Optimise an EA and get the best of the optimised ones. - page 10

 
Serqey Nikitin:

The answer is simple: a sophisticated TS on the right idea gives stable profits from month to month...

This is the main point I disagree with. In my opinion, a complex TS - gives exactly the same stability as the simplest one. And you (you) with your complex system, in my opinion, will have about the same thing as I have with my League of "dumb" systems.

Well... We'll come back to this conversation in a year's time, when you (you) will have... If I'm not mistaken, just under 800 thousand on demo without reinvestment, or 650 million with reinvestment.

 

The EALeague general EA file and the file with the individual optimisation EAs have brief instructions attached.

As a reminder, both files are on Yandex disk.

Besides, a set-file is attached to the file with a set of separate EAs for quick setting of required parameter limits (it is the same for all EAs).

EALeague
EALeague
  • yadi.sk
View and download from Yandex.Disk
 

So, the current situation for the TS in the League:


Charts (without MM, with constant minimum lot):

As before, trading systems of Inverse Trailing Stop are the best. But I am glad that among favorites there are also systems with fixed TP-SL. I personally like them the most.

 
The EA brief on the Yandex CD includes a list of the currently used TS. There are 260 in total.
EALeague
EALeague
  • yadi.sk
View and download from Yandex.Disk
 
George Merts:

Let's keep it on a first-come, first-served basis.

For the paranoid, I can give you the demo account number and the investment password for MT4:

Login: 9968945
Investor Password: TCxG16r9
Server: Alpari-ECN-Demo


It seems that I am one of those paranoiacs. ) Logged in, saved the report. The account has been trading at constant 0.01 lots for more than 7 months, during this time over 13 thousand trades were made. During this time the account was deposited 4 times for $1000 each time. How is it possible to do this on a demo account? But more surprising is the consistency of trading with a steady drain. Or maybe I got something wrong? Attached is the account statement. What is the trick?

State

Files:
287469DE.zip  679 kb
 
Grigori.S.B:

Looks like I'm one of the paranoid ones. ) Logged in, saved the report. My account has been traded with 0.01 lot more than 7 months and made more than 13 thousands trades within that time. During this time the account was deposited 4 times for $1000 each time. How is it possible to do this on a demo account? But more surprising is the consistency of trading with a steady drain. Or maybe I got something wrong? Attached is the account statement. What is the trick?

To clarify.

The account given is not the account of "the best TCs". It is an account in which ALL the TSs work. Regardless of whether they are profitable or unprofitable. All 250 TCs. (And there will be more.) Without MM. Constant minimum lot.

And because there are always more losing TPs - it is clear that the total account is constantly decreasing. And, of course, that I am constantly adding funds there.

I need the account to select deals of those 13 thousand deals of profitable TS. Above I purposely cited a table showing magiks of the best TS and time of their placement on the account. If you want to be sure of the performance of these TS, you should follow only deals with these magicians, and not the entire account.

Moreover, each of the TS has boundary parameters, after exiting from which it stops at once and goes on overoptimization. Its build time, accordingly, changes to a newer one.


When I fully launch the League - most likely areal account signal will be opened by the best TS - "favourites" with MM enabled. But this requires - firstly - running regular re-optimisation of "outsiders" and developing clear characteristics for selecting "favourites".

I gave the account just to make it clear - I'm not trying to deceive anyone. My idea is very clear. There is a "League of Trading Systems", which allows us to shift the question from "What should I invent to make the TS work" to "How to choose the most stable TS among those that work".

 
George Merts:

To clarify...

It's clearer now.

George Merts:

The account given is not that of the "best TCs".

Where is the account of the "best TCs"?

 
Grigori.S.B:

It makes more sense now.

Where's the "best TCs" account?

Erm... I don't get it... It's the same account - it's the best and the worst... Take your pick.

In addition - you can choose TS, by magician - find it in the list, look at the name, and immediately understand the principle on which it operates - you can do the same. I'm not making a secret...

I'm offering to help me optimize them, in exchange for which I'll give you access to any TS you choose. You tell me which one you want, I'll give you a regcod for it.


Or do you want me to choose for you? My top choice is right there. I showed you ten. For confirmation - I gave you an account, look at the magiks, and make sure that the charts show the trades of this particular TS.


Or what else? Do you want me to just open another signal with those TSs, which I think are the best? There will be a signal. When I clearly formulate the selection criteria.

 
George Merts:

Erm... I don't get it... It's the same account - it's all there, the best and the worst... Take your pick.

Moreover - you can choose a TS, by magician - find it in the list, look at the name, and immediately understand the principle on which it works - you may do the same. I am not making a secret...

I'm offering to help me optimise them - and in return I'll give you access to any TC you choose. You tell me which one you want - I'll give you a redcoded one.


Or do you want me to choose for you? My top choice is right there. I showed you ten. For confirmation - I gave you an account, look at the magiks, and make sure that the trades on the given charts are accounted for by this particular TS.


Or what else? Do you want me to just open another signal with those TSs, which I think are the best? There will be a signal. When I have clearly formed the selection criteria.

Interesting movie ... How are you going to trade and earn on the real account? I thought you were carrying out a forward test of the system. There may be self-deception error: you (or me) are choosing the best TS according to current trade results for last 3 months. You have selected it. But what are the reasons to hope that these values are not random, and that immediately after a choice 9 out of 10 selected by you/me TS will not start losing like all the others? After all, purely statistically, the "best" TCs should always be if there are only a few hundred of them. You don't believe in the possibility of randomness of the best TCs and self-deception? So here are the main questions:

  1. Why don't you do a forward test, select the best TS according to your methodology and put them on a separate account of the best TS (at least a demo)?
  2. Do you believe that the results of the best selected strategies may be random? What do you think the probability is? If not, why?

 
Grigori.S.B:

Interesting movie ... how are you going to trade and make money in real time? I thought you were performing a system forward test. There may be self-deception error: you (or me) are choosing the best TS according to current trade results for last 3 months. You have selected it. But what are the reasons to hope that these values are not random, and that immediately after a choice 9 out of 10 selected by you/me TS will not start losing like all the others? After all, purely statistically, the "best" TCs should always be if there are only a few hundred of them. You don't believe in the possibility of randomness of the best TCs and self-deception? So here are the main questions:

  1. Why don't you do a forward test, select the best TS according to your methodology and put them on a separate account of the best TS (at least a demo)?
  2. Do you believe that the results of the best selected strategies may be random? What do you think the probability is? If you do not, why?

1. All TS are subjected to annual optmization with 5-month backtest and 7-month forward test.

That's what I'm suggesting participants help me do !

Take the League, and run it in the strategy tester - you'll see that a year before build time they all perform quite well. Just at the time of back and forward tests. And before that, at an earlier time - I think most of them will stop working (haven't tested it, what difference does it make how they worked before).


2. Indeed, we have no guarantee that the TS we've chosen won't start draining right away. But we don't know the future. We only have back and forward testing, and results on the demo account after installation.

Here we take the best performing TS - EMA FlatRTS on EURCAD, it was set on 14.01.18, it means that from 14.01.17 to 14.06.17 it had a back test, and then from 14.06.17 to 14.01.18 - the forward test, the maximal parameters in terms of back and forward quality were selected (that is, among all the parameter sets, those with the minimum quality between back and forward were selected). On the annual historical period it showed a quality of 0.92.

Now - we can see the results of its demo trading. And we can see that they are even better than before - the integral quality index is 1.4. Quite a good job. The only thing that worries me is Reverse Trailing Stop - a system of reverse trailing, which has a small TP and a large SL. In this case 0.15 and 1.95 daily TR respectively. That is, one SL will wipe out the profits made on 15-20 trades, but... So far, everything is working.

So. I'm absolutely sure that the results of this TS are not random, it's not a "statistical artifact", but a real correspondence to the symbol behavior.

BUT.

This is only for now, at this moment. At any next moment - everything can change. Furthermore, I don't tend to classify myself as a "dabbler of fortune", and therefore I have a fear that any TS only shows good results until I put it on real. As soon as I do so - it will immediately show a "test shot" and stop. But, alas, there is nothing else I can do but replace it with another one that shows good results (also at the moment).