Five ways a backtest lies to you, and how to catch each one

2 September 2026

A backtest is an argument, not evidence. Its job is to find reasons your strategy will fail, cheaply, before capital is involved. Most backtests we are shown are built to do the opposite.

1. Look-ahead bias

The classic: using the day's closing price in a decision made at 10am, or an indicator computed over a window that includes bars the strategy could not have seen. It is easy to introduce accidentally with vectorised code, and it produces equity curves that look too good to question.

2. Survivorship bias

Testing a basket of today's index constituents over ten years quietly excludes every company that was delisted or removed. The universe you test must be the universe as it existed on each date.

3. Costs applied as an afterthought

• Brokerage, exchange charges, STT, stamp duty and GST all apply in Indian markets.
• Funding rates matter on crypto perpetuals and compound quickly.
• Slippage grows with size — a model tested at one lot behaves differently at fifty.

A strategy with a small edge per trade and a high trade count is the most vulnerable. We have seen a 38% annual return become negative once realistic costs were applied.

4. Overfitting to one regime

Parameters tuned on 2021 to 2023 describe that market, not markets in general. Walk-forward testing, out-of-sample holdout periods and parameter sensitivity checks are the minimum. If a small change to a threshold collapses the result, the edge was never there.

If your strategy only works with exactly these parameters on exactly this period, you have not found an edge. You have found a description of the past.

5. Ignoring fills

Assuming you get filled at the signal price, at any size, in any conditions, is the most common unrealistic assumption in retail backtests. Model partial fills, rejections and the gap between signal and acknowledgement — especially around open and close.

The honest process

Backtest, then paper trade on live data for an agreed window, then deploy a fraction of intended capital behind hard risk limits. Every strategy we onboard to Nexa Flux or Nexa Quant goes through all three. Some do not survive the second step, which is exactly the point.