Education
Survivorship Bias in Trading: The Silent Performance Killer
You only see the funds that survived. The strategies that blew up vanish from the data. This single bias distorts everything retail investors believe about returns.
SIGMASEVEN ResearchIn short
Survivorship bias occurs when failed strategies, blown-up accounts, and delisted assets are excluded from historical data. The surviving subset shows artificially inflated performance because the failures are invisible. Every return statistic, every benchmark comparison, and every fund track record you encounter is affected by this bias unless explicitly corrected.
What survivorship bias actually is
Survivorship bias is the logical error of drawing conclusions only from entities that survived a selection process, while ignoring those that did not. In finance, this means analysing only the funds that still exist, only the assets still traded, and only the strategies that did not fail catastrophically.
The result is systematically overstated returns. If you look at "all hedge funds" today, you are looking at the ones that survived. The median fund that existed ten years ago but shut down due to poor performance is excluded from every database you can access.
How it distorts crypto analysis
In cryptocurrency specifically, survivorship bias is extreme. Of the thousands of tokens launched in 2017-2018, only a small fraction still trade at meaningful volume. Any analysis of "crypto returns" that begins with today's top 100 coins and looks backward is studying only the survivors.
The same applies to crypto trading influencers and signal services. The ones you see today are the ones that happened to be right during the latest cycle. Hundreds of equivalent services from 2021 that failed during 2022 have been memory-holed from the collective narrative.
Correcting for survivorship bias in strategy research
The primary correction is inclusion of dead assets, dead strategies, and dead funds in your data. Backtest on the full universe of assets that existed at each point in time, not only those that exist today. This is called "survivorship-bias-free" data and it materially changes conclusions about what strategies actually work.
For crypto strategies, this means testing on all coins that existed during the backtest window - including those that went to zero. Any strategy that only shows results on the coins that survived is implicitly benefiting from hindsight selection. Our systems are validated against the full asset universe available at each historical point in time.
Why this matters for evaluating any provider
When someone shows you a five-year track record, ask: what else did they launch during those five years that is no longer running? A provider with one surviving strategy and four quietly retired failures has a very different hit rate than their marketing suggests.
The honest correction: evaluate all strategies ever launched, not just the ones currently published. Weight track records by the number of attempts, not the results of the best attempt.
Frequently asked questions
- What is survivorship bias in investing?
- Survivorship bias is the tendency to overestimate returns by studying only successful entities (funds, assets, strategies) while ignoring those that failed and disappeared from the data.
- How does survivorship bias affect cryptocurrency data?
- Most crypto indices and analyses only include currently active tokens. The thousands of tokens that went to zero between 2018 and 2023 are excluded, making historical crypto returns appear far higher than they were for the average investor.
- How can I correct for survivorship bias?
- Use survivorship-bias-free datasets that include delisted assets and dead funds. When evaluating providers, ask about all strategies they have launched, not just the ones currently running.
- Does survivorship bias affect index returns?
- Yes. Market indices regularly remove underperforming companies and add outperforming ones. The historical return of the current S&P 500 constituents is meaningfully higher than the actual historical return of the index because composition changed over time.