Signs of a REAL system

 

After the excitement in the "Who needs JAPAN CHANGES? I tried to systematize my criteria for assessing the correctness of this or that trading system and the Expert Advisor built on it. Here is a brief summary of my rules that came to my mind.

Signs of fitting:

  • the balance growth curve and the obtained result radically change if you change just a little bit any parameter in the tester
  • the curve of balance growth and the result radically change when the method of checking (all ticks/control points/open prices) is changed in the tester.
  • Stop and Take parameters are not calculated, but strictly specified by specific numbers.
  • Trading signals are calculated using formulas from external programs of neural networks or AI (usually they are kilometre long formulas with ratios of a dozen decimal places, like 1.28677263556*Open[i]).
    joo: This function approximation is one application of neural networks. Applied to use in EAs is a fit.
  • joo: Uneven balance growth. I mean when 60%-80% of profit made 2-5% of transactions
  • Svinozavr: dramatic difference in results on different quote histories (from different brokerage companies).
  • niko1312: difference in results for different history intervals (for the same brokerage company)
  • goldtrader: results on another TF are very different (for the worse) from the baseline,
  • goldtrader: results on another similar (for example GBPUSD after EURUSD) trading instrument are very different (worse) with respect to the base instrument,
  • goldtrader: balance curve outside (both left and right) the optimization area is very different from the view in the optimization area,
  • goldtrader: a large number of parameters to be optimized in an Expert Advisor (more than 2-3).

Logical errors in algorithms:

  • use of redrawing and blinking indicators.
  • Use of indicators and blocks for calculation of trading signals, working with prices and zero bar indicators
  • terrible drawdown figures (over-exposure)
  • Goldtrader: absence of stops. actually this is a duplication of "overfixing", but not everyone will see the overfixing, but the absence of stops is immediately apparent.
  • long stops, the size of which is many times the size of takeovers can be equated to their absence, because this is essentially a camouflaged absence of takeovers.
  • any martingale modifications (no matter how good they are at pulling your profit, the chance to lose everything in just a few operations = 100% - it's just a matter of time)
  • the use of techniques that work for some timeframes for others (the classic candlestick combinations designed for day trading will never work for minute timeframes)
  • All kinds of "flip" strategies (take a losing Expert Advisor and swap buy and sell)
  • ULAD: Using only one source of the signal. There should be a confirmation of the signal and filtration itself.
  • Testing and debugging of the system is made only for one type of market behavior and does not include the others (i.e., you need to test where there is a growth, and a fall, and flat)
  • goldtrader: there is an evident bias towards buy/sell deals in test results (from 30% and higher) with full coverage of all types of market behaviour
  • Low amount of trades on the period under test

    goldtrader: tests with a number of trades below a few hundred are virtually unreliable:

    - up to 100 trades: fthopu adnistratsiya ("unsatisfactory"),

    - up to 300 trades: already something ("satisfactory"),

    - up to 600 trades: good ("good"),

    - over 800 trades: trustworthy ("excellent").

  • Avals: there is no logic tying together all components of the system (i.e. the system is like a haphazard collection of indicators and signals included for example just because everyone says it is a good indicator)

External bans:

  • Pips-slippers will not work - sooner or later the broker will find a way to withdraw your deposit or prohibit the trade referring to some clause of the agreement
  • lots (?)

I would be glad if this list is supplemented by your rules, backed up by practical experience of real trading.

As long as I have access to edit this post I will move all sensible criteria into it.

All that was (from my point of view) sensible - collected in this post. Unfortunately, at about 5-th page substantive communication (as usual) ended and 6th went empty crap .... but since the 15th page they seem to have returned to the topic...

 

I would add uneven balance sheet growth to the signs of fitting. I mean when 60%-80% of profits were generated by 2-5% of trades.

Please explain the words: "trading signals are calculated using formulas from externalneural networks or AI programs (usually they are kilometre long formulas with coefficients with a dozen decimal places like 1.28677263556*Open[i])".

 

Uh-huh. I agree with all the pp. With some on faith (haven't encountered it myself, but logical). I'm thinking of what to add.

You know, this could turn out to be an article. I'll think of a cleverer title... Here: "Apophatic Expert Writing." )))

 

Yes indeed, the Expert Advisor I have given out to practically everyone is just a rough draft of the main idea....

It had both errors in the code and algorithmic errors, after fixing them, the result is MUCH better, the idea makes sense!!! And I think that analyzing the indicators is stupid!!! Why analyze the derivative of price? instead of studying the price chart, and just Japanese candles bear the most information (depending on the TF).

At the expense of "the tester radically changes the growth curve of the balance and the result obtained when you change just a little bit of one parameter"

I say... on our (improved) TS tried to test on all sorts of parameters that differ from each other not by 10-20%, but at times!!!!

The worst result we got was the prof. 1,18 at 1044 transactions since 2008

and the best is already a secret )))

 
joo >> :

Please explain the words: "trading signals are calculated using formulas from external neural network or AI software (usually kilometre-long formulas with coefficients with a dozen decimal places like 1.28677263556*Open[i])"

it was like this

= (1.00002*IF(AND(0.00026974 <= 1*G2,1*G2 < 0.00026974 + 141.828),1/G2,0.121066)*G2*G2*G2-1.25485*IF(AND(0.00026974 <= 1*G2, 1*G2 < 0.00026974 + 141.828),1/G2,0.121066)*IF(AND(0.00026974 <= 1*G2, 1*G2 < 0.00026974 + 141.828),1/G2,0.121066)*G2*G2*G2)/(IF(AND(0.00026974 <= 1*G2, 1*G2 < 0.00026974 + 141.828),1/G2,0.121066)*G2*G2-1.25521*IF(AND(0.00026974 <= 1*G2, 1*G2 < 0.00026974 + 141.828),1/G2,0.121066)*G2+0.000389753-0.00000823394*IF(AND(0.00026974 <= 1*G2, 1*G2 < 0.00026974 + 141.828),1/G2,0.121066)*IF(AND(0.00026974 <= 1*G2, 1*G2 < 0.00026974 + 141.828),1/G2,0.121066))



 
Svinozavr >> :

Uh-huh. I agree with all the pp. With some on faith (haven't encountered it myself, but logical). Thinking of what to add.

You know,this could be an article. Let me think of a cleverer title... Here it is: "Apophatic Expert Writing." )))

Shit... wish I could get into a discussion with you.... But the condition is the same as what you had yesterday....

Apophatic hangover :) Anyway, I'm not fit for a vivid discussion today, sorry....

 
RomanS >> :

Yes indeed, the advisor I've pretty much given out to everyone is just a draft of the main idea....

this thread is not to criticise your work, it's just a thought provoking discussion - nothing personal.


this one's personal: don't go all "i got a little something" on me.

I'll say... on our (improved) TS tried to test on all sorts of parameters that differ from each other not by 10-20%, but at times!!!!

The worst result we got was the prof. 1,18 at 1044 transactions since 2008

and the best is already a secret ))))

You should not read that on a forum where people are looking for help and support. >>) : it's unseemly to read on a forum where people are looking for help and support. if it's a real secret - why whistle about it ;) otherwise it might turn out that you just haven't got all the clues yet.

 
RomanS >> :

Shit... i wish i could get into a discussion with you.... but it's the same condition you were in yesterday....

Apophatic hangover :) Anyway, I'm not fit for a vivid discussion today, sorry....

))) Yeah. I understand. I was disappointed we didn't get a shot back from the moon myself yesterday - it was so hard.

Apophatic means "through negation". That is, the expert is NOT ...

 

Regarding the ForexTools advisor

https://www.youtube.com/watch?v=wKRYFVcPv5E

 
RomanS >> :

Regarding the ForexTools advisor

https://www.youtube.com/watch?v=wKRYFVcPv5E

Zato breeding!!!!

 

Of course, you can giggle all you want. But the only criterion for the correctness of theory is practice. Real life. ;)

You'd better think of something useful to add to your first post...