You're Backtesting Ghosts
You've seen the Reddit post. The TradingView strategy with 47% annual returns. The YouTube guy's backtest showing 200 consecutive wins. The Discord alpha hunter's EA that "broke" the market in 2024. All winners. All published. All posted because they made money—at least in the past.
But you haven't seen the 9,847 strategies that lost money. They don't get tweeted. They don't trend on TradingView. They sit in forgotten hard drives, unremarked and unlauded.
That invisible graveyard is survivorship bias. And it's why 98% of DIY traders who backtest published strategies blow their accounts live.
What Survivorship Bias Actually Is (And Why It Kills Your Edge)
Survivorship bias is a logical error: you only see the winners because losers disappear. If you're studying the best traders, you're studying traders who made it. You're not studying the thousands who tried the same approach and quit. The selection is rigged before you even see it.
In trading, it works like this:
- A trader develops a strategy. It fails for 8 months. They stop trading it.
- Same strategy, different trader, different market condition. It wins for 6 months. They post it everywhere.
- You see the winner. You backtest it. You get 47% returns on historical data. You go live.
- You lose 60% in 4 months because the conditions that made it work have changed.
The strategy wasn't good. It was lucky. And you never saw the evidence of failure.
The Published Strategy Graveyard (What You Don't See)
MQL5 hosts 660+ published Expert Advisors. That's the public list. But how many EA traders built strategies that never got published? 50x that number? 100x? No one knows. Only the ones that worked made it public.
Here's how survivorship bias compounds in backtesting:
- Selective start date: I backtest a strategy from January 2022 onward. It crushed it during a 2-year bull run. But I never test from March 2020 (the crash) to January 2022. If I had, my 47% return becomes 23% because of volatility.
- Curve fitting: I test 50 different indicator combinations on 5 years of historical data. One combination wins 47% annually. That's not a strategy—that's statistical luck. If I tested 100 more combinations, I'd find another "winner." It's just overfitting.
- Optimizing on the data you see: I optimize entry points using the exact prices I know happened historically. Live trading has slippage, spread, and rejections. My perfect entries become imperfect ones. My 47% return drops to 23%, then goes negative.
- Ignoring the failures: If I tested this strategy during March 2020, May 2022, or September 2023 volatility spikes, it would have blown up. But I didn't test those periods. So I never saw the failure. That's survivorship bias working in real time.
Survivorship bias isn't a mistake you make. It's a trap the data sets up for you.
Why Paper Trading Still Kills Your "Good" Strategy
Here's the thing: backtesting is a lie. A useful lie, but a lie.
When you backtest, you're assuming:
- You get filled at the exact price you wanted. (Live: you don't. Slippage costs 5-50 pips per trade.)
- You can exit whenever you want. (Live: the market moves, orders gap, your broker re-quotes.)
- You follow the rules perfectly. (Live: you add to winners, exit early, hold losers longer when emotions kick in.)
- Transaction costs don't exist. (Live: spreads, commissions, and swaps eat 10-30% of your profits.)
Your backtest shows 47% annual returns. You paper trade it. Returns drop to 23% because slippage and spread are real. Then you go live with real money. Emotional trading kicks in—you second-guess entries, hold winners too long, exit losers too early. Returns crash to -8%.
That's the graveyard most traders fall into. Their strategy wasn't bad. Their test was fictional.
The Math That Proves You Can't Backtest Your Way Out
Let me be direct: if you're testing strategies from social media, you're testing the survivors of a selection bias that's already eliminated 99% of the failures.
Here's how it works:
You find 10 published strategies on TradingView. You backtest all 10 on 5 years of data. One shows 47% annual return. You think you found gold.
But whoever published that strategy tested it against 100 different start dates, or 100 different indicator combinations, or 100 different market conditions. One of those 100 iterations won big. The other 99 lost or broke even. They published the winner. You're seeing 1 out of 100.
This is why professionals don't trade published strategies. They know that if 1 out of every 100 random backtests wins, you can't know which 1 you're looking at without seeing all 100 failures. That's the graveyard working against you.
Why Custom Beats Published (Every Single Time)
Professional traders don't backtest strategies from Reddit. They build custom strategies for their account size, risk tolerance, and the specific market they trade.
A custom strategy is different because:
- It's built for YOUR account: A strategy that works on a $10k account breaks on a $100k account due to slippage and market impact. Professionals build it for your exact account size.
- It's tested on out-of-sample data: Instead of optimizing on all available data (curve fitting), they optimize on 60% and test on the remaining 40% they never touched. This catches overfitting instantly.
- It accounts for real costs: Spread, slippage, commissions, swap fees. Real costs for your specific broker, not hypothetical ones.
- It's iterative: After live deployment, it's refined based on real results, not backtest results. Because reality and backtest are never identical.
This is why traders hire Alorny. Not to learn backtesting. But to skip the graveyard entirely.
The Cost of Getting This Wrong
Survivorship bias isn't a knowledge problem. It's a data problem. You can't fix it by reading another blog about backtesting.
If you test a published strategy and go live, you're gambling that:
- The person who published it tested correctly (they didn't—they cherry-picked the data)
- The conditions that made it work haven't changed (they have—markets evolve constantly)
- Slippage and spread won't destroy returns (they will—they always do)
- You'll execute perfectly under stress (you won't—no one does)
The average trader who backtests published strategies loses $2,400-$4,800 within 12 months. That's the cost of survivorship bias.
A custom EA from Alorny runs $300-$500. Cheaper than one margin call on a strategy that was never tested for your account size or broker conditions.
The Way Out: Stop Testing Published. Start Building Custom.
You can't backtesting your way out of survivorship bias. The data is rigged before you ever see it. But you can avoid it entirely by going custom.
A custom EA built specifically for your strategy, account size, and risk tolerance gets you:
- A working demo in 45 minutes so you see exactly how it'll perform before you deploy
- Full backtest report with realistic assumptions—spread, slippage, your actual broker conditions, not fantasy numbers
- Revisions until it aligns with your expectations and strategy, not until it hits some arbitrary performance target
- Deployment on MT4, MT5, TradingView, cTrader, or any platform you use
- Full project delivery in hours, not weeks
Most traders spend their first year backtesting 50 published strategies. Professional traders spend their first week hiring someone to build 1 custom strategy and spend the year refining it on real data.
You can backtest published strategies forever and fail forever. Or you can spend $300 once and stop testing ghosts.