🏠 Home 🔍 Today's Scan 📰 Daily Report 📈 Hit Rate 📊 Base Rate ❓ Methodology 📚 Learn 🪙 Crypto Scanner 📋 All Tools ⏪ Investment Simulator 🧾 Tax Calculator🧮 Pension vs. Direct ⚡ Leverage ⚖️ Rebalancing 💹 DCA 📉 Averaging-Down B/E 💰 Dividend Calendar
← Back to the Learn hub
📉 Statistical Traps · Verification

Survivorship Bias
— Drop the Failures and Performance Improves on Its Own

A number like "this strategy's return over the last 5 years" can hide a trap. If you don't count the stocks that disappeared, performance improves on its own. We explain why this happens, and how to screen for it.

Written by Dawn · IT Engineer · Published
💡 Key takeaway — Look only at what survived, and everything always looks like it did well. That's because the worse an outcome is, the more likely it is to vanish from the sample first.

What Survivorship Bias Is

Survivorship bias is "the distortion that comes from analyzing only the subjects that remained until the end." The problem is that what stays and what disappears isn't random. Usually, it's the failures that disappear.

The famous example is the World War 2 bomber story. When researchers gathered up the bullet-hole patterns on returning bombers, the damage was concentrated on the wings and fuselage. The conclusion was "reinforce those spots" — until the statistician Abraham Wald objected. The data was missing the planes that never came back. A plane hit in the engine never made it home at all, so the spots that actually needed reinforcing were the spots with no bullet holes.

How This Shows Up in Stocks

In stock data, the "planes that never came back" are delisted or trading-halted stocks.

  • Backtest the past using today's list of listed stocks, and the stocks that failed and disappeared during that period are already missing from the start.
  • The stocks that disappeared are mostly the ones that performed badly.
  • As a result, "what the return would have been if you'd used this strategy back then" comes out higher than it actually was.

The same thing happens with fund performance statistics. Poorly performing funds get liquidated or merged into another fund and drop off the list. The average return of the funds that remain naturally looks better than reality.

⚠️ Why This Is Especially Dangerous for a Surge-Stock Scanner — The group of stocks that tends to show a clean pre-breakout signal is small-cap, low-price, high-volatility stocks. But that exact same group is also the one with the highest delisting probability. Drop the failures, and it's precisely the failure cases from the riskiest segment that vanish wholesale.

How DawnScan Handles This

The base-rate proof statistics keep delisted and trading-halted stocks in the sample.

  • If data for a stock being tracked cuts off, it isn't deleted — it gets closed out with a "delisted" status.
  • The high and the return up to the closeout point get locked in as the final label and included in the tally.
  • These stocks mostly miss the target, so the hit rate goes down. That's an intended result.

If the number on the hit-rate page feels low, some of that is due to this handling. It means the failures haven't been erased.

What to Check When Looking at Someone Else's Track Record

  1. "What point in time was the stock list taken from?" — if the past was simulated using today's list of listed stocks, there's bias.
  2. "How were delisted stocks handled?" — no answer usually means they're missing.
  3. "Can I see failure cases along the way?" — showing only successes is itself a signal.
📌 Summary — Survivorship bias happens automatically, with no deliberate manipulation needed. Just take the data as it comes and it's already biased. So "does it include the failures?" is the first question to ask when looking at a performance statistic. The same principle applies when computing a base rate too.
📮 Daily US Market Morning Brief — We send an analysis of the previous day's top 10 US gainers (TOP10) and what they had in common, every day at 8am (KST). Telegram @dawnbrief · Free · No ads · Not stock recommendations.

Frequently Asked Questions

What is survivorship bias?

It's a statistical distortion that comes from analyzing only the subjects that remained until the end. Because what disappears isn't random — it's usually whatever performed badly — looking only at what's left produces a better-than-real result. In stocks, this shows up as delisted or trading-halted stocks being dropped.

Why is this especially a problem in backtesting?

If you simulate a past period using today's list of listed stocks, the stocks that failed and disappeared during that period were never candidates to begin with. The losses you would have actually experienced investing back then get dropped entirely, inflating the return. The distortion is larger for strategies that trade small-cap, low-price, high-volatility stocks.

How does DawnScan handle delisted stocks?

Instead of erasing them from the sample, it closes them out with a "delisted" status and includes them in the tally. The high and the return up to the closeout point get locked in as the final label. These stocks mostly miss the target, which lowers the hit rate — but that's the number closer to reality.

What should I ask when looking at someone else's performance statistics?

① What point in time was the stock list based on ② how delisted stocks were handled ③ whether failure cases are published too. If there's no answer to any of the three, it's safer to assume survivorship bias is present.
📊
Base-Rate Proof
Real measured statistics including delistings
📈
Scanner Hit Rate
Every failure published too
📐
What a Base Rate Is
Why a comparison baseline matters

Related Reading