How to Spot a Curve-Fit Backtest Before You Buy an EA: 5 Checks From a Year of Real-Tick Testing
9 July 2026, 04:50
0
19
We build EAs and, before we publish anything, we run it on real broker tick data across multiple feeds and across full market cycles. Doing that for a year taught us an uncomfortable lesson: most beautiful backtests you see in marketplaces are artifacts, not edges. Here is a practical checklist you can use before buying any EA — ours included.
1. Balance drawdown is not risk. Equity drawdown is.
Grid and averaging systems can show a 5% balance drawdown while the floating loss on open baskets reaches 50% or more. The balance curve only records closed trades. Always ask for the equity drawdown. If the seller only quotes balance DD on a grid system, assume the real number is several times larger.
2. Check the test window, not just the numbers.
A strategy tested only on 2023-2026 gold has never met a real crash. When we re-ran popular-looking configurations over windows that include 2020 and 2022, results that looked like PF 3+ with tiny DD often turned into blown accounts. If a demo or report is locked to a recent date range, that is not an accident.
3. Win rate above 90% is a warning, not a feature.
The signature of martingale and deep averaging is many small wins and rare catastrophic losses: average loss 4-10x the average win. The account survives until it doesn't. Look at average win vs average loss, not the win percentage.
4. One data feed is not verification.
We have watched a strategy print PF 2.2 on one feed and lose money on another broker's real ticks, same dates, same settings. Anything that only works on one feed is fitting noise, not trading a market. Test on at least two brokers' data before trusting it.
5. Hard stop-loss on every trade, or walk away.
"Smart recovery", "zone hedging" and "no-SL management" are different names for holding losers and hoping. A real edge does not need to hide its exits.
What this means for our own products
We publish losing years and worst drawdowns in our own listings, because a buyer who knows the real risk keeps the product longer than one who was promised magic. Our trend systems are verified on real ticks across multiple brokers and full cycles — the numbers are smaller than the marketplace headliners, and they are real.
Examples of how we present results (including the ugly parts):
MEGAMAX Donchian Trend (JPY pairs, hard SL): https://www.mql5.com/en/market/product/181838
Bitcoin Comet Trend Follower (BTC/ETH, hard SL): https://www.mql5.com/en/market/product/181831
All products: https://www.mql5.com/en/users/app.develop.sk/seller
If you take one thing from this post: ask every seller for equity DD, a 2020-2022 window, and a second broker's feed. The honest ones will show you.
