Almost every success story you can read about was selected for having succeeded. The ones that would have made the lesson honest are not in the sample.
8 min read · Level 1
The most dangerous thing about survivorship bias is that correcting for it does not just lower your expected return a little - it often reverses the conclusion entirely. A strategy that looks excellent among survivors can be worse than doing nothing once the casualties are put back in, because the same feature that produces the spectacular winners is what produced the silent losses. If you only ever adjust the number downwards a bit and keep the original conclusion, you have not really applied the idea.
Survivorship bias is what happens when the things that failed drop out of the data before you look at it, leaving a sample made entirely of things that made it. The result looks like evidence about a strategy, an asset class or a type of founder, and it is really evidence about what happens to the ones that worked - which tells you almost nothing about your own odds.
It is everywhere in investing because failure is quiet. Companies that go under stop filing and stop being written about. Funds that perform badly get closed or merged into better-performing ones, and their record leaves the database with them. Investors who lost money are not the ones writing about their method.
Because it systematically inflates the returns you think are available and deflates the risk you think you are taking, in the same direction, every time. Any historical average built from survivors is too good, and any list of 'what successful investors did' is compiled from people for whom it worked, with no way to see how many did the identical thing and are not on the list.
Educational, not personalised financial advice. Nothing in the Investor Academy is a recommendation to buy, sell or hold anything — what is right for you depends on your own circumstances and the rules where you live.
Take a fund family that launched 100 funds a decade ago. Sixty are still running and have averaged 9% a year. Forty were closed or merged away after poor results, averaging -4% before they disappeared. The advertised figure - the average of what still exists - is 9%. The average an investor picking blindly from the original hundred would actually have experienced is (60 x 9 + 40 x -4) / 100 = 3.8%. The gap is 5.2 percentage points a year, and nothing in the marketing was false. The forty simply are not in the sample any more.
Studying successful investors to extract their method, without any access to the people who used the same method and failed. If a hundred people concentrate heavily in a single volatile position, a few will end up conspicuously wealthy and the rest will be quiet. Reading only the few produces a confident conclusion about concentration that the full sample would flatly contradict.
For any track record, performance table or list of examples, ask explicitly what happened to the ones that are not shown, and whether their results are included anywhere in the figure you are looking at.
Reputable index and fund research explicitly reports whether closed funds are retained in the sample. Where a provider says nothing about it, assume they are gone.
Deliberately seek out post-mortems, insolvency accounts and honest write-ups of things that did not work. They are harder to find, which is exactly the problem.
Whenever you meet a historical return figure, a manager's track record, or an argument that takes the form 'the people who succeeded all did X'.
This is not a reason to dismiss all evidence from successful cases. Success stories can genuinely show you what is possible and how a thing was done - they just cannot tell you how likely it is, and it is only the likelihood claim that survivorship bias destroys.
Providers of fund performance research routinely publish survivorship and persistence statistics showing that a substantial share of funds available at the start of a long measurement window no longer exist at the end of it, having been closed or merged. These reports exist precisely because comparing only the surviving funds produces a materially better-looking average than the full original set.
When a research provider makes a point of saying its data is survivorship-free, that is a signal about how different the two numbers are.
A widely recounted wartime analysis concluded that armour should be added where returning aircraft showed no damage, on the reasoning that the aircraft hit in those places had not returned to be examined. The sample of surviving aircraft was systematically missing the very cases the study was trying to learn from.
The absence of damage in the sample was information about who was missing from it, not about where damage does not occur.
Create a free account to read the rest and apply it to your own investing.