Survivorship Bias in Stocks: Why Best-Stock Lists Lie

There is a famous story from the Second World War. The military studied bombers that returned from raids, mapped where they were riddled with bullet holes — wings, tail, fuselage — and proposed armouring those areas. A statistician named Abraham Wald pointed out the flaw: they were only looking at the planes that came back. The holes showed where a bomber could be hit and still survive. The places to armour were the ones with no holes on the returning planes — the engines, the cockpit — because the bombers hit there never returned to be studied.

Every "best stock of the last ten years" list you have ever read is the bullet-hole map. It shows you the planes that came back. The companies that looked exactly as promising a decade ago — same sector, same story, same confident guidance — and then quietly delisted, halved, or drifted into irrelevance are simply not in the dataset. They didn't return, so nobody studied them. And there were a lot of them. Across more than 1,500 listed Indian companies on Inve, 13,280 management commitments are tracked, and of the ones that have actually come due, a large share have not been delivered as stated; 934 were quietly dropped and never mentioned again (Inve data, as of 2026-06-19). One caveat worth stating up front: Inve's transcript record is only about two years deep, and "delivered as stated" can only be counted for commitments that have already resolved — many are still open, on track, and simply not yet due — so these counts describe a young, still-filling record, not a settled scorecard on Indian management. The lists never show you the companies built on the commitments that didn't hold.

This piece is about what those lists systematically hide, why the deletions matter more than the entries, and where you can watch the same deletion happen — in miniature, in real time — inside a company that is still very much on everybody's screen.

What is survivorship bias, and why do 'best stock' lists run on it?

Survivorship bias is the error of drawing conclusions from the things that made it through a selection process while ignoring the things that didn't — because the ones that didn't are no longer visible. You judge the success rate of a strategy by looking only at its successes, and the strategy looks far more reliable than it was.

A "top performer of the decade" list is survivorship bias in its purest commercial form. It is assembled today, looking backward, from the set of companies that are still around and still worth writing about. The selection happened after the outcome. Every name on it is there because it won — which tells you nothing about whether you could have identified it as a winner ten years ago, when it sat in a list of two hundred companies that all looked equally plausible.

Here is the homely version. Reading a "best stocks of the decade" list to learn how to invest is like studying only the people who won the lottery to learn how to get rich. Every single winner did, in fact, buy a ticket — so you conclude that buying tickets works. What you can't see, because they don't write articles, are the millions who bought identical tickets and won nothing. The winners are real. The lesson drawn from them is an illusion, because the losers were deleted from the sample before you ever looked.

What do these lists delete, exactly?

Three categories, all invisible by the time the list is written, and all of which would change the lesson if you could see them.

The dead. Companies that delisted, went bankrupt, or were suspended. India has produced plenty — names that were market darlings before governance blow-ups, debt spirals, or fraud erased them. They had stories every bit as compelling as the survivors. By the time the "best of" list is compiled, they have been scrubbed from the index and from memory. You will never see "and here are the forty companies that looked just as good and went to zero" — that article doesn't get written.

The has-beens. Worse, in a way, because they're still listed and still trading at a fraction of their peak — the once-celebrated leaders that compounded downward for a decade. They rarely appear on either the winners' list (they lost) or the cautionary list (they didn't dramatically die). They just faded. Their absence quietly inflates the win rate of whatever strategy supposedly identified the survivors.

The deleted commitments. This is the one specific to how the loss actually happens, and it is measurable. The companies that fell out of the lists usually didn't collapse overnight. They eroded — and the erosion was visible, quarter by quarter, in guidance that was given and then silently abandoned. Of the 13,280 commitments on Inve, 934 were ghosted — stated once on a call and never raised again — and about a third of companies (534 of the ~1,500 tracked) carry at least one (Inve data, as of 2026-06-19). A ghosted target is a deletion in miniature: a piece of the original investment story that quietly left the record. The "best stock" list deletes whole companies; managements delete individual commitments. The mechanism is the same — and the second one is the early warning for the first.

What does a deletion-in-miniature actually look like?

Take a name still firmly on everybody's screen: SG Mart, the steel-distribution venture. Treat what follows strictly as a mechanism illustration — how a once-anchored number can recede from the record — and not as a view, recommendation, or "avoid" call on the stock; the point is the pattern, not the company. When the venture was relaunched, the public framing was a steep multi-year ramp: revenue scaling toward the ₹12,000-13,000 crore range and, further out, much larger numbers by the end of the decade. It was a clean, confident, list-worthy story: a young company with cash on the books and a credible right to win, guiding to near-doubling year after year.

Now hold that early framing against what actually showed up in the financials. For FY26, SG Mart's consolidated revenue was about ₹6,315 crore (Inve data, FY26 — the sum of the four quarters to March 2026); the standalone figure is lower, around ₹5,540 crore, but because the original ramp was framed as a whole-venture number, the consolidated figure is the fairer comparator and the one we use here. Either way, the gap is the point: ₹6,315 crore is roughly half of the ₹12,000-13,000 crore range that had anchored the story — against a number the venture had set for itself. The early framing didn't fail loudly. It simply stopped being the number anyone was measured against.

A fair counter-argument deserves stating plainly. A young, fast-scaling business should revise its own targets as it learns; walking a number down once, and saying so, is prudence, not deletion — legitimate guidance revision is a normal part of how honest managements operate. What turns prudent revision into the pattern this article is about is silence: a number that recedes with no owned restatement, dodged or unaddressed across several quarters, until there is nothing left to hold management to. The danger sign is not the lower number — it's the missing acknowledgement.

We can only verify so much of the call-by-call record here. Inve's transcript archive for SG Mart currently holds a single, recent concall, so the quarter-by-quarter "staircase" of earlier verbal targets cannot be reconstructed from the parsed transcripts at the time of writing — the comparison above rests on the reported FY26 financials against the publicly framed ramp, not on a chain of quoted ghosted commitments. That limit is itself the lesson of this piece: a story can be loud at launch and thinly documented later, and the parts that go quiet are exactly the parts a backward-looking "best stock" list will never miss on your behalf.

Hold the two ends side by side: a steep multi-year growth story narrated at launch, against a business reporting roughly half the near-term number it had pointed at — and watch how easily the early framing recedes from view. The "best stock" list deletes the whole plane after it goes down. A commitment tracker's job is to record each rivet that worked loose first — and to flag, honestly, where the record is too thin to be sure.

Why is this more dangerous than it looks?

Because survivorship bias doesn't just mislead you about the past. It rewires the questions you ask about the future, and it does so invisibly.

When you study only winners, you learn the wrong rules. You see that the survivors had bold guidance, charismatic promoters, and exciting stories — and you conclude that bold guidance and charisma predict winners. But the companies that died also had bold guidance and charisma. Those traits were not what separated the survivors; they were common to the entire starting population. The list can't show you that, because the dead aren't in it. So you go looking for exactly the features that are equally present in future disasters.

Invert it the way Wald did. Don't ask "what did the winners have in common?" Ask "what did the losers have in common that the winners didn't — and is that thing visible before the outcome?" That question is far more useful, and it points somewhere specific: the losers, again and again, gave guidance they then quietly walked away from, and dodged the questions that would have exposed it. Inve has flagged on the order of a thousand such evasive answers across its tracked calls (Inve data, as of 2026-06-19). The winners' lists can't teach you to spot that pattern, because they've deleted every company that displayed it.

The lollapalooza here is how the biases stack. Survivorship bias hides the losers; hindsight bias makes the surviving winners look inevitable ("obviously that was a great company"); and recency bias (the NISM curriculum's projection bias — extrapolating the recent past indefinitely) makes the latest winner feel like the safest bet. All three converge on the same false conclusion — that picking winners is easier than it is — and they reinforce one another precisely because the disconfirming evidence has already been removed from view.

How do you invest against survivorship bias?

You can't recover the deleted companies. What you can do is stop learning from a censored sample and start watching, in your own holdings, for the early signs of the erosion that gets companies deleted in the first place.

Study the failures, not just the successes. Where the record exists, the more instructive exercise is the post-mortem: take a company that faded and read its concalls from before the fade. The warning signs — guidance that drifted down and went silent, a sore-spot question dodged for quarters, a flattering metric that quietly stopped being mentioned — were almost always on the record before the price confirmed them, and the parsed concall summaries make that record readable without wading through years of transcripts. The winners can't teach you this. The fallen can.

Watch for deletions in real time, in your own portfolio. The defence against owning a future deletion is not a better list of past winners. It is catching the small deletions — the ghosted commitment, the receding target, the new evasion streak — while they're still small. Inve's Promise Tracker holds each commitment's original wording against its current verdict, so a target that has quietly left a company's record surfaces as a flag rather than vanishing the way it's designed to. Across a 15–25 stock portfolio, that's the difference between noticing the erosion in Q2 and reading about it in a cautionary article three years later.

Distrust the backward-looking list as an input. A "top performers of the decade" list is fine entertainment and useless as a research method. If you must use one, use it to generate names to investigate from scratch — not as evidence that those names share a winning formula. The formula is an artefact of deletion.

Where we could be wrong: survivorship bias is not a reason to dismiss every long-term winner as luck. Genuine quality compounds, and some businesses really were identifiable as superior years ago — durable moats and disciplined capital allocation are real and were visible in the record at the time. The error is not believing winners exist. It is inferring the rules for finding them from a sample that has secretly removed all the look-alikes that lost. Quality is real; the lists just can't teach it, because they've thrown away the control group.

What should you actually take from a 'best stock' list?

One thing only: humility about your own visibility. For every name on it, picture the dozen companies that stood beside it a decade ago — same sector, same conviction, same confident guidance — and have since been deleted from the conversation. You are not seeing the choice as it looked then. You are seeing the answer key, with all the wrong answers erased.

The useful work isn't admiring the survivors. It's watching your own holdings for the quiet deletions — the commitment that goes silent, the question that gets dodged — because that erosion, caught early, is the only part of the next "best stock" list you actually get to influence: keeping your companies on it rather than in the part nobody writes about.

Frequently asked questions