{"slug":"survivorship-bias","term":"Survivorship Bias","category":"Quant & Stats","definition":"Survivorship bias is drawing conclusions from the things that lasted while ignoring the ones that did not — backtesting only coins that still trade, studying only traders who are still trading, admiring only launches that graduated. The failures vanish from the dataset, so the survivors look like a rule instead of a residue.","street":"Every memecoin millionaire posts screenshots; the ten thousand who lost everything on the same coins do not post. The feed is a survivor list.","depth":["In a coin universe, testing a strategy on today's top 300 quietly excludes every delisted, rugged and dead token that was in the top 300 when the trades would have happened — and those are disproportionately the ones a trend-following rule would have bought. Our own ledger includes every call ever issued, including the ones on coins that later fell out of the list, because a record has to keep the corpses.","The meme corpus is the sharpest example: 5,139 launches recorded, 570 graduated. Studying the 570 tells you what graduations look like; only the full 5,139 tell you what predicts one, which is why our cabal report computes odds ratios against the whole population and reports the base rate first. Whenever a dataset arrived through a filter, ask what the filter threw away."],"live":{"label":"Graduation odds against every recorded launch (free sign-in)","href":"/meme"},"related":["sample-size","overfitting","backtest","look-ahead-bias","rug-pull"]}