On Sunday 26 October 2008, Porsche put out a statement. It held 42.6% of Volkswagen's ordinary shares outright and cash-settled options on another 31.5%, which came to a little over 74% of the company. The state of Lower Saxony held 20.1% and was not selling, because it never sells. Anyone with a calculator could work out what was left. Something under 6% of the ordinary shares were actually available to trade, and roughly 12.8% of the company had been sold short.

What happened over the next two sessions is one of the few things in market history that still reads as impossible. Volkswagen ordinary shares went from around 210 euros to an intraday print of 1,005. For part of Tuesday 28 October, in the middle of the worst global credit crisis since the 1930s, with new car sales falling off a shelf everywhere on earth, a German carmaker was the most valuable listed company in the world.

The people who were short were not idiots. They were right. Volkswagen's business was heading into a genuinely terrible year, the valuation was absurd, and within a few months the shares were back near 200 euros. Every single thing in the analysis was correct. It made no difference at all, because the analysis was about the company and the thing that killed the position was about the float.

I have thought about that week for years and I have come around to a claim that sounds cynical and is actually mechanical. Your edge and your ruin come out of the same hole. You are paid, every day, for standing in front of one specific risk. The payment is real and it is why the style works. The bill is also real, it is drawn on exactly the same mechanism that was paying you, and it arrives on a schedule you do not control. Skill decides how much you collect. Style decides what the bill looks like when it comes.

What to try: click any cell to change the focus, or use the pills underneath. Every path in the panel belongs to a trader who is good at this, with a positive drift and modest noise. The only thing that differs between the four is what the instrument does to the losses the stop was supposed to cap.

The two questions

Most style debates are about temperament. Are you patient or twitchy, do you like doing the reading or reading the tape, can you sit through a drawdown. I have written about that side of it before, in The Mirror and the Magnifying Glass, and I still think it matters. It is not what this piece is about.

There are two questions you can ask about an instrument that have nothing to do with the person holding it, and between them they do most of the work.

The first is whether your loss has a floor. Price is bounded below at zero and has no ceiling at all, which is such a familiar sentence that it slides past without anyone taking it seriously. Take it seriously for a second. If you buy at $P$ and the position is later marked at $mP$, your return on the position is $m - 1$. Because $m \ge 0$, that return can never be worse than $-100\%$. The worst case exists, you can name it before you enter, and no additional amount of being wrong makes it worse than the worst case.

If you sell short at $P$ and the position is later marked at $mP$, your return is $1 - m$. There is no upper bound on $m$. Which means there is no lower bound on that expression, and the phrase "maximum loss" has no referent. You can define a maximum loss for a short position only by naming a specific $m$ you are prepared to survive, and then the question becomes whether the market agrees to stay under it.

$\text{long: } r = m - 1 \ge -1 \qquad\qquad \text{short: } r = 1 - m, \quad m \in [0, \infty)$

What to try: push the multiple out to the right and watch the long line flatten against its floor while the short line keeps going. Then move the position size slider and watch where the vertical red marker lands, because that is the multiple at which a short of that size is the whole account.

At 20% of the account, a short is gone at 6 times the entry. That is not an exotic number in a small cap. I have watched worse than that happen between 9:30 and 9:34.

The second question is whether price is continuous. Everyone checks liquidity. The question sitting underneath it is whether price is obliged to print every level between here and there. A stop-loss order is a conditional instruction that quietly assumes it is: it says take me out when price reaches this level, which only means anything if price has to reach that level on its way to the next one. That assumption holds on almost every day, in almost every instrument, which is exactly what makes it so easy to stop noticing you are making it.

Put those two questions on a pair of axes and you get four cells. Long or short sets whether your tail is bounded. Small cap or large cap sets whether the tape can stop. Neither axis asks anything about you.

Where skill stops being defined

Here is the sharpest way I can put the thesis, and it is worth being precise because the imprecise version sounds like fatalism.

Skill is a function defined on the region where price is continuous and liquid. Everything a good trader is good at lives there: reading a book, sizing into a move, knowing when the tape has changed character, getting out of something at a price of roughly their choosing. All of it is real, all of it compounds, and all of it requires the market to keep printing.

Every catastrophic loss happens where those conditions break. So skill is defined precisely on the region where it does not matter, and undefined on the region where it does.

That is why the Volkswagen shorts lost. Their skill was fully engaged on the question of what the company was worth and completely irrelevant to the question of how many shares existed to buy back. And it is why a small cap halt is so much more brutal than its size suggests, because the halt does not make the news worse, it makes your order inapplicable.

The panel below runs the same event twice. Same news, same destination, two routes. On one the tape keeps printing and your stop is one of the levels price walks through. On the other the tape stops for a while and the first print anyone sees afterwards is already on the far side of you.

What to try: leave the stop at 8% and drag the reopening gap. The two routes always end the session at the same price, because it is the same news. Only one of them was obliged to pass through your order to get there.

I want to be careful not to oversell the small cap column here. Large caps gap too, over weekends and around earnings and on macro prints, and a gap is a gap. The difference is one of frequency and of scale, and it is large enough to be a difference in kind. A liquid mega cap gapping 8% through your stop on an earnings miss is a bad Tuesday. A $60m market cap biotech reopening from a T12 at a third of the halt price is a different event, and it happens to somebody every week.

Four premiums

Every cell on the map pays you for standing in it. That is the actual reason each style has an edge, and naming the payment tells you in advance what the bill will be.

Long, small cap. You are paid for the range a thin float will travel. A stock with four million shares available and a story attached can move 200% in a session, and if you can read which ones are going to, that range is your income. What you are absorbing in exchange is the company's need for money. Small caps are, overwhelmingly, businesses that do not generate cash and therefore have to sell shares to exist, and the moment your buying creates a price at which they can sell shares, they will. The bill arrives overnight, priced at a discount to the close, usually with warrant coverage attached. You were not wrong about the setup. You were paid for the range and charged for the shelf, and both came from the same feature of the instrument, which is that the float is small and elastic. I wrote the architecture of this out in detail in Toll Booths, Not Traps, so I will not repeat it here beyond the one line that matters: the dilution is the other half of the setup rather than a betrayal of it.

Short, small cap. You are paid for the fade. The same structural facts that make these things run make them come back, and a trader who can sit on the right side of the give-back can make a living out of one repeating pattern. What you are absorbing is everyone else who wants the same borrow, plus the mechanical consequence of a float too small to service them. Your bill is a squeeze or a halt, and it is delivered inside a minute, with your stop switched off for the part that matters. This is the only cell where both problems compound: the tail is unbounded and the tape is discontinuous, which is why it produces the fastest and most complete account deaths on the board.

Long, large cap. You are paid the equity risk premium, which is the most honest name for a payment anywhere in this article, because the finance literature calls it a premium out loud. You are compensated for holding an asset whose value is contingent on a regime continuing. The bill is that regime ending, and it is charged in years. The Nikkei closed at 38,915 on 29 December 1989 and did not see that number again until 22 February 2024, thirty four years later. The Nasdaq Composite peaked at 5,048 on 10 March 2000, bottomed 78% lower in October 2002, and took until 2015 to get back. Neither of those is a story about bad execution. They are stories about being paid to hold regime risk, and then holding it.

Short, large cap. This one does not fit the pattern, and the fact that it does not fit is the most interesting thing on the map.

The box that runs backwards

Three cells collect a premium. The fourth pays one.

Short a broadly held large cap and you are charged borrow every day, you are on the wrong side of the long-run drift of an equity market, and you fund any dividend rather than receiving it. None of that is contingent on being wrong. It is the price of the seat, and it accrues whether the thesis is brilliant or stupid.

Which means the thesis is never "this is overvalued." The thesis is "this is overvalued by more than the carry between now and whenever the market agrees," and the second clause has a date attached to it that most people never compute.

What to try: leave borrow at 4% and drag the wait out to three years, then push borrow toward 100% and watch the expiry date collapse into a single quarter. That collapse is the boundary between the two columns of the map.

I find this cell clarifying because it explains a pattern that otherwise looks like a psychology problem. Short sellers of large caps are famous for being early, and being early gets discussed as though it were an impatience issue or an ego issue. The arithmetic says something plainer. In a position with negative carry, early and wrong are financially similar, and the more thorough your research, the longer you are willing to hold, and the longer you hold the more of the eventual payoff the seat has already consumed. The quality of the work extends the exposure to the thing that charges you. That is a genuinely uncomfortable feedback loop and no amount of discipline dissolves it, because it is priced into the instrument before you arrive.

It also explains why the successful practitioners in this cell are almost all catalyst traders rather than valuation traders. A catalyst is a way of putting a date on the second clause.

How long until the bill

Now the part I did not expect when I started laying this out, and the part I would keep if I could only keep one paragraph.

The four claims are about seven orders of magnitude apart in how long they take to be delivered. A halt reopens in the time it takes to read this sentence. An offering is priced overnight. A short thesis bleeds out over quarters. A regime takes a decade to finish with you.

That spread does something to the people standing in each cell, and it does it invisibly. If claims arrive at a rate of $\lambda$ per year, then after $t$ years the chance your entire experience contains none of them is

$P(\text{no claim yet}) = e^{-\lambda t}.$

For a small cap short seller, $\lambda$ is large and that probability collapses inside a season. They have been shown their own tail, repeatedly, at their own expense, and their beliefs about risk are built from evidence they paid for. For a long-biased large cap trader, $\lambda$ might be one in eleven years, and a decade at the desk can easily contain zero completed instances of the event that ends them.

What to notice: set the slider to your own answer and read the right-hand column rather than the lanes. Both traders feel experienced. Only one of them has been shown anything.

This inverts the usual hierarchy of respectability, and I think the inversion is correct. The parts of the market that are treated as serious and the parts treated as a casino are ordered almost exactly by how long their bills take to arrive, which is to say by how long a participant can go on believing the bill does not exist. The small cap trader who has been squeezed four times this year is not the least risk-aware person in the room. They are the one who has actually seen the distribution they are drawing from.

There is a related trap I have written about elsewhere. The plateau in The Six Phases of a Trader kills people because the feedback goes quiet while the skill is still rising. This is the same failure of feedback wearing a different hat: a long quiet stretch feels like confirmation and is only ever a small sample.

The same skill, four endings

Everything above is an argument. Here is the argument with the numbers put in.

Four hundred careers per cell. Every one of them runs the same edge: the same hit rate, the same reward-to-risk, the same fraction of the account risked per trade. Nothing about the trader changes between lanes. The only thing that changes is what the instrument does to the losses the stop was supposed to cap, which is to say the standing carry, how often the claim lands, and how many multiples of the intended risk it takes when it does.

What to try: set the hit rate wherever you think yours is, then push the risk slider up and watch how much sooner the unbounded lanes give way than the bounded ones. The "re-run the world" button reseeds everything without touching the parameters.

One caution about the lane that comes out looking best. The panel counts trades, and the long large cap claim is denominated in years, so a six hundred trade horizon flatters that cell in the same way a quiet decade flatters it in real life. It has the mildest bill on the board per unit of trading and the slowest one per unit of calendar, and the second of those is the reason people who have lived in it for twenty years say the confident things they say.

Sizing off the claim

If the map is right, then the standard risk formula has a hole in it, and the hole is not subtle.

Divide your risk budget by the distance to your stop, and you get a position size. That calculation is correct on every day the tape is continuous and the move is bounded. It is also the calculation that produced every account that has ever been ended by a single print, because the number it depends on, the distance to your stop, is precisely the number that stops existing at the moment it matters.

The alternative is to divide the same budget by the distance to the claim, meaning the distance to wherever the instrument is capable of going without asking you first: a reopening gap, a squeeze, or a regime that outlives your patience.

What to try: pick a cell with the pills, then compare the two bars on the right. The left pair is what the discipline costs you every ordinary day, and the number in the bottom left corner is the multiple by which your position shrinks.

This is a real cost and I want to state it plainly rather than sell it. Sizing off the claim will hand you a position several times smaller than everyone around you, on every single ordinary day, for as long as the claim does not arrive. In a good stretch you will watch people with worse reads make more money than you, and the arithmetic that makes that happen is the same arithmetic that keeps you solvent later. Almost nobody pays this price, and the reason is structural rather than stupid: the premium is visible every day and the claim stays invisible until it is the only thing on the screen.

What matters more than the rule is what the rule forces you to go and look up.

The claim distance is a property of the instrument, so you can look it up. What did this thing do the last time it halted and reopened. What is the largest single-session move in this float in the last two years. What does the borrow cost and what has it cost when it got tight. That research is available, it is boring, and it is the only kind of research that speaks to the part of the distribution that ends careers.

On the discontinuous side of the map a stop is an efficiency tool rather than a risk control, and a good one, for the losses that were always going to be ordinary. Position size is the only instrument that works on both sides.

And your worst outcome is knowable in advance in exactly two of the four cells. In the bounded row you can name it. In the unbounded row you can only name a level you have chosen to survive, which is a different kind of statement and should be made out loud, in a number, before entry.

October again

The Volkswagen shorts had done the work. They understood the business, they had the macro right, they were early on a call that paid out within months for anyone who was still there to collect it. What they had not done was ask the second question, which was not about the company at all. It was about how many shares could actually be bought, and by whom, on a Tuesday morning when everyone needed them at once.

I do not think there is a version of that trade with better execution. The premium those short sellers were collecting, week after week, was payment for absorbing exactly the risk that a float can be smaller than the position built on top of it. The trade worked because of that fact and it ended because of that fact, and there was never a moment where those were two separate things.

Your style has a hole in it. You can find it in about ninety seconds by asking whether your loss has a floor and whether price has to come to you. The question worth sitting with is what your position size would have to be for that hole to be survivable, and whether you would still take the trade at that size.

If the answer is no, then the thing on your screen is already the claim, fully written, with only the date left undecided.