EA scams - page 2

 
stuartmiles # :

I am pretty convinced there is something underhand going on and others on here have said there is no way of checking so for me the marketplace is not an option anymore. I am not saying that all sellers are cheating but if there is no way of checking then it is all pointless. Most of the ones I tried don't even pass a backtest so at least he made an effort to make it look like it works :-)

I agree that the programs need to be regularly optimised but it is no good if they don't work on future data. I don't think that is what happened with this EA as it has given 90% win history over 6 years even after the 'opimisation' has been changed, but it doesn't make a profit after the latest upload date.

Anyway, if one person reads this and it makes them look before they leap then it has gone a little way to help.

EAs using Historical Database can be very easy to detect. You can run it against a Custom Symbol (using a totally cryptic name, but copying the original symbol data).

If the results are the same, it is not using historical database. If it is totally different, then yes, it is most likely using historical database.

Another way, is to run optimisations and see if the various passes are different enough, or run optimisations on different symbols and see how they behave.

 
Fernando Carreiro #:

履歴データベースを使用する EA は非常に検出しやすいです。カスタム シンボルに対して実行できます (完全に不可解な名前を使用しますが、元のシンボル データをコピーします)。

結果が同じ場合は、履歴データベースを使用していません。まったく異なる場合は、はい、履歴データベースを使用している可能性が高くなります。

別の方法は、最適化を実行してさまざまなパスが十分に異なっているかどうかを確認するか、さまざまなシンボルに対して最適化を実行してそれらがどのように動作するかを確認することです。

I wanted to know the way of custom symbol a few days ago. I will try it with the EA I bought recently.

 
Fernando Carreiro #:

EAs using Historical Database can be very easy to detect. You can run it against a Custom Symbol (using a totally cryptic name, but copying the original symbol data).

If the results are the same, it is not using historical database. If it is totally different, then yes, it is most likely using historical database.

Another way, is to run optimisations and see if the various passes are different enough, or run optimisations on different symbols and see how they behave.

not most likely , but possible i.e 50/50.

If an ea has a neural network that has a configuration for a symbol and you change the symbol name then it cannot know what configuration to load , it will fail.

It works both ways too , so , if you create a history reader you add "neural network" lingo in the description .

 
stuartmiles #:

Lorentzos, I am not sure about neural networks or "change symbol name" but I think what you are saying is that there is no way of telling if someone is cheating as there are several ways to do it.

Yeah they really can't check unfortunately . 

For neural nets they could do some sort of the vendor is sending the onnx model to the server and the server sends it to the eas bought . Then they could indicate the EA has a verified NN , and , uses it to trade , and prohibit the use of neural network keywords and AI etc in products without that classification. But they would have access to the neural network , i.e. the core of the EA .  (they might be probably considering a pivot like this as they are preparing an AI trading championship where contestants will upload models not eas)

 
Lorentzos Roussos #:

Yeah they really can't check unfortunately . 

For neural nets they could do some sort of the vendor is sending the onnx model to the server and the server sends it to the eas bought . 

I had turned my internet connection off as I suspected they were connecting to an external database and that was ok so I thought it was ok but it was only when I saw the other EA didn't work after the update date that I realised it was probably something within the code. 

 
Lorentzos Roussos #:

not most likely , but possible i.e 50/50.

If an ea has a neural network that has a configuration for a symbol and you change the symbol name then it cannot know what configuration to load , it will fail.

It works both ways too , so , if you create a history reader you add "neural network" lingo in the description .

If it is using a hard-coded pre-trained neural network that does not re-train itself, then it is effectively equivalent to using Historical Database.

The end result will be the same, that it will not adapt to the market and will fail in future conditions.

 
Fernando Carreiro #:

If it is using a hard-coded pre-trained neural network that does not re-train itself, then it is effectively equivalent to using Historical Database.

The end result will be the same, that it will not adapt to the market and will fail in future conditions.

No . These are miles apart , we cannot conflate them into the same category .

If you place them in the same category then you should say not "most likely" but "definitely".

The "EA review" broadcasters do the same thing , they conveniently fail to mention that there is a chance the poor thing cannot trade if it does not know what its supposed to trade.

However that does not mean the scammers are not taking advantage of that . 

EAs with NN are easy targets but also an easy cover (for an EA to "pretend" its an NN EA) 
 

It is easy to find fake neural network  EA in market place:

  • many updates in a year like more than 1 per month i.e version 15,
  • they just code bot not to trade on date when martingale blew up.
  • Upward equity curve is an uninterrupted constant straight-line, red flag.
  • If it is too good to be true...

I think the people who were fooled by the above, wanted to be fooled.

 
Lorentzos Roussos #:

No . These are miles apart , we cannot conflate them into the same category .

If you place them in the same category then you should say not "most likely" but "definitely".

The "EA review" broadcasters do the same thing , they conveniently fail to mention that there is a chance the poor thing cannot trade if it does not know what its supposed to trade.

However that does not mean the scammers are not taking advantage of that . 

EAs with NN are easy targets but also an easy cover (for an EA to "pretend" its an NN EA) 

Sorry, but I am unable to agree with you! But it does not really matter to me, as I don't use Market products.

I will leave it up to the OP and others to come to their own conclusions.

 
Philip Kym Sang Nelson #:

It is easy to find fake neural network  EA in market place:

  • many updates in a year like more than 1 per month i.e version 15,
  • they just code bot not to trade on date when martingale blew up.
  • Upward equity curve is an uninterrupted constant straight-line, red flag.
  • If it is too good to be true...

I think the people who were fooled by the above, wanted to be fooled.

That is the biggest one or a J curve

This is not necessarily a red flag , many vendors update often as they optimize sets in order for the product to be a one click away from the "proper" test . 

I mean if you are competing in  a market where EAs without explanation and politicians names get sales then you want the test process to be as hassle free as possible.