Backtesting that doesn't lie

A beautiful equity curve is the easiest thing in the world to manufacture. Give a backtester enough freedom over fills, enough parameters to tune, and enough hindsight, and it will happily draw you a line that goes up and to the right forever. Then it loses money the first day it meets a real order book. The job of an honest backtest is not to flatter your strategy — it's to disappoint you now instead of later.

Simulated fills, not fantasy fills

Algod walks real historical NSE candles one bar at a time and simulates fills at prices you could plausibly have gotten. It won't let you buy at the exact bottom wick of a candle just because the data touched that price for a millisecond. Entries and exits are filled on the terms the bar actually offered, so the trade log you read is a trade log you could have lived through — not a highlight reel assembled with perfect hindsight.

Walk-forward on real candles

The classic way to fool yourself is to optimize a strategy on the same history you then test it on. Of course it looks brilliant — you fit the answer to the exam. Walk-forward testing breaks that loop: you tune on one slice of history, then evaluate on a later slice the strategy has never seen, and repeat as you roll forward through time. A strategy that survives out-of-sample, across bull runs, chop, and drawdowns, is telling you something real. One that only shines in-sample is telling you it memorized the past.

Metrics that matter

Total return is the number everyone quotes and the number that matters least on its own. Algod streams the metrics that actually describe whether you can live with a strategy:

• Maximum drawdown — the deepest peak-to-trough loss, and the real test of whether you'd hold on.
• Win rate paired with average win/loss — a 40% win rate can beat an 80% one if the winners are bigger.
• Sharpe ratio — return measured against how much it lurched to get there.
• Trade count — thirty trades is an anecdote; hundreds across regimes is closer to evidence.

If a backtest has a low drawdown, a high win rate, and only nineteen trades, you haven't found an edge — you've found a coincidence.

The deadliest trap is overfitting: adding one more condition until the curve is perfect. Every extra rule that only exists to dodge a specific past loss is a rule that will misfire on a future you can't see. Keep strategies simple, judge them out-of-sample, and treat a suspiciously smooth curve as a warning, not a trophy. And remember the honest part — backtested performance never guarantees future results, and no simulation removes the risk of trading real money.