Scam EAs in Market - page 4
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
Why do you interpret that the use of AI in an EA is obligatorily linked to ChatGPT?
Beyond ChatGPT, AI has been used for several decades to a greater or lesser extent.
I'm aware of that, but in this specific example, ChatGPT is used to forecast the market, if we believe the author's advertisement.
There are some logic gaps in this product. The 48-month history rises perfectly smooth like butter.
So why wait 4 years to publish it? Just to integrate ChatGPT that was released a few months ago? Then this "forecast feature" has nothing to do with the trading of the past 4 years.
It must be said that the authors are getting more and more clever to cheat the customers. They have years of experience, and most customers do not.
It is really difficult, even for Metaquotes, to establish a certain quality standard here when they work with all available tricks and even brokers help in the game (unproven but obvious).
I'm aware of that, but in this specific example, ChatGPT is used to forecast the market, if we believe the author's advertisement.
There are some logic gaps in this product. The 48-month history rises perfectly smooth like butter.
So why wait 4 years to publish it? Just to integrate ChatGPT that was released a few months ago? Then this "forecast feature" has nothing to do with the trading of the past 4 years.
It must be said that the authors are getting more and more clever to cheat the customers. They have years of experience, and most customers do not.
It is really difficult, even for Metaquotes, to establish a certain quality standard here when they work with all available tricks and even brokers help in the game (unproven but obvious).
If they find a way to "enforce" that they will also weed out the real neural network eas . The authors will then have to resort to files being loaded per config externally (by user input) and they will lose the default demo functionality and the sales that result from that . I imagine part of the advantage the "scams" have (the real scams) is that the user clicks on test and it goes up , they must alter no parameters , no need to read any instructions ,even if they test it by accident while browsing its appealing.
I'm aware of that, but in this specific example, ChatGPT is used to forecast the market, if we believe the author's advertisement.
There are some logic gaps in this product. The 48-month history rises perfectly smooth like butter.
So why wait 4 years to publish it? Just to integrate ChatGPT that was released a few months ago? Then this "forecast feature" has nothing to do with the trading of the past 4 years.
It must be said that the authors are getting more and more clever to cheat the customers. They have years of experience, and most customers do not.
It is really difficult, even for Metaquotes, to establish a certain quality standard here when they work with all available tricks and even brokers help in the game (unproven but obvious).
In the title of the signal it says that the AI Filter is not used, so the author could finally be being honest. Time will tell.
Otherwise I agree with you. Customers have to jump through too many hoops before they find a genuine product and it is difficult for MetaQuotes to control.
An AI/NN is a function f(x) = y...
I really get your point. Now let's say, they used GPT for coding and/or generating an AI/NN, if you fit it to a generalized dataset, it could be argued it is not over fitted, but actually "detecting" features that are relevant to changes/reversals/continuations in the market. This could be also working on more historical data. Which could be even seen for a proof of working concept/fitting.
Nether the less, I was thinking if it were possible to build an EA/Service that enables the detection of scam EAs. Something along the lines of creating a custom symbol with interchanged and shifted data, to create a Datastream that a developer could impossibly foresee and therefore an overfitting would be revealed. Also preconfigured open/close sequences could be found this way.
One major issue is for sure, what is a scam? This is for sure including a gray scale where, as mentioned by Lorenzo's, the developer is not even aware of an overfitting. The problem I see here is the fact of general working versus slightly over fitted. From what point onwards is it a scam?
But I am sure, a good analysis environment for evaluating EAs for sure includes some scale, from 0.0 to 1.0, or non Scam to def scam. With an area where the result is questionable.
Such an environment could include also a general, centralized statistic database to evaluate the results against.
Could be a very interesting project.
Arguably... Yes, indeed.
An AI/NN is a function f(x) = y...
I really get your point. Now let's say, they used GPT for coding and/or generating an AI/NN, if you fit it to a generalized dataset, it could be argued it is not over fitted, but actually "detecting" features that are relevant to changes/reversals/continuations in the market. This could be also working on more historical data. Which could be even seen for a proof of working concept/fitting.
Nether the less, I was thinking if it were possible to build an EA/Service that enables the detection of scam EAs. Something along the lines of creating a custom symbol with interchanged and shifted data, to create a Datastream that a developer could impossibly foresee and therefore an overfitting would be revealed. Also preconfigured open/close sequences could be found this way.
One major issue is for sure, what is a scam? This is for sure including a gray scale where, as mentioned by Lorenzo's, the developer is not even aware of an overfitting. The problem I see here is the fact of general working versus slightly over fitted. From what point onwards is it a scam?
But I am sure, a good analysis environment for evaluating EAs for sure includes some scale, from 0.0 to 1.0, or non Scam to def scam. With an area where the result is questionable.
Such an environment could include also a general, centralized statistic database to evaluate the results against.
Could be a very interesting project.
Great idea . Maybe something with uploading the onnx model could be done too , if the server was "airtight" of course , so you get in the AI/NN category if you upload the model , but then you have issues when you update it , weights change etc . If i were they i would actually enforce the rule about promising $$$ , that's step 1 and that's the first thing a scammer will advertise.
Great idea . Maybe something with uploading the onnx model could be done too , if the server was "airtight" of course , so you get in the AI/NN category if you upload the model , but then you have issues when you update it , weights change etc . If i were they i would actually enforce the rule about promising $$$ , that's step 1 and that's the first thing a scammer will advertise.
ow i thought you meant a generated EURUSD chart based on EURUSD price action .
No then i disagree . This is the method the "saviors of clients (come buy mine not theirs)" use .
So as in the first reply : There is no way for a neural net using ea (or ml using ea) with an internal configuration to know which model to load other than reading the symbol (as in which symbol its trading on).
If you have good models per symbol and the service uses AUDCAD instead of EURUSD then they will brand you a scammer automatically.
This is the nuance missed by many that you can't tell if its a history reader or has an internal ML config for real , and it goes the other way around too.
Guys, this topic is not really about AI or NN or ChatGPT.
There is no way to avoid someone to try to scam others beforehand. There is also no way to detect that automatically and reliably, you will always have false positive and false negative.
What could be done is to request higher standards for the EA being published, like a minimum track records (and even there you will see people trying to cheat).
An other thing to do is to invest for time to investigate the complaints, but that's a difficult problem because you will have all kind of fake complaints against honest products fro example. And even honest complaints are not necessarily meaning the product is a scam.
It's a very difficult issue to deal with.
ow i thought you meant a generated EURUSD chart based on EURUSD price action .
No then i disagree . This is the method the "saviors of clients (come buy mine not theirs)" use .
So as in the first reply : There is no way for a neural net using ea (or ml using ea) with an internal configuration to know which model to load other than reading the symbol (as in which symbol its trading on).
If you have good models per symbol and the service uses AUDCAD instead of EURUSD then they will brand you a scammer automatically.
This is the nuance missed by many that you can't tell if its a history reader or has an internal ML config for real , and it goes the other way around too.
Well, either I am not understanding your point, or thats exactly what I was trying to aim at.
You generate a Symbol, EURUSD, using some other symbols data, like AUDCAD and you apply a shift, so that it has values comparable to EURUSD. Then you apply a time shift as well (Maybe not even necessary, but anyways).
Now you let the EA run on this synthetic symbol in a backtest. Even if the EA should detect based on the price stream, it is EURUSD and loads the corresponding NN, if it is overfitted, it will fail, if it is not overfitted, it will succeed, because it is able to identify the relevant features.
Now by the detailed report it should be able, when compared to the detailed report of the real EURUSD results, if the EA is a history-reader or has an overfitted NN....
Both should be now quantifyable by applying such a method. And this is your measurement value, normalized to 0.0 to 1.0 (Or as percent 0% to 100%) And therefore give you a good reliability interpretation of the underlying algorithm.
Guys, this topic is not really about AI or NN or ChatGPT.
There is no way to avoid someone to try to scam others beforehand. There is also no way to detect that automatically and reliably, you will always have false positive and false negative.
What could be done is to request higher standards for the EA being published, like a minimum track records (and even there you will see people trying to cheat).
An other thing to do is to invest for time to investigate the complaints, but that's a difficult problem because you will have all kind of fake complaints against honest products fro example. And even honest complaints are not necessarily meaning the product is a scam.
It's a very difficult issue to deal with.