Why a verdict can still be wrong.
We believe understanding the limits of a backtest is just as important as understanding its results.
Most trading platforms focus on what a backtest can show. We also explain what it can't. A verification is the strongest evidence we can hand you before you risk a rupee — real NIFTY and SENSEX data, real option prices, every Indian cost counted, stress-tested across thousands of paths. But evidence from the past is not a promise about the future, and we'd rather you know exactly where a verdict can break than discover it with live capital. Here's where, honestly.
1. Market regimes change
A strategy that thrived in a trending market can bleed in a choppy one. Your verdict reflects the regimes that happened to fall inside the test window — not the one you'll trade into next week.
2. Liquidity isn't guaranteed
We model fills on historical prices, but on a real bad day the spread widens, depth thins, and the strike you wanted isn't there at the price you saw. Thinly-traded strikes are exactly where a backtest flatters and a live account suffers.
3. Our data coverage has limits
Every verdict carries its coverage — how many sessions, which instruments, what's missing. A strategy verified on 248 sessions of NIFTY weeklies has said nothing about monthlies, about SENSEX, or about the months we didn't hold data for. Read the coverage; it's there for a reason.
4. Corporate actions reshape the past
Splits, dividends and adjustments mean historical prices aren't always the prices a trader actually saw. We handle the common cases, but reconstructed history is never a perfect photograph.
5. Black swans aren't in the sample
A surprise RBI move, a circuit breaker, a 2020-style crash — the rare, violent days that decide whether a strategy survives are, by definition, under-represented in any finite history. A spotless backtest has often just been lucky enough to miss one.
6. The future is not the past
This is the one that contains all the others. Backtesting answers “would this have worked?” It can never answer “will this work?” Anyone who tells you otherwise is selling something.
Questions to ask every backtest
The habit that separates careful traders from hopeful ones: before you trust any result — ours or anyone else's — make it answer these.
- How much historical data was tested — months, or years?
- Which market conditions were included — trending, choppy, high-volatility?
- Were realistic trading costs included — brokerage, STT, slippage?
- How many trades generated this result — enough to be more than luck?
- Would I trust this with my own money — honestly?
Every Algoshastra verdict is built to answer the first four out loud. The last one is always yours.
So why verify at all?
Because the alternative is trading on a hunch. A verification won't make you bulletproof — it filters out the ideas that never even worked on paper, and it tells you how confident the evidence is. That's what the Trust Engine is for: coverage, data confidence, walk-forward and out-of-sample tests, Monte-Carlo paths, and the Deflated Sharpe and PBO that catch strategies which only “worked” because we tried many. We quantify the uncertainty instead of hiding it. The full methodology walks through every gate.
What we ask of you
Read the limitations on every verdict. Size every position for the worst case, not the verdict's average. Review the exported code before it ever touches your broker. And treat a strong verdict as what it is — good evidence, not a guarantee.
A good backtest doesn't predict the future. It helps you make better decisions by understanding how an idea behaved in the past. That's exactly what Algoshastra is built to give you — honest evidence, openly bounded.