Why do many EAs fail while the same strategy works perfectly in manual trading?

New comment
 
Why do many EAs fail while the same strategy works perfectly in manual trading?
Manual trading lets us adapt, filter trades, and read the market in real time. EAs, on the other hand, follow strict rules — they can’t “think” like a human. Automation often reveals the gaps in a strategy rather than fixing it.

Have you ever automated a strategy and been surprised by the results?
 

I have thought about this many times.

From what I’ve concluded, it has to do with how the human brain works. It constantly learns (consciously or unconsciously) from any experience, ideally to improve. This directly correlates with trading.

Imagine giving the same set of instructions to different people—each individual will execute them in a unique way. Meanwhile, a bot will always perform them the same way.

 

There are 2 sides to every story─a coin's head versus tail, a religion's god versus devil, a trader's brain versus bot, etc.

Arguably, just as many or even more human brains fail versus bots. Keep in mind that it is the human who developed the logic of a strategy, even if the human is cavalier enough to use AI to code it. Therefore, the GIGO principle applies─garbage-in, garbage-out. At the end of the day, a trader either has a system of rules that yields profits or does not. That is to say that there is a scientific way to trade versus an anecdotal way to trade, regardless of whether the strategy is automated.

If the system is anecdotal, then it can't be empirically tested. A professional FX trader once said that you should be able to explain your system to a hermit that doesn't speak your language with nothing more than a chart and a pointer. Obviously, this was an exaggeration to demonstrate a concept, but the idea is that the rules of the system must be well defined. If a trader is unable to follow her/his own rules, then those rules are merely soft suggestions. This creates space for indecision, hesitation, delayed response, and emotional fatigue to intervene.

Also keep in mind that a bot never gets hungry, gets overcaffeinated, oversleeps, under-sleeps, sleeps at all (ok ok... Sleep()'s maybe), gets hungover, argues with its life partner, has diarrhea, gets sick, etc. All of those physiological/psychological aspects of being human have an effect on everything, including trading decisions.

More than likely, the folks that are unable to profitably implement their manual system as a bot are just that─unable to codify what it is, exactly, that they're doing manually. You don't necessarily have to be an expert programmer, but you do need to be able to identify the plain language logic in your system, research the code elements required to implement that logic, and understand how those elements fit together.

 
Ryan L Johnson #:

There are 2 sides to every story─a coin's head versus tail, a religion's god versus devil, a trader's brain versus bot, etc.

Arguably, just as many or even more human brains fail versus bots. Keep in mind that it is the human who developed the logic of a strategy, even if the human is cavalier enough to use AI to code it. Therefore, the GIGO principle applies─garbage-in, garbage-out. At the end of the day, a trader either has a system of rules that yields profits or does not. That is to say that there is a scientific way to trade versus an anecdotal way to trade, regardless of whether the strategy is automated.

If the system is anecdotal, then it can't be empirically tested. A professional FX trader once said that you should be able to explain your system to a hermit that doesn't speak your language with nothing more than a chart and a pointer. Obviously, this was an exaggeration to demonstrate a concept, but the idea is that the rules of the system must be well defined. If a trader is unable to follow her/his own rules, then those rules are merely soft suggestions. This creates space for indecision, hesitation, delayed response, and emotional fatigue to intervene.

Also keep in mind that a bot never gets hungry, gets overcaffeinated, oversleeps, under-sleeps, sleeps at all (ok ok... Sleep()'s maybe), gets hungover, argues with its life partner, has diarrhea, gets sick, etc. All of those physiological/psychological aspects of being human have an effect on everything, including trading decisions.

More than likely, the folks that are unable to profitably implement their manual system as a bot are just that─unable to codify what it is, exactly, that they're doing manually. You don't necessarily have to be an expert programmer, but you do need to be able to identify the plain language logic in your system, research the code elements required to implement that logic, and understand how those elements fit together.

totally agree with you.
Also what if you can add Machine learning to the process so the bot can learn from it's mistakes adjust and tweak the strategy a little bit so it can fit the market continues changes, that's what I am trying to achieve through my project this can simulate a human brain as much as possible.
 
It is because of dynamics in my opinion. 
 

 

RENATA: "Manual trading lets us adapt, filter trades, and read the market in real time. EAs, on the other hand, follow strict rules — they can’t “think” like a human. Automation often reveals the gaps in a strategy rather than fixing it."

  


  Indicators adapt to price as each bar forms.  So adaptation as a filter doesn't necessarily work in either case manual or robotic.

  Logic doesn't lie, it doesn't start creating new exceptions to trading system rules or adding extra indicators to satisfy any belief.

  Statistical methods are the only way to test 1000's of iterations of logic to know if you are wasting your time looking at a screen.

  Optimization can be avoided too if the case for the signals is good enough.

  

  Enjoy your day everyone.

 
Most of the times if a strategy is proven to work manually, it's either not back-tested with enough trades or the algorithmic approach isn't applied correctly 1:1. Either way you need to test on the same data, sometimes the data might be wrong and you're ending up with poor EA performance even if the strategy works 
 
RENATA PALIOKIENE:
Why do many EAs fail while the same strategy works perfectly in manual trading?
Manual trading lets us adapt, filter trades, and read the market in real time. EAs, on the other hand, follow strict rules — they can’t “think” like a human. Automation often reveals the gaps in a strategy rather than fixing it.

Have you ever automated a strategy and been surprised by the results?


I could simply 'adapt' my EA/Algo though...?


The only surprise I got from automating a strategy is to see how much time I manage to free up because the algo saves me time, removes emotion and enforces discipline. 

 
I think it may depend on the strategy as well. Some bots could more effectively execute a swing based strategy, but human beings could more effectively execute scalping strategies
 
RENATA PALIOKIENE:
Why do many EAs fail while the same strategy works perfectly in manual trading?
Manual trading lets us adapt, filter trades, and read the market in real time. EAs, on the other hand, follow strict rules — they can’t “think” like a human. Automation often reveals the gaps in a strategy rather than fixing it.

Have you ever automated a strategy and been surprised by the results?
Might be a sign that the strategy is just a ticking 💣. In the long run it'll have a net negative.
 
RENATA PALIOKIENE:
Manual trading lets us adapt, filter trades, and read the market in real time.
This is discretionary trading added to a systematic approach. Often this addition will not help in the long run.
12
New comment