What You Cannot See
In World War II, the US military analyzed bullet hole patterns on returning bombers to decide where to add armor. They initially wanted to reinforce the areas with the most bullet holes. Abraham Wald, a statistician, pointed out the error: the holes showed where planes could be hit and still make it back. The planes that were hit in other areas did not return. They needed to armor the places with no bullet holes.
This is survivorship bias, and it operates in financial markets constantly. You are always looking at the returning planes.
Survivorship Bias in Crypto
The crypto market is one of the most survivorship-biased environments in finance. When you look at Bitcoin's all-time chart and see a 60,000%+ return, you are looking at the single biggest winner out of thousands of cryptocurrencies launched since 2009. The vast majority of tokens launched in 2017 and 2018 went to zero. They do not appear on CoinMarketCap. They do not appear in your backtest data. They are invisible.
If you backtest a "buy the top 50 crypto tokens" strategy, your backtest is biased because the current top 50 are the survivors. The tokens that were in the top 50 three years ago but fell out (and possibly went to zero) are not in your data. Your backtest shows better performance than the strategy would have actually delivered because it only includes things that survived.
Survivorship Bias in Fund Performance
The same bias afflicts fund performance data. Hedge funds that perform poorly shut down and disappear from databases. The average performance of remaining funds improves not because funds got better, but because the worst performers stopped being counted. Studies estimate this effect inflates reported hedge fund returns by 1-3% annually.
Survivorship Bias in Trading Signals
When someone shows you a chart of "every time this pattern formed, the price went up 20%," ask how many times the pattern formed in assets that subsequently went to zero and are no longer in the dataset. The pattern might have appeared 100 times, with 60 positive outcomes in surviving assets and 40 instances in assets that no longer exist. The true success rate is 60%, not the 100% suggested by only looking at survivors.
How to Protect Yourself
Point-in-time data is the gold standard. A proper backtest uses the universe of assets that was available at each historical moment, including those that subsequently failed. This is harder to source and process, but without it, your results are systematically overstated.
When evaluating any strategy, system, or historical analysis, ask: what am I not seeing? What assets, funds, or traders were part of this universe at the time but are no longer here? What would the results look like if I included the full universe rather than just the survivors? This question does not always have a precise answer, but asking it protects you from the most common form of analytical self-deception.