Ellen Marsh bought her first share of Halcyon Grid on a Tuesday in February 2021, at 9:52 in the morning, for $26.40. She sold nothing. Five years later the position was worth nothing either, and the difference between those two sentences is the whole subject of this note.
We want to walk through the life of a listing the way an ordinary shareholder actually experiences it — the bell, the good year, the letter from the exchange, the queue at the end — and then set that experience against the arithmetic. The arithmetic is not a secret. It is simply never presented at the moment it would be most useful, which is the moment you are deciding to buy.
01 / THE BELLThe day the window opens.
Halcyon Grid priced its offering at $18 a share. By the closing bell of the first session it was $24.10. Every account of the day called this a success, and in one narrow sense it was: the company had raised its money and the shares had gone up.
Ellen did not get $18. Almost nobody outside the allocation process does. The offer price is available to the institutions and the favoured accounts on the book; everyone else buys in the aftermarket, at whatever the aftermarket decides. She bought the next morning at $26.40, which felt like discipline — she had waited a day.
That gap between the offer price and the first close has a well-measured size. Across 9,253 US initial public offerings from 1980 to 2024, Jay Ritter's data puts the average first-day return at 18.9%. It is the most reliably positive number in the entire lifecycle of a listing, and it is almost entirely captured before the general public can participate.
The first day is the part of the story everyone reports. It is also the only part most outside investors cannot buy.
There is a second fact about that morning worth holding onto. Halcyon Grid was not profitable and did not claim to be. That was unremarkable for the window it listed in — the share of US IPOs coming to market unprofitable has climbed from roughly 20% in the 1980s to around 80% in recent years. Ellen was not buying earnings. She was buying a plan, priced as though the plan had already worked.
02 / THE GOOD YEARWhen the thesis appears to be confirmed.
By June the stock was $41. Two banks initiated coverage with price targets above it. A trade publication ran a feature. A supply agreement was announced with a utility Ellen had heard of. Her position was up 55% and the position size, which she had deliberately kept small, now looked like a failure of nerve rather than a piece of prudence.
This is the phase that does the real damage, and it does it quietly. A rising price is read as evidence. It is not evidence; it is a price. But the experience of being right is powerful enough that most people respond to it by getting larger, and by moving the sell discipline they wrote down in January.
Ellen added at $38. Her average cost went from $26.40 to $31.60. She would later describe this as the decision that mattered, which is very nearly correct — though as we will see, the decision that actually mattered came three years later.
03 / DAY 181The lockup expires.
Most US IPOs carry a lockup of 180 days, during which insiders, employees and pre-IPO investors cannot sell. The lockup is not a secret and its expiry date is not a secret; it is disclosed in the prospectus. Almost no retail holder has it in the calendar.
On day 181, volume in Halcyon Grid roughly tripled. The stock went into the expiry around $31 and came out under $28. Research on lockup expirations finds an average price decline of a few percent that is statistically significant and typically not recovered — with the important caveat that roughly a third of expirations produce a price increase, so the effect is real on average and unpredictable in any single case. Thinly traded small caps tend to take it worse than large ones with heavy analyst coverage.
What made the day disorienting for Ellen was not the size of the drop. It was the absence of news. Nothing about the business had changed. The only thing that had changed was who was allowed to sell, and the answer was now: everybody who had bought in at a fraction of her price.
04 / THE SLOW PARTDilution, and the arithmetic of averaging down.
The first genuine miss came fifteen months in. A deployment slipped two quarters; guidance came down; the stock went to $11 over a fortnight and stayed there. Ellen read the filing carefully for the first time since the prospectus and found the sentence that mattered buried in the liquidity discussion: cash on hand was sufficient for approximately twelve months of operations.
A company that says it has twelve months of cash is a company that will be selling shares within nine. Halcyon Grid did, at $6.40, in a registered offering that increased the share count by 31%. Ellen's ownership of the business shrank by nearly a quarter without her selling anything or being asked.
She averaged down at $7.00. This is the point at which most people's losses stop being caused by the original decision and start being caused by the response to it. Her cost basis was now $16.21 across 3,100 shares, well over twice what she had held, and the position had gone from a small one to a meaningful one at exactly the moment the balance sheet was deteriorating.
Averaging down converts a small mistake into a large one whenever the thing that is falling is the business rather than the price.
05 / THE LETTERBelow a dollar, and the compliance clock.
Two more financings and eleven more months later, Halcyon Grid closed below $1.00. Nasdaq's rule is mechanical: 30 consecutive business days under a dollar triggers a deficiency notice, and the notice starts a 180-calendar-day compliance period. Companies on the Nasdaq Capital Market can often obtain a second 180 days, which in practice is usually cured the same way — a reverse stock split.
Halcyon Grid did a one-for-fifteen. Ellen's 3,100 shares became 206. The price went from $0.62 to $9.30 overnight and her position was worth exactly what it had been worth the evening before: $1,922, now as $1,916 in shares and $6 in cash for the fraction of a share the split could not issue.
A reverse split creates no value. It is a compliance action, and one of the more reliable pieces of information on a screen, because companies that are healthy do not generally need one. Nasdaq has been tightening around exactly this behaviour — a security trading at or below $0.10 for ten consecutive days can be delisted without a compliance period, cumulative reverse splits of 250-to-1 or more over two years disqualify a company from the process entirely, and since 2025, a company that has already done a reverse split within the prior year gets no compliance period at all when it falls below a dollar again.
That last rule is a regulator saying, in the driest possible language, that the second reverse split is the tell.
06 / THE QUEUEWhat common stock is, at the end.
Halcyon Grid filed for Chapter 11 in the spring of its fifth year as a public company. The shares stopped trading on Nasdaq and reappeared over the counter with a "Q" appended to the symbol, at eleven cents. Delisting itself did not cancel anything — this is a point worth being clear about, because it is widely misunderstood. A delisted share is still a share. It trades in a thinner venue, with wider spreads and less disclosure, but it exists.
What cancels it is the plan of reorganisation. In a bankruptcy, claims are paid in order: secured bondholders, then unsecured bondholders, then subordinated debt, then preferred stock, and then — only if something is left, which there rarely is — common stockholders. FINRA's guidance to investors is blunt on the point: shareholders of bankrupt companies see little to no compensation, and cases where old shares are exchanged for shares in the reorganised company are especially uncommon.
Ellen's shares were cancelled in the confirmed plan. The eleven cents she could have taken in the OTC market during those months was, in hindsight, the entire remaining value of five years of conviction. She did not take it, because taking it would have meant admitting the thing she had spent three years not admitting.
07 / THE ARITHMETICHow often this happens.
Now the wide view, which is the part we built this site to measure properly.
Start with the population. The number of operating companies listed on Nasdaq and the NYSE peaked at 7,451 at the end of the second quarter of 1997. It fell to 3,631 by the first quarter of 2013 and stood at 3,657 at the end of 2025. The US listed market has roughly halved in thirty years, and it has done so while the economy grew.
Then the flow. In their study of the US listing gap, Doidge, Karolyi and Stulz counted 17,303 delistings between 1975 and 2012 — 9,749 mergers, 7,120 delistings for cause, and 434 voluntary. The average annual rate at which listed companies left the market rose from 7.29% before the 1997 peak to 9.49% after it. Something close to one listed company in eleven stopped being a listed company each year.
And then our own count, which is the reason this note exists. In building the screen that runs on this site, we pulled every company that filed financial statements with the SEC between 2015 and 2025. There were 11,766 of them. When we checked in September 2026, 4,821 still carried a current listing.
We found this the hard way. An early version of our backtest could only price companies that still have a ticker today, which meant it was quietly measuring a market composed entirely of survivors — and reporting results that were correspondingly flattering. Every historical record you are shown has already been edited, and the editor was survival.
08 / WHAT "GONE" MEANSThe honest caveat.
Six in ten is a striking number and it would be easy to misuse, so here is what it does not mean.
It does not mean six in ten companies failed. The largest single category of delisting is merger — 9,749 of those 17,303, or about 56%. Being acquired is frequently the best outcome a shareholder gets, often at a premium, and it removes a company from the listed universe just as thoroughly as bankruptcy does. A count of disappearances cannot distinguish the shareholder who was bought out at $60 from the one whose certificate was cancelled.
Nor is our own 59% a pure failure rate. The SEC filer population includes entities that never had a common listing to lose, and a company can stop filing for reasons that have nothing to do with distress.
What the number does establish is the base rate of impermanence. Whatever the cause, the listed company you are researching today is, on the historical evidence, meaningfully unlikely to be a listed company in ten years. And the delistings-for-cause — 7,120 of them, roughly two in five exits — describe the specific path Ellen walked.
09 / THE RETURNSWhat the median listing actually did.
The story above is one outcome. The distribution is more useful than any single path, and it is unsentimental.
Measured from the first closing price — that is, from the price an ordinary investor could actually pay — the average US IPO from 1980 to 2024 returned 20.5% less than the market over the following three years. Across 9,195 offerings from 1975 to 2021, only 39.9% produced a positive three-year return. Some 38.5% lost more than half their value. At five years the median market-adjusted return was −32.0%, and 42.6% of IPOs were down more than half.
Size matters within that distribution. Among IPOs of companies with sales of $100 million or more, 48.8% produced a positive three-year return — still not a majority, but a materially different proposition from the pre-revenue end of the market where Halcyon Grid listed.
10 / WHAT WAS KNOWABLEThe parts Ellen could have read.
We are not going to tell you Ellen should have seen it coming. Plenty of companies with worse balance sheets have done fine and plenty with better ones have not. But it is worth listing what was on the public record, in filings anyone could open, before each leg down.
The lockup expiry date was in the prospectus, in a numbered section, eighteen months before it happened. The twelve-month liquidity statement was in the 10-Q two quarters before the dilutive offering. The share count increase was in the 10-Q after it, and could be compared to the one before it in about ninety seconds. The reverse split was announced in an 8-K. The second reverse split — the one the regulator now treats as disqualifying — was announced in another.
None of those is a prediction. Each of them is a fact that was available and free, and that collectively described a company consuming its own equity to stay listed. That is the whole premise of screening: not to forecast, but to make the facts that already exist show up in the same place at the same time, so a person can see them without having to already suspect something is wrong.
The purpose of a screen is not to predict the fall. It is to make the balance sheet impossible to avoid reading.
Which is also why we publish the whole ranked list rather than a handful of picks, and why our own testing found the top of the ranking performing roughly in line with the market rather than beating it. A rank is a filter for what to read about. It is not a forecast and it is not a recommendation. We would rather say that plainly than sell you the version of Ellen's story where somebody had the answer in advance.
11 / A NOTE ON THIS STORYWhat is real and what is assembled.
Ellen Marsh is not a real person and Halcyon Grid is not a real company. The story is a composite, constructed to walk through a sequence of events that are entirely real and extremely common: an aftermarket entry above the offer price, a first-year advance, a lockup expiry, a guidance miss, a dilutive financing, averaging down, a minimum-bid-price deficiency notice, a reverse stock split, a Chapter 11 filing, and a plan of reorganisation that cancelled the common stock.
Every one of those steps has happened to real shareholders of real companies many thousands of times, and the mechanics described at each step — the 180-day lockup, the 30-day bid-price trigger, the two compliance periods, the order of claims in bankruptcy — are the actual rules as they stand. The prices, dates, names and the person are invented. They are invented so that the sequence can be shown in one continuous line, which is how it is lived and not how it is usually reported.
No part of this note describes any specific company, and nothing in it should be read as a characterisation of any actual security, past or present.
The statistics, by contrast, are not invented and are sourced below. Where a figure comes from our own data, we have said so and described how we counted.
12 / SOURCESWhere the numbers come from.
Listed company counts and IPO long-run returns: Jay R. Ritter, University of Florida, Initial Public Offerings: Updated Statistics and Number of Operating Companies Listed on US Exchanges, figures current to year-end 2025. Delisting composition and delisting rates: Doidge, Karolyi and Stulz, The U.S. Listing Gap, covering 1975–2012. Minimum bid price and reverse split rules: Nasdaq Listing Rules 5810 and 5815, including the amendments approved by the SEC in January 2025. Bankruptcy claim priority and shareholder recovery: FINRA, What a Corporate Bankruptcy Means for Shareholders. Lockup expiry price effects: published research summarised by Aswath Damodaran, NYU Stern. Filer counts for 2015–2025 are our own, drawn from the SEC's Financial Statement Data Sets and checked against the SEC's current company-ticker file.
Corrected September 21, 2026: two figures in Ellen's story did not add up and have been fixed — how much the offering diluted her stake, and her cost basis after averaging down. None of the statistics changed.