During World War II, military engineers studied the planes that returned from combat to decide where to add armor. They marked every bullet hole and saw a clear pattern: the wings and tail were riddled with holes, while the engine had almost none. The conclusion seemed obvious: armor the wings and tail, where the bullets land.
The mathematician Abraham Wald said exactly the opposite. Those planes were in front of them because they had come back. The ones hit in the engine never returned: they crashed. The holes they saw did not mark the dangerous areas; they marked the areas where a plane could take a hit and still survive. They needed to armor precisely where the surviving planes had no holes, because those were the invisible hits from the planes that never came home.
That is survivorship bias: drawing conclusions by looking only at the ones who made it to the end, ignoring everyone who fell along the way and vanished from the picture.
What survivorship bias is
It is a very human logic error: we judge a situation based on the visible examples that survived, without noticing that the failures have gone silent. It is not that we lie; it is that we cannot even see half the story. Missing data does not shout. It simply is not there.
And when the failures are missing, everything looks easier, safer and more profitable than it really is.
Survivorship bias in trading
Trading is full of planes that never came back. A few examples:
- The funds that closed. When you read that "funds return X% on average," that average is usually calculated using only the funds that are still open. The ones that did badly closed, merged and disappeared from the statistics. The average of the survivors always looks better than reality was.
- The traders who quit. On social media you see screenshots of accounts that multiplied their money. You do not see the thousands who lost, stayed quiet and deleted the app. Winners post; losers disappear. That is how it looks like "so many people are winning."
- Backtests on stocks that are still listed today. If you test a strategy using only the companies that are still on the market, you are cheating without meaning to: you are picking the survivors. The ones that went bankrupt, got delisted or collapsed are no longer on your list. Your backtest looks beautiful because you only tested it on winners.
Why it inflates your expectations
Survivorship bias is dangerous precisely because it is encouraging. It makes you think winning is normal, that almost everyone manages it, and that a strategy that "worked" in the past will keep working. But the past you see is carefully filtered: it only contains the survivors.
If you only count the winners, winning looks easy. Reality includes everyone who fell in silence.
Making decisions with biased data is like armoring the plane's wings: it seems reasonable and is exactly the opposite of what you need.
How AlphaLab handles this
AlphaLab is built to show you the planes that did not come back too. It does this in two ways:
- Survivorship-free (point-in-time) data wherever possible. Instead of testing strategies only with the companies or instruments that exist today, AlphaLab tries to use data that reflects the universe as it actually was at each moment in the past, including the ones that later went bankrupt or disappeared. That way the backtest is not playing only with winners.
- It also publishes the strategies that FAIL. AlphaLab does not show you only the pretty discoveries. It is a falsification lab: it subjects each strategy to brutally hard statistical tests and shows you which ones do not survive. Seeing the failures is not depressing, it is honest: it is the only way to get the full picture and avoid the survivor's illusion.
Remember: AlphaLab is a research tool that runs on your own PC. It is not a signals service nor a promise of profit. Trading always carries a risk of loss.
Key takeaways
- We tend to judge the world by looking only at survivors; failures become invisible.
- In trading, closed funds, traders who quit and backtests on still-listed stocks all inflate the appearance of success.
- Survivorship bias makes you believe winning is easier and safer than it is.
- AlphaLab fights this with survivorship-free point-in-time data and by also publishing the strategies that fail.
If you want to see the full picture of your research, including the planes that never came back, you can try AlphaLab free for 14 days (card required, cancel anytime) at whop.com/alphalab-005b/alphalab-pro.