Reading our 3-year backtest like a skeptic

The trap of beautiful backtests

Any strategy can look brilliant if you fit its parameters to the past. The honest test is out-of-sample: fix the rules first, then see how they survive data they've never seen.

How ours was run

• Parameters from decades-old trend-following literature — fixed before the test window
• Three full years of real hourly market data across 15 liquid assets
• Realistic execution costs charged on every rebalance
• The exact same portfolio code that runs live — not a research copy

What it showed

Profitable in every calendar year of the window — including a full bear market — with a Sharpe ratio near 0.9 and roughly half the drawdown of buy-and-hold.

Past performance still doesn't guarantee future results. But a test you can't quietly re-fit is the difference between evidence and marketing.

Verify it yourself

The performance page charts the actual curve, and the strategy detail page in the portal shows the parameters — unchanged from the test.