Why sell profitable EAs! - page 6

 
The hardest part is selling machines and subscriptions. Everything else (indicators, libraries, even semi-automatics) is an order of magnitude easier.
 
Assuming that there is no such thing as a consistently profitable system, but rather a constantly optimised one, then buying signals is fine, and a support team of reputable developers will constantly tweak the signals
 

Signals, on the other hand, are different! It's the same as not selling a hen laying golden eggs, but just selling those eggs for scrap! :)))

 
Belford писал (а): Systems that have passed the test described in Pardo's book are extremely rare (but they do exist) and are of great interest to serious investors. And, as practice shows, these systems are indeed quite robust.

This is encouraging, Belford, thank you: serious guys must demand real guarantees. Would it be too much of a secret to roughly hint at the real order of magnitude of the following figures for such systems:

1. Maximum percentage drawdown for the year

2. Profitability per year (percentage)

3. Cost

4. Size of initial capital?

5. Nature of the MM (permanent lot, reinvestment, etc.)

If you do not want to here, you can, and in private, the address is given in the profile.

P.S. I strongly suspect that the serious guys pay enormous money not for the space profitability, but for reasonable guarantees of a modest profitability at the level of tens of percents per year, with acceptable drawdowns. And the question of how the guarantees influence the system's value has not been raised here at all - because it is assumed by default that the system already meets these guarantees...

 
Mathemat: But I have the same question for you: are there any other criteria and tests, apart from those posted in the Alpari thread, by which you have evaluated your EAs - and if there are, can you post them as well?

I don't know what criteria I'm talking about. I seem to have only posted reports there, not my methods of evaluation. :)

I determine myself as follows: when developing an algorithm and then optimizing it, I leave "untouched" about the last year and a half of the quotes history. It is important not to "tweak" not only during optimization, but also at the stage of development - because adding new "features" like a trailing stop or a filter is already a fitting, though without parameters. Next, when optimization is complete, I run final result on this year and a half. If the parameters suit this last piece - I include the system in my portfolio. If not - to the "dustbin of history". Without any further modifications.

Unfortunately, this method cannot be used to prove the profitability of an Expert Advisor to OTHER people - since no one can check how the tests have been performed. That is where monitoring or investment password from the account for about six months + backtest for a long period (several years) is needed. The backtest shows how the EA is behaving in the long term, and the fragment shown in the real time proves that the backtest 1) is not faked 2) it is from this EA and 3) it is not a fitting (the real time fragment should fit into the long-term trend of the backtest).

That's it, in my opinion that's enough.

 
Igonter:

When developing the algorithm and further optimization I leave "untouched" about the last year and a half of the quotes history. Is it not better to choose another currency pair. For example, all do on the pound/dollar, and tested on the euro/dollar, as in the pictures above (when exhibited were each other. Please help the moderator to correct)

 

Who says that an EA for one pair must work on another? This popular phrase comes from the writings of the "classics" who worked in the stock market. The charts of different currencies are not the same in their "physical sense", and they do not have to behave in the same way.

 

I really liked the "platform" idea

 
Igonter:

Who says that an EA for one pair must work on another? This popular phrase comes from the writings of the "classics" who worked in the stock market. The charts of different currencies are not the same in their "physical sense", and they do not have to behave in the same way.


They should at least not lose much money on similar pairs, or even better - at least to 0. This is my way of testing for the purpose of adjustment. Although, to be fair, not every EA is initially designed to work with another pair, even because of its design features.
 
brOOt:

I really liked the "platform" idea


This conversation has been going on for a long time. If you're interested, check out 'How to publish your code in Code Base' here and 'BUY A COUNTRY COUNTRY' here. And there's a lot more on the forum.