Survivorship bias quietly inflates returns and understates risk. In an Indian context, think of backtesting a strategy on Nifty 50 constituents by downloading the current list and using that as historical input. That ignores firms that were part of the index earlier but got delisted, merged, or bankrupted. The same problem happens with mutual fund data if closed or merged schemes are missing from the dataset. The result is a strategy that appears robust on paper but fails in live markets.
Why it matters
Real market history includes losers, corporate actions, and closures. When you remove those events, two bad things happen:
- Returns look higher because only surviving winners remain.
- Drawdowns and volatility look smaller because the worst outcomes are removed.
Both give false confidence. For retail and institutional traders in India, this can mean allocating capital to strategies that would not have survived actual market stress.
Common ways survivorship bias appears
Many sources of historical data are culprits:
- Downloading current symbols list from an exchange and assuming they existed forever.
- Using candle data that omits days or symbols around delisting.
- Backtesting mutual fund performance using only funds that currently exist.
- Relying on vendor databases that do not keep records of dead tickers.
Signs your backtest may be biased
- Extremely high compounded returns with low drawdown compared to market history.
- Strategy performs too consistently across long periods without big failures.
- Historical trades include only familiar, large-cap names and never obscure or delisted stocks.
How to fix and run realistic simulations
Start with the right data. Use survivorship-free databases that preserve historical symbols and their status at each date. For Indian equities, prefer exchange-provided historical listings, or vendors that explicitly state they maintain delisted ticker records. For mutual funds, use NAV histories that include merged or wound-up schemes.
Account for corporate actions correctly. Splits, bonuses, rights issues, mergers and delistings change share counts and prices. Ensure your price series are adjusted consistently and that delisting events appear as real losses or zero value if warranted. Ignoring these leads to wrong returns.
Simulate real-time decision-making. A common mistake is to use end-of-day data for signals but then trade at the close with perfect fill. Use realistic order fills, intraday slippage, and partial fills when liquidity is low. In India, many small-cap stocks have thin volumes—assume wider spreads and higher slippage for them.
Include costs. Brokerage, taxes, stamp duty, and impact cost matter. Model brokerage per trade, Securities Transaction Tax (STT), Goods and Services Tax (GST) on brokerage, and stamp duty. These can reduce performance significantly for frequent strategies.
Rebalance and survivability rules. If your strategy periodically rebalances using ranks or screens, simulate the list as it was at that rebalancing date. Do not peek at future constituents. If a stock gets delisted right after purchase, the backtest must capture that loss or recovery, not ignore it.
Practical checklist before trusting results
- Use survivorship-free datasets for equities and funds.
- Adjust prices fully for corporate actions, and keep delisting events.
- Simulate order execution with slippage, liquidity limits, brokerage, STT, and GST.
- Avoid look-ahead by reconstructing the investable universe as of each decision date.
- Test sensitivity: remove the top X% of winners to see how performance changes.
Final tip
Start small and validate. Run your strategy on a period or segment you can verify manually—look up a few random trades and confirm the historical status of those names. If your simulated trades match real historical events, your framework is more likely survivorship-free.
If you remember one thing: good backtests mirror messy history, not a cleaned-up highlight reel.