How Much Does Only Keeping Surviving Coins Overstate Returns?
"Keeping only surviving coins" inflates return statistics by more than 50%, depending on your asset allocation method. Academic research quantifies this bias: equal-weight portfolio annualized returns are overstated by 62.19%, and the small-cap factor premium is overstated by 50.3%.
A leading global cryptocurrency platform,suitable for both beginners and experienced traders.
New user benefit: 20% off trading fees upon registration!！
In the real market, nearly 40% of crypto assets eventually delist or go to zero. If a backtest only includes coins that survived, it's like "knowing the answer before solving the problem."
1. First, verify how your backtest sample is constructed
Different sample construction methods produce completely different bias magnitudes.
Case A: Equal-weight portfolio If you assume "invest an equal amount in each coin," the bias from keeping only survivors is the largest. In data from 2014 to 2021, the annualized return of an equal-weight portfolio including all coins was about 82.94%, while the portfolio with only surviving coins returned about 145.12%—an extra 62.19% fabricated from nothing. This happens because many small coins in the equal-weight portfolio are dragged down by those that go to zero; dropping them naturally makes returns jump dramatically.
Case B: Market-cap weighted portfolio If you allocate capital by market cap (BTC dominates), the bias is much smaller. The all-coins market-cap-weighted portfolio had an annualized return of about 72.9%, while the survivors-only portfolio returned about 73.8%, a bias of just 0.93%. In a market-cap-weighted portfolio, top coins like BTC and ETH carry most of the weight, so delisted small coins have minimal impact.
Case C: Small-cap strategy If you specifically target small-cap coins, the bias is extreme. After removing coins that went to zero, the small-cap strategy's weekly return is overstated by 0.95%, which annualizes to an overstatement of 50% or more. This is one core reason why the "small-coin premium" looks so strong in backtests but is so hard to realize live.
Key data: Between 2014 and 2021, a total of 3,904 crypto assets were traded, of which 1,222 were delisted—a rate of 31.3%. Among the delisted coins, 76% of investors suffered a total loss.
2. Check your backtest for hidden "survivor filtering"
Many backtesting tools default to keeping only coins with "complete historical data"—that is hidden filtering.
Common hidden filtering methods:
Setting a later start date, which automatically excludes coins that had already gone to zero earlier
Requiring "the coin must have data throughout the entire backtest period" (i.e., it must have survived to the last day)
Only selecting coins that are "currently ranked in the top XX by market cap"
Using data sources that only provide historical data for coins that still exist
Completion criterion: You have confirmed that your backtest coin list includes delisted coins. If you cannot obtain data for dead coins, your backtest results carry an inherent bias by default.
3. Assess the actual impact of the bias on your strategy
This is not a "minor issue"; it can directly affect your capital decisions.
| Strategy type | Survivorship bias impact | Real trading consequence |
|---|---|---|
| Equal-weight buy and hold | Annualized overstatement 62%+ | Live returns far below backtest |
| Small-cap screening strategy | Annualized overstatement 50%+ | Small-coin premium nearly disappears live |
| Grid/mean reversion strategy | Moderate bias | Gaps from zeroed coins ignored |
| Market-cap-weighted large-cap strategy | Minimal bias (<1%) | Bias acceptable |
But note: Although top-asset portfolios suffer less from survivorship bias, such backtests can still be dominated by specific period volatility; don't rely on a single metric to judge a strategy's robustness.
A leading global cryptocurrency platform,suitable for both beginners and experienced traders.
New user benefit: 20% off trading fees upon registration!！
4. Fix: Obtain data that includes delisted coins
To truly validate a strategy, you need historical data that includes "dead" coins.
How to do it: Some academic databases and research platforms offer survivorship-bias-corrected datasets (such as certain cryptocurrency academic datasets), which are harder to obtain than regular market data services.
Alternative approach: Use a "rolling window simulation"—assume a certain percentage of coins go to zero each year (referencing the historical ~31% delisting rate). Randomly drop that proportion of coins in the backtest and record a -100% return for them, then see whether the strategy still shows a profit.
Completion criterion: Your backtest includes some form of zero-coin loss simulation, rather than assuming all coins "survive to the end" by default.
How to confirm you've properly accomplished this assessment:
The next time you see a strategy claiming "annualized 50%+ returns," first ask: "Does the sample include coins that went to zero?" If the answer is no, apply a 40% haircut to the returns for an equal-weight strategy—if the haircut result still outperforms the market, then it's a signal truly worth live testing. Also, look at the strategy's survivor count: of the 511 coins in 2014, only 127 remained by 2021—a mortality rate as high as 75%. Any strategy that ignores those 75% of failures is replacing a "full battle report" with a "photo of the survivors."
