Quantitative System with 20 Years of Evolutionary Robustness

1. What's a Backtest?

A backtest is a simulation of how a trading strategy would've performed using historical data. It's like "replaying the past" to see if your idea would've worked.

But here's the thing: a backtest is only useful if it's done with scientific rigor. Otherwise, it's just a pretty story — one that'll fall apart in live trading.

2. What's Optimization — And Why You Need It?

Optimization is the process of fine-tuning your strategy's parameters (like stop loss, take profit, moving average periods) to find the best possible combination.

This is essential — without optimization, you're basically trading with random settings. But there's a catch: optimize too much and you'll fall into overfitting.

3. What's Overfitting — The Silent Killer

Overfitting happens when a strategy gets so finely tuned to past data that it basically "memorizes" history instead of learning real market patterns.

The result? It looks amazing in backtests… but crashes and burns in live trading, because the future never perfectly matches the past.

4. The 3 Fatal Mistakes 99% of Traders Make

1. Static optimization on the entire dataset: They optimize parameters using all available data and then test on that same data — it's like taking a test with the answer key.
2. Ignoring out-of-sample testing: A valid backtest must separate training and testing data.
3. Using bad metrics: Relying only on "total profit" or "win rate" leads to terrible decisions.

How Earendil Does It Differently

Earendil was built to eliminate overfitting by design. Instead of a single optimization, it uses:

• ✓ Optimization in short windows (2 years of training)
• ✓ Rigorous out-of-sample testing (2+ years of real validation)
• ✓ Continuous revalidation with expanding windows (up to 20 years of testing)
• ✓ Proprietary metric (Melão Index) that penalizes short strategies and adjusts for inflation
• ✓ Automatic discard of genotypes that lose homogeneity

What Are P5, P50, and P95?

These are percentiles — numbers that divide the 100,000 simulations into equal parts.

• P5 (5th Percentile): The worst result among the best 5%. Your investment has a 95% chance of performing better than this.
• P50 (50th Percentile): The median result — half the simulations do better, half do worse.
• P95 (95th Percentile): The best result among the worst 5%. Your investment has a 5% chance of performing better than this.