Why every backtest works
Every fund track record beats the market and every backtest is a winner — because the losers have been deleted. A look at survivorship bias in fund databases, stock indices and strategy backtests.
Every backtest works. Every fund advertisement shows a track record that beat the market. Every “our model would have turned £10,000 into £240,000” pitch has the equity curve to prove it. If the markets are so efficient and most active managers underperform, how is it that almost every historical record you are ever shown is a winner?
The answer is the most financially expensive instance of a single statistical illusion: survivorship bias. The records that lose are not refuted — they are deleted. Funds that fail get merged or liquidated and vanish from the database; companies that go bankrupt drop out of the index; strategies that flop never make it into the pitch deck. What reaches you has passed through a filter that quietly removes the failures, and an average computed over survivors is not the truth about investing. It is a memorial to the ones that made it.
The fund graveyard
Start with the number you see most often: the average return of some category of funds. It is almost always computed over the funds that still exist — and a startling share of them do not survive. Over long horizons, roughly half of all funds are merged into other funds or shut down entirely. In one fifteen-year US sample, only 42 of every 100 starting funds were still around at the end; over twenty years, the SPIVA scorecards find nearly two-thirds gone.
This would not matter if funds died at random. They do not. The funds that get quietly folded away are overwhelmingly the underperformers — a struggling fund is merged into a sibling with a better record, its embarrassing history erased, while the winners stay open and keep reporting. So the survivors you average over are a sample selected for good performance, and the “typical fund return” is systematically too high. The studies that have measured this put the overstatement at roughly one to two percentage points a year — Burton Malkiel found about 1.5% — which, compounded over an investing lifetime, is the difference between a comfortable retirement and a disappointing one.
The effect is so mechanical it can make history improve in hindsight.
There is no fraud in any of this. Each average is computed correctly over the funds that happen to exist when it is taken. The losers simply aren’t there to be counted, and their absence is invisible precisely because they are absent. You cannot see the gap a deleted fund leaves; that is what makes the bias so reliable, and so persuasive.
The backtest that tests only the living
Now turn from fund records to the thing every quantitative pitch is built on: the backtest. Here survivorship bias hides in a subtler place — the universe the strategy is tested on. Suppose you want to check whether some rule for picking stocks would have worked over the last twenty years, so you run it on the companies in a major index. The trap is that you are running it on the companies in the index now — which is to say, on the firms healthy enough to have survived to the present.
Every company that went bankrupt, got acquired in distress or was delisted for failing has been removed from your sample before the test begins. Your strategy never has to live through holding Enron into oblivion, or Lehman, or any of the hundreds of dead tickers that a real investor would have owned. Tested only on the firms that made it, almost any plausible-sounding rule looks profitable, because the single biggest risk in real investing — catastrophic loss — has been quietly engineered out of the data. Serious researchers use “point-in-time” databases that record exactly which companies were in the index on each past date, dead ones included. Most casual backtests do not, and silently test a strategy against a universe that already knows who wins.
The strategy that survived the search
Even with perfect, graveyard-inclusive data, one last survivor problem remains — and it is the one that sells the most products. It is survivorship not among funds or stocks, but among strategies.
Imagine testing a thousand trading rules against the historical data. Pure chance guarantees that some of them will show spectacular backtested returns, the way that flipping a thousand coins guarantees a few long streaks of heads. The rule with the best backtest is, more often than not, simply the luckiest — and it is the only one that ever appears in a brochure. The hundreds of rules that failed are never mentioned, so you see a single gorgeous equity curve with no idea how many ugly ones were tried to find it. This is the Texas sharpshooter fallacy wearing pinstripes: paint the target around the winning shot, and a coincidence becomes a system. It also explains a notorious finding in the academic literature — that once you correct fund performance for survivorship, the apparent ability of past winners to keep winning largely disappears. The persistence was mostly the bias.
How to read a track record
The cure is not cynicism but a single reflex: whenever you are shown a glittering history, ask what is missing from it. In markets the missing are almost always the failures, and they are exactly the evidence you need to judge the survivors honestly.
None of this means investing is hopeless or that all history is a lie. It means that the raw, survivor-only record overstates how well funds, stocks and strategies have done, by an amount large enough to change real decisions. It is why a cheap, broadly diversified index fund — which simply owns the whole haystack, survivors and stragglers alike, and doesn’t need a flattering backtest — quietly beats most of the dazzling track records waved in front of it. The brochure shows you the winners because the winners are all that is left to show. The losers are not evidence against the pitch; they are the part of the evidence that was thrown away.
So the next time a return looks too good — a fund category that always seems to beat the index, a strategy that turned a little into a fortune, a manager with an immaculate decade — do not ask how the survivors did. Ask how many started, and where the rest are buried. The honest answer is almost always in the graveyard nobody put on the chart.