popular version of the recurrent neural model. Values of the American Dollar Index DXY, or "Dixie", are used as inputs to the neural model. Make sure you have the 6 symbols making up the American Dollar Index in your terminal: EURUSD, USDJPY, GBPUSD, USDCAD, USDSEK, and USDCHF. Otherwise the advisor will not work
а на RannForex он как USDX - наверное надо передавать имя символа советнику в параметрах?
Пошёл тестить на IC Markets ;)
Likewise, right now I am seeing that it is QuantumPip slowing down, while LuminaFX is really pushing up. They differ in the way inputs are fed: stacked sequences for QuantumPip, ordered sequences for LuminaFX. There is a reason why they may be opening opposite trades.
And as we all know, it is better to open more trades because more risk creates more profit. If the forecast quality is at least 51% (it is so for both QuantumPip and LuminaFX), a lot of trades would always be winning at the end of a month. A few good trades may be pure luck, it may last 1 month, 2 months... But a lot of trades will always create enough opportunities to grab profits!!
What do you think?
The advisor "QuantumPip" is a fully automated advisor which can trade simultaneously 10 symbols: AUDCAD, AUDUSD, EURUSD, EURUSD, GBPCAD, GBPCHF, GBPUSD, NZDUSD, USDCAD, USDCHF, and USDJPY. The advisor also uses prices of Gold and Silver to calculate inputs for the symbols. Currently I am doing tests to add 10 cross-currency symbols more. The advisor uses 2 types of recurrent neural model - 1 network (decisions "buy" or "sell") and 2 networks (decisions "buy" or "uncertainty" and "sell" or
The thing is QuantumPip, which is my most advanced system so far, still shows better results. I can't make LuminaFX outperform QuantumPip. But I will keep on searching. I think if the past has the same random prices as the future does, a good strategy should be able to pass the history with a satisfactory result. LuminaFX struggles to do it. Do I have to come to the conclusion that it is different: To be able to pass the random prices of the future, a good strategy should not fit to the prices of the past at all???
IDixie is an indicator which shows open and close values of the American dollar index, DXY or "Dixie". The indicator is calculated from weighted values of 6 major symbols: EURUSD, USDJPY, GBPUSD, USDCAD, USDSEK, and USDCHF. In addition to "Dixie" values, the indicator also shows a moving average. You can set the period of the moving average. When bars of the indicator break through the moving average, this creates good opportunities to enter the market. As a rule, signals from this indicator are
The indicator IQuantum shows trading signals for 10 symbols in the daily chart: AUDCAD, AUDUSD, EURUSD, GBPCAD, GBPCHF, GBPUSD, NZDUSD, USDCAD, USDCHF, and USDJPY. Signals of the indicator are produced by 2 neural models which were trained independently from one another. The inputs for the neural models are normalised prices of the symbols, as well as prices of Gold, Silver and markers of the current day. Each neural model was trained in 2 ways. The Ultimate mode is an overfitted neural model
The good thing to know is it can't grow endlessly. It must fall. There is no level that is higher than 2100. So no target value to reach and to roll back. It can roll back any day. Any time. Can it roll back today?
The bad thing, or the bad consequence is evident. No one expected it to grow that high. It just broke all indicators. It has made us incur losses and suffer the pain of frustration. It is not fair! Go back, you brick of bright shiny gold! Go back now!
My two advisors - Gold Chaser and Neural Rabbit - both trade Gold. Their neural model for Gold is identical, and it could not foresee the trend beyond 2100 because it has never happened in the history on which it was trained. Clearly, I will have to retrain the model. To avoid that kind of miscalculation in the future, I am going to try the following ways:
1) Extend the training period.
2) Drop the trade-off between a good historical chart and a good performance on the unseen future data. What this means, is I tried with these 2 advisors to accomodate the 2 approaches. I wanted to please those traders who want to be sure it can pass the history well. And I tried to find such values of neural loss which would allow the strategy make correct trading decisions on the unseen future data. That is why it broke. No more trade-offs! Trade-offs don't work! Or else they work until the market spirals out of control.
3) My latest approach is to leave out 75% of training data to avoid repetitions in the training dataset and make it less addicted to such repetitions. The following validation includes a 100% of the dataset, so I can always make sure it is a valid neural model.
We do not succumb to the frustration wrought by the power of the unwinding market! We adapt and we become stronger! Profit will be ours!
Below is the picture of how it may look like. I am still thinking of what could be done. I hope I can publish it this week.
It is going to be based on the same data as the advisor "QuantumPip", but it will be a free indicator, not an advisor by itself. Some users like to see how trading decisions were made.
If you have any comments or suggestions, you are invited to comment here or you can write to me via the messages.
To address this issue, you may consider implementing the following strategies:
- **Educational Materials**: Develop educational materials or presentations that explain the concepts of overfitting, generalization, and the importance of future performance in a simple and accessible way.
- **Comparative Analysis**: Compare the performance of overfit models and properly trained models on real-world data to demonstrate the impact of overfitting on future performance.
- **Risk Assessment**: Discuss the risks associated with overfitting and the potential consequences of basing decisions solely on visually appealing historical charts.
- **Engagement**: Encourage customer engagement by involving them in the model evaluation process and demonstrating the value of robust, generalizable models over visually appealing but overfit models.
Neural networks open possibilities. However, too many users these days do not understand the possibilities.
Neural networks learn prices, fluctuations, patterns... In fact, whatever you feed into a neural network (really - whatever!!), it will learn it. Can I feed random prices? Yes! Can I feed random prices of random symbols? Yes! Can I feed bullshit? Yes! It will learn it.
It will learn it. It is not a problem. The problem arises from learning too much.
Why this post? Because too many users buy bad strategies that show an incredible growth in the historical chart. Too many users do not buy good strategies because they think that a mediocre historical chart will not allow them to make a profit on Forex. Good strategies fall into oblivion, that is to the bottom of the ranking system, which is based on greed and not on analysis or credibility.
So what? Make a real live signal, run it for half a year and show a profit with your good strategy. Prove that it works. Yes, this is what I tried. And they still don't buy, because they found hundreds of strategies that show an incredible growth (from 1k to 1 trillion in several years, literally) in the nice historical chart.
Why this post? I do not want to be a liar. I do not want to sell nice historical charts.
This new strategy "Intraday Rush" shows several levels of accuracy in the training. These levels are defined by the property "Trading pattern". The "reliable" level has a much higher chance to win in the unknown period. While the "ultimate" level, despite the mind-blowing profitability in history, has a much worse chance of winning in the unknown period. This is it. I have written it. I have explained it.
Does the market never repeat itself? Is it true that, from this standpoint, an absolutely unprofitable strategy in history should have the best chance to win in the unknown period? Maybe yes!!!
This is what I think. History has some valuable patterns. A good strategy should try to learn the very basic pattern-like moves from history. Therefore, it should demonstrate profitability in any long period running for at least a year. A good strategy should have drawdowns and wrong trades leading to frustration in some periods. And it should win in other periods. This is what makes a good strategy.
Sometimes users do not want a profit on Forex. They want to buy a hope at this particular moment of their lives and be happy for a few days with that false hope. Yes, I understand it. Switch to the "ultimate" level, this is your false hope. Some users can't stop believing that a nice historical pic will not lead to a profit in the unknown period. Yes, I understand this, too.
Why this post? After months or even years of playing around with so many "nice" strategies out there, come back and read this post again. And switch to the "reliable" trading pattern of "Intraday Rush".
It is going to be the same logic, but you will be able to choose between daily and hourly trading. 2h-timeframe has 12 times more trades than daily. It also bring more flexibility of course, because the advisor can react to market fluctuations much faster.
The indicator "IRush" will also be updated! So stay tuned! The new update will be published no later than next week!
myfxbook.com/members/otvazhnii
It has all my best strategies - Neural Rabbit, Gold Chaser etc - updated live. More strategies will be added in the very near future. All in one chart, it is so easy to compare!
If you want to drill down into details, it has lots of nice diagrams and piecharts. Profit per month, missed profit and more. Go check it out!
Intraday Rush 95$ down from 385$
Neural Rabbit 85$ down from 215$
Tokyo 85$ down from 215$
Merry Christmas and Happy New Year!!
End of Year Sale!!
Limited Time Only!!