Case study · Data & market research
Edge Verify: building an honest backtester
Edge Verify is a backtesting and paper-trading platform for Indian markets that I build solo. Its job is to stop the accidental cheating that makes most backtests look brilliant: it tests strategies on 8 years of 5-minute NSE data across 428 symbols, with a realistic cost model, no look-ahead execution and deliberately harsh stress tests.

At a glance
- What
- Backtesting and paper-trading platform for Indian stocks, indices and crypto
- My role
- Founder and solo builder
- Data
- 8 years of 5-minute NSE equity data (2018–2026, 428 symbols), plus indices and crypto history
- Costs modelled
- Brokerage, STT, exchange charges, GST and liquidity-tier slippage
- Status
- Beta — backtesting live, paper trading live, capital-timing research in development
The problem
Most backtests lie by accident
A backtest that lies is worse than none. The usual lies are small and invisible: filling a trade on the same candle that generated the signal, ignoring brokerage and taxes, testing only on today’s winners, or trusting a result built on a handful of trades. Each one inflates the result; together they turn a losing idea into a convincing one.
Architecture
From raw data to a verdict
- 01
Market data
8 years of 5-minute NSE equity data across 428 symbols, plus index and crypto history, so ideas meet bull, bear and sideways markets.
- 02
Strategies
Opening-range breakout, EMA, VWAP, volume and SuperTrend — plus a no-code builder for custom rules.
- 03
Execution guards
No same-candle or look-ahead execution, and position size from the true entry-to-stop distance, not the range width.
- 04
Cost model
Brokerage, STT, exchange charges, GST and liquidity-tier slippage in every result — net, never gross.
- 05
Stress tests
Re-run the same trades with 2× and 3× slippage, late and worst-case fills, Monte-Carlo percentiles and a bull / bear / sideways split.
- 06
Verdict
How much of the edge survives, with warnings for survivorship bias and for samples under 30 trades.
Decisions
Decisions and trade-offs
Refuse to flatter
The engine warns when a test uses today’s universe of stocks (survivorship bias) and when a sample has fewer than 30 trades. A warning makes a result look worse — which is the point: if a feature would make results look better than they are, it doesn’t ship.
Costs in every number
Brokerage, STT, exchange charges, GST and slippage by liquidity tier are applied to every trade, so every figure is net. Gross numbers are where most backtests hide their problems.
Stress before trust
The same trades are re-run under harsher fills — double and triple slippage, late and worst-case fills — with Monte-Carlo percentiles and a regime split. A strategy that only works with perfect fills isn’t a strategy.
Paper trading as the second proof
History proves the idea; live paper trading proves it can be executed. Orders run against live NSE prices with stop-loss and target, automatic exits, and discipline limits such as daily trade caps and a maximum number of open positions.
Principle
What it deliberately doesn’t do
No signals, no tips, no predictions. Edge Verify measures whether an idea survives reality; it doesn’t tell anyone what to buy.
This case study describes software engineering and research method. It is not investment advice or a guarantee of future performance.
Written by
Contact
Need a data-heavy product built properly?
Tell me what you want to build, who it's for, and your timeline — projects are delivered by the Rana Brothers team.