EA Performs Well in Backtest but Fails on Recent Data – Any Ideas?

New comment
 

Hi everyone,

Over the past few months, I’ve developed and tested more than 50 trading bots (EAs), each with different variations and strategies. From my backtesting results, a handful of them have shown strong performance.

Recently, I started running backtests using IC Markets real tick data, which is supposed to be more accurate. Initially, everything looks great — the equity curves are smooth, steadily rising, with controlled drawdowns and proper risk management built into the EA.

However, I’ve noticed a consistent issue: as soon as the backtest reaches around April 1st (up to current date), performance drops significantly. The equity curve, which was previously trending upward, starts declining.

I’ve tried multiple approaches to figure this out:

  • Reviewed and adjusted the code (even used AI tools to assist)
  • Tested different parameter optimizations
  • Tried reversing the entry logic to see if the strategy could benefit from the downturn

Despite all this, the issue persists — there’s a very noticeable “switch” in performance.

I also experienced something similar earlier this year. I had an EA that performed very well throughout 2025, but as soon as January came, it suddenly stopped working, and the equity curve turned downward.

At this point, I’m trying to understand what’s really causing this:

  • Could it be an issue with the data (even though it’s real tick data)?
  • Is this just a case of changing market conditions/regime shifts?
  • Or does this mean the strategy simply isn’t robust enough?

For context, I’ve also started forward testing on a small live account, but it’s still too early to draw conclusions. So far, results align with backtesting, but again, limited sample size.

I’d really appreciate input from more experienced traders or developers who have encountered similar issues.

I will share all my ea with anyone willing to assist :) 

Thanks in advance!

 

In my experience, U.S. Thanksgiving week through the end of January is a bad time to trade likely due to institutional traders going on vacations.

Thus far, April has been a mess too. If you ask me what happened, I'll say U.S. Government "policy."

 
Ryan L Johnson #:

In my experience, U.S. Thanksgiving week through the end of January is a bad time to trade likely due to institutional traders going on vacations.

Thus far, April has been a mess too. If you ask me what happened, I'll say U.S. Government "policy."

Yea i thought so as well since the asset im testing my bots on is gold. So far, i've been able to  get a positive equity curve on the first 3 months of this year after vigorous testing but now April just started and immediately from April first the curve immediately dipped. Now the only logical thing i can think off is that maybe IC market doesn't have real tick data for the month as yet. But i'll keep testing and hopefully expect good results as i've seen in the previous months and years on backtester. 
 
may be spread /slippage issue, or your developer if you developed  by other than mql5 require licence key for live market
 
salemtrading5 #:
Yea i thought so as well since the asset im testing my bots on is gold. So far, i've been able to  get a positive equity curve on the first 3 months of this year after vigorous testing but now April just started and immediately from April first the curve immediately dipped. Now the only logical thing i can think off is that maybe IC market doesn't have real tick data for the month as yet. But i'll keep testing and hopefully expect good results as i've seen in the previous months and years on backtester. 
IMHO, the situation that performance of previously optimized EA is dropped sometimes (abruptly or gradually) is quite normal. It's not normal if you did not see this effect before with any of your 50 bots. The market is changing.
New comment