Do futures backtests suffer from survivorship bias?

Less at the instrument level for major index contracts, because a continuous series of ES or NQ exists throughout. The bias remains in strategy selection and in the choice of which markets to study, often made because they performed well recently.

How is survivorship bias different from lookahead bias?

Lookahead bias uses information that was not available at decision time inside a test. Survivorship bias selects the sample itself using an outcome known only afterwards, such as which instruments survived or which strategies worked. Both inflate results, through different doors.

Survivorship bias

Survivorship bias is the distortion that arises when a study only includes the instruments, strategies or traders that still exist or are still visible, leaving out those that failed. The surviving sample looks better than the population it came from, so backtests and track records built on it overstate what could have been expected.

Senzoukria · Glossary · Updated September 2026

Three places it enters

The data. A universe of instruments chosen from what is listed today excludes those that were delisted, merged or collapsed. In crypto, exchanges regularly delist pairs; a test on today's listed pairs never holds a coin through its failure.
The strategies. When many ideas are tested and only the ones that worked are kept and reported, the published set is a survivor sample. This overlaps with data snooping: the failures are what would have corrected the picture.
The people. Visible track records, public payout screenshots and success stories come from those who stayed. Traders whose accounts failed rarely publish, so the visible sample describes the winners of a lottery as much as a skill.

A coin-flip illustration

Give 100 people a fair coin and ask each to call ten tosses. The probability of calling at least seven correctly by pure chance is 176 ÷ 1,024 ≈ 17%, so about seventeen people will show a 70% or better record. If only those seventeen are interviewed, the sample suggests a group of skilled forecasters; the other 83 are the part of the population that makes the result meaningless.

Nothing in the seventeen records is false. The bias lies entirely in which records reached the reader.

In Senzoukria

The automatic backtest runs on the bars of the contract you choose, so which instruments and periods enter a study is the researcher's decision, and the survivorship risk lies in that choice. The Performance panel keeps every finished replay session and automatic backtest in its list, with a delete button for each run; deleting the runs that failed recreates survivorship inside your own records. The Gauntlet counts every configuration of a sweep when it deflates the Sharpe ratio, but it cannot count ideas you tried and discarded in earlier runs. A research record that lists rejected ideas alongside kept ones is the practical countermeasure.

On crypto charts, history is rebuilt from the exchange's public daily archives for the pair you open; a day whose archive is not published is listed as missing rather than filled from another source, so gaps are visible instead of being quietly skipped.

Common mistakes

Building a crypto universe from the current top pairs by volume and testing it over years.
Reporting the three strategies that worked out of twenty tried, without the twenty.
Judging a method by the traders who publicly succeeded with it.
Deleting losing backtest runs to keep the results list clean.

In the same section

This page in other languages

Français

Frequently asked questions

Do futures backtests suffer from survivorship bias?: Less at the instrument level for major index contracts, because a continuous series of ES or NQ exists throughout. The bias remains in strategy selection and in the choice of which markets to study, often made because they performed well recently.
How is survivorship bias different from lookahead bias?: Lookahead bias uses information that was not available at decision time inside a test. Survivorship bias selects the sample itself using an outcome known only afterwards, such as which instruments survived or which strategies worked. Both inflate results, through different doors.

Keep reading

Strategy researchTest a trading strategy, from observation to challenge rulesDefine an observation, question it with the copilot, write a rule and explore costs, drawdown and fictional challenge limits in an interactive example.Quant · Start hereQuant trading for beginners: from idea to backtestDefine a testable rule, choose suitable data and evaluate later periods after costs.Order flow · HubOrder flow trading: the complete guideThe hub. What order flow is, what it shows — aggression, delta, imbalance, absorption, liquidity — and the tools that read it, each linked to a full guide.