Can an EA last over 10 years?

New comment
 

Hello everyone,

I’ve seen Expert Advisors that have managed to run profitably for years. Even a few martingale strategies have kept going much longer than expected. But sooner or later, most seem to blow the account or simply stop performing the way they once did.

It makes me wonder what it really takes to keep an EA running well over the long term. Whether it’s possible to stick with one approach for years without any adjustments, or if market changes eventually force some kind of update.

I’ve looked into different testing methods, including backtesting in Pine Script, but I keep coming back to MT5 Strategy Tester with real tick data as the most accurate option. It also makes me think about whether there’s any real shortcut to proving long-term viability, or if the only way is to let it run on demo or live and see what happens over time.

 
Valentin Pandarov:

No strategy lasts forever without adjustments because markets change. Backtesting with real ticks in MT5 helps measure robustness, but it is still a simulation.

The real test is to run it on demo or live and set predefined limits, such as a certain drawdown or a drop in average profit, to stop and adapt it when needed.

 
Miguel Angel Vico Alba #:

No strategy lasts forever without adjustments because markets change. Backtesting with real ticks in MT5 helps measure robustness, but it is still a simulation.

The real test is to run it on demo or live and set predefined limits, such as a certain drawdown or a drop in average profit, to stop and adapt it when needed.

True, no strategy can stay untouched forever. Even EAs that look bulletproof in a decade-long backtest can start struggling when the market shifts in ways we didn’t plan for.

What’s interesting is how some traders manage to stretch the life of an EA far beyond the “expected” limit not by changing the core logic, but by adjusting little things around it.

 
any strategy depends on few factors which you see on report of strategy tester such as drawdawn etc. people test on chart for years and keep optimising to find something working for them which don't blow account.

if any of those factors are compromised in future trades then ea won't work anymore as you expect. so ea may work for 1000 years or 1 year depending on how the data of strategy report is performing.
 
Valentin Pandarov #: What’s interesting is how some traders manage to stretch the life of an EA far beyond the “expected” limit not by changing the core logic, but by adjusting little things around it.
In fact this is usually the case unless the strategy and its foundation are truly poor.

Sometimes a small adjustment is enough. It can be like the flap of a butterflys wings in the Amazon as Edward Lorenz described it.

This can happen every six months or a year or even five. We never know where we stand within that mathematical expectancy. There are too many factors involved and we cannot determine it. All we can do is be ready.
 
The landscape keeps changing. Price psychology changes from one year to the next. Political and economical factors also change everything. 
You need to have a news filter implemented in order to have a higher win rate. If you want swing trading to be consistent then you ideally need to have multi timeframe analysis. Scalping is possible to work consistently on single timeframe but you need a statistical edge and use h1 or h2 for scalping with a small take profit or tight trailing stop that activates after 100 - 120 points. The entry price is important, if the bot is entering on noisy areas, or too soon where HFT trading is taking place, it's entering on unwanted volatility. You need to ideally enter on the right momentum and that way you can use a stop loss safely on every trade
 
Valentin Pandarov:
Whether it’s possible to stick with one approach for years without any adjustments...

It really depends on what you mean by "adjustments."

  1. Yes, Martingale strategies are destined to blow up unless one has infinite trading capital and, in that case, why would one be trading at all?
  2. To some extent, any regular indicator with "period" or "length" settings is always adjusting based on ever changing live sample data.
  3. An adaptive indicator alters more of its calculation on the fly which means it's more adjusting than the above regular indicator.
  4. A repainting indicator alters all of its history every time that new live data prints--very adjusting.
  5. Any EA that is based the aforementioned indicators adjusts to the same respective extents.
  6. Neural networks and machine learning can make even more adjustments on the fly.
  7. Finally, the human user can intervene and manually adjust, disable, enable, etc.
There are, indeed, EA's that run on Custom Charts having unchanging "timeframes"--using only #2 and #5, and that test profitably for over 20 years.
New comment