How to map a Hedge Fund Strategy at a Structural Level

The book is read before the label. The long book and the short book are established first, then the gross and net exposure, then what the result depends on if the market does nothing, then the instruments and the counterparty, then whether what the fund holds can be sold as fast as it has promised to pay redeeming investors. The label is the last thing to be looked at, not the first.

A reading order is seven questions asked in a fixed sequence, with each answer striking names off a list of approaches under consideration. The sequence does not explain what any of those names means. A procedure that also explains its own subject stops being a procedure and becomes an essay with numbers in it, and an essay cannot be run against a document at four in the afternoon. So the mechanisms live somewhere else, and each approach is covered separately under its own name. The sequence supplies the order, the questions, and the discipline of writing down what was actually found.

Why does the label go last, and what is a label anyway?

A shop on an ordinary street makes the point. The question is what it actually sells. One answer is the board above the door, reading general store. The other comes from standing inside for ten minutes and looking at what is on the shelves and what people are carrying out: four kinds of rice, two kinds of soap, a phone recharge counter, and a fridge at the back that is full of cold drinks and nothing else. Both descriptions are about the same shop. Only the second one shows what would happen to the shop if the price of rice moved, or if the fridge stopped working for a week.

A strategy label is the board above the door. Somebody chose it. The people who chose it had reasons that are perfectly reasonable and are still their reasons rather than the reader's: it is short, it is the phrase the people they talk to already use, and it puts the fund in a group where a reader will know roughly what to expect. None of that makes the label wrong. A label is therefore a summary written by somebody else, produced before the reader arrived, and pointed at an audience that may not have included them.

Two funds carrying the same label can hold entirely different books, and one fund can carry a label that fitted it three years ago and does not fit it now. Neither of those is dishonesty. A label is a name, not a contract, and nothing in a name is obliged to move when a book moves. The consequence is practical rather than moral: a reading that starts from the label has already decided what it is going to find, and what follows will be a search for agreement rather than a description.

So the order in this procedure is inverted on purpose. The description is built first, out of positions and arithmetic and contracted terms, and only when the description is finished does the question arise of which name fits it. The name has to earn its place against something already written down. Building the description first is the whole idea, and the seven steps below are just the disciplined version of it.

Step 1: what is held long, what is held short, and against what net assets?

Three numbers and one question. The three numbers are the size of the long positionsHoldings that gain if the price rises., the size of the short positionsPositions sold without holding the thing sold, which gain if the price falls., and the net assets of the fund those two sit inside. The question is what the unit of analysis is: is this book made of companies, or is it made of markets?

The worked case is Nilgiri Absolute Return Fund, an invented vehicle, and every figure attached to it belongs to the record date. The record date is the end of Nilgiri Growth Partners Fund II's Year 9 Quarter 2. At that date the fund held long positions of Rs 6,50,00,00,000 and short positions of Rs 2,50,00,00,000 against net assets of Rs 5,00,00,00,000. Its positions are shares, and the shares it has sold short are borrowed stock. So the answer to step 1 is that a short book exists, it is not small, and the unit of analysis is a company rather than a market.

Much of the work is done by the phrase "a short book exists". A short book existing is not a small finding dressed up. Half the names on any list of approaches are names for books that hold something short, and the other half are not, so the very first answer written down cuts the list. People forget to look for that answer. A fund with a large long book reads at a glance like an ordinary long book, and the short side sits in a different part of the same document.

The unit of analysis question is the second half of step 1, and it is the one that most often gets skipped. The question feels obvious once the answer is known and is invisible before then. A household deciding whether to buy a second scooter is thinking about one thing it can see and touch. A household deciding whether to move cities because work is drying up in the whole area is thinking about something much larger that contains many things it cannot see. The two decisions use different information, get made on different evidence, and go wrong in different ways, even though both are the same household with the same money. The unit of analysis is the difference between reasoning about one thing and reasoning about the set that thing sits in, and a book is built on one or the other.

On this record the answer is a company. The company answer, on its own, takes two names off the list: number 1, global macro, and number 2, managed futures and systematic trend following. The answer moves those two names, and the reason it does is not needed in order to run the procedure. A reason is needed in order to argue with the result, and that argument belongs with the material that teaches each approach as a mechanism.

Step 2: what do the two exposure figures say that one alone cannot?

Now the arithmetic, and there is very little of it. Gross exposureLong plus short, added without regard to sign, as a share of net assets. is the long book plus the short book, added together without caring which way either points, expressed as a share of net assets. Net exposureLong less short, as a share of net assets. is the long book less the short book, over the same net assets. Two additions, one subtraction, two divisions. Anybody can do it, and almost nobody does both.

Run it on Nilgiri Absolute Return Fund at the record date. Long Rs 6,50,00,00,000 plus short Rs 2,50,00,00,000 is Rs 9,00,00,00,000 of gross exposure, and against net assets of Rs 5,00,00,00,000 that is 180.0 per cent. Long Rs 6,50,00,00,000 less short Rs 2,50,00,00,000 is Rs 4,00,00,00,000 of net exposure, and against the same net assets that is 80.0 per cent. Taken separately: the long book alone is 130.0 per cent of net assets and the short book alone is 50.0 per cent, and 130.0 plus 50.0 is the gross figure while 130.0 less 50.0 is the net figure.

The two percentages describe the same book on the same day and they answer two completely different questions, so a description that gives only one of them has described neither. The gross figure answers how much is going on: how much is bought and sold, how much has to be financed, how much has to be settled, how much moves when anything moves. The net figure answers which way the book points and by how much: what is left over after the long side and the short side have been set against each other.

The everyday version: two people run a vegetable stall. The first buys twenty crates in the morning and sells twenty crates by evening. The second buys a hundred crates in the morning, sells ninety of them to another trader by noon and the remaining ten in the evening. At the end of the day both are holding nothing. Their net position is identical. But one of them handled twenty crates and the other handled a hundred and ninety, and if a crate is dropped, mispriced or refused by a buyer, the second person meets that risk nine and a half times as often. Net says what is left at the end of the day. Gross says how much passed through the stall to get there.

The 80.0 per cent figure is the one that does the narrowing at this step. A book that is net long by four fifths of its own net assets is a long way from a book whose net figure sits at or near nil. Writing that down removes three more names: number 3, equity market neutral, number 4, relative value, and number 7, short selling as a standalone approach. Six of the eight names are now gone, on two additions, one subtraction and two divisions.

Step 3: what does the result depend on if the market does not move at all?

Step 3 turns two percentages into an actual claim about the book. The net figure is already in hand. Taking the general direction of the market away leaves either something or nothing. If the answer is "nothing much", then the direction was the answer, and a great deal is now known. If the answer is "quite a lot", then something other than direction is producing the result, and the next job is to find out what.

Step 3 is the first of three branching steps, and the branch is not optional decoration. The next question genuinely depends on what step 2 produced. If the net figure is near nil, the direction was already taken away by the fund itself, so step 3 asks what is left in its place and there had better be something. If the net figure is large, step 3 asks how much of the answer the direction is, and what the rest of it is. The order of the seven steps never changes, but the previous answer chooses the question inside a branching step, and that is what stops the sequence being a checklist somebody can fill in without reading anything.

On this record the branch is the second one. Net exposure at 80.0 per cent means a general market move reaches most of this book, so the answer is not independent of direction. A net figure of 80.0 per cent does not settle everything: 80.0 per cent is not 100.0 per cent and something is going on with the other side. One thing it does settle is that name 8, arbitrage in its precise sense, comes off the list. Two names are left standing, numbers 5 and 6.

Step 4: is the dependency a price, a spread, a date or an event?

By now the size of the book is roughly known, along with which way it points and whether direction is doing the work. Step 4 asks what kind of thing has to happen for a position in it to resolve. There are four kinds worth separating, and they are separated by what would go in a calendar entry rather than by anything technical. A price: the position resolves when a number reaches somewhere. A spread: it resolves when the gap between two numbers changes, and neither number on its own is the point. A date: it resolves when a day arrives, whatever anything costs on that day. An event: it resolves when a specific thing either happens or does not, and if it does not happen the position is simply still there.

Everyday version, and it is closer than it looks. A household waiting for the price of a scooter to come down before buying is waiting on a price. A trader who has bought onions in one town to sell in another is waiting on the gap between two prices, and does not much mind whether both are high or both are low. A fixed deposit maturing is waiting on a date and on nothing else at all. A shopkeeper who has taken a lease on the promise that a metro station will open two streets away is waiting on an event, and the event either arrives or it does not.

On this record the dependency is a price relationship rather than a date or an event. A price relationship takes name 5, event-driven, off the list. One name is left standing: number 6, long-short equity, and the record fixes it for Nilgiri Absolute Return Fund.

Step 4 has a second half, and this is where the discipline of this procedure gets tested for the first time. Ask whether the decision rule is written in advance or exercised as judgement each time. In plain words: is there a rule that decides, or is there a person who decides? On this record that question has no answer at all, and the correct thing to do with a question that has no answer is to write it down as a question with no answer, not to guess. The material fixes what the fund holds, what its exposures are, who its prime broker is and what it has contracted with its investors. The material does not say who or what pulls the trigger.

Step 5: what instruments carry it, and who is on the other side?

Two questions that are usually asked as one and should not be. The first is what the positions are actually made of. The second is who holds them, lends against them, and can change the terms on which they are held.

For Nilgiri Absolute Return Fund the record answers the second question completely and the first one only partly. The positions are shares held long and shares sold short with borrowed stock. Whether any derivative sits alongside them is not stated anywhere in this record, and that is a real hole in the description rather than a detail. Write it down as a question with no answer.

The counterparty answer is the interesting one, and it is a single sentence with four things inside it. Marudhar Securities Private Limited, invented, is this fund's prime brokerThe counterparty that holds positions, lends stock and cash, and sets margin.. Marudhar holds the positions. Marudhar lends the stock that is sold short. Marudhar lends the cash that funds the long book. And it sets the marginCollateral a counterparty requires against a position.. Four separate roles, one party, and the structural consequence is arithmetic rather than opinion: anything that changes at that party reaches the book by four routes at the same moment rather than one.

The shape is familiar from ordinary life. A household that rents its home from its employer, banks with the employer's tied bank, and has its children in the employer's school is not in a bad position and is not in a good one. The household is in a concentrated one. If anything at all changes at the employer, four things change on the same morning. Nobody needs to be at fault for that to be true, and describing it is not the same as complaining about it.

The distinction between a concentrated arrangement and a bad one is the whole of step 5 for a reader running this procedure. The counterparty is named, the roles are counted, and there it stops. Whether the concentration matters, what it would cost, and how anybody manages it are questions about a mechanism, and the prime broker as a mechanism is covered separately.

Step 6: can what the fund holds be sold as fast as it has promised to pay?

Step 6 is the step everybody skips. The structural reason why is worth naming, and naming it is the only defence against skipping it again. Every other step in this procedure reads one document. Step 6 is the only one that reads two, and puts them side by side. The holdings sit in one place. The promise the fund has made its investors about getting money back sits in another place, written by different people for a different purpose. Nobody is hiding anything. The two halves simply live in separate rooms, and comparing them is a thing somebody has to decide to do.

The liquidity matchWhether what a fund holds can be sold as fast as it has promised to pay. is not a number and it is not a ratio. The match is a comparison, and like all comparisons it needs both sides present.

A household that has promised to pay a school fee at the start of every quarter and holds its savings in a plot of land two districts away shows the shape of it. Nothing is wrong with the plot and nothing is wrong with the promise. The problem, if there is one, lives entirely in the relationship between them, and it cannot be seen by looking at either the plot or the fee schedule on its own.

Run it on this record. Nilgiri Absolute Return Fund deals quarterly on notice, and among its contracted terms are a lock-up, a gate and a side pocket. The terms establish that the promise is not a promise of immediacy, and that the fund has written down in advance several ways in which paying out can be slowed or set aside. Each of those four terms, and the order in which they bite when they are used together, is covered separately.

Now the other half. How fast could this book actually be sold? The record does not say, so the honest output of step 6 is not a match and not a mismatch: half of the comparison is missing, and the missing half is the finding. A missing half may feel like a failure to conclude. The opposite is true. A reader who writes "no mismatch found" here has quietly converted an absent answer into a reassuring one, and that conversion is the single most expensive habit this whole procedure exists to break.

Step 7: what would end the position, and what has the fund contracted to do?

The last question, and it has two halves like step 4 did. Which events could end a position without anybody deciding to end it? And what has the fund written down in advance about what happens then?

On this record the first half has three answers, all of them structural. A recall of the borrowed stock ends a short position, and the decision belongs to the lender rather than to the fund. A change in the margin required against the book changes what the book can carry, and that decision sits with the same party. And a designation into the side pocket moves a holding into a separate class. Every one of those three is an ending that arrives rather than an ending that is chosen, and that is exactly why step 7 asks the question in that order.

The second half is the fund's own contracted answer, and it is the same set of terms step 6 already found: quarterly dealing on notice, with a lock-up, a gate and a side pocket named. The fund's contracted terms are established, and the manager's own rule for closing a position is not established anywhere in this record. The closing rule is the third question this reading has finished with no answer, alongside the decision rule at step 4 and the sale speed at step 6.

What does each answer rule in, and what does it rule out?

The whole narrowing sits in one view. Eight names, numbered as this subject area numbers them: 1 global macro, 2 managed futures and systematic trend following, 3 equity market neutral, 4 relative value, 5 event-driven, 6 long-short equity, 7 short selling as a standalone approach, and 8 arbitrage in its precise sense. Each is covered separately as a mechanism, and none of them is recommended anywhere on this platform.

Step 1's answer, a company rather than a market, removes names 1 and 2. Step 2's answer, net exposure of 80.0 per cent, removes names 3, 4 and 7. Step 3's answer, that the result is not independent of direction, removes name 8. Step 4's answer, a price relationship rather than a date or an event, removes name 5. Steps 5, 6 and 7 remove nothing. Describing rather than narrowing is their job, and a procedure that pretends every step eliminates something will start eliminating things it has no grounds to.

One name is left, number 6, long-short equity, and the record fixes it for this invented fund. The list ended at exactly one name here because this particular record answered every question that eliminates, and a thinner record would have stopped at two or three names with nothing wrong with it. The stopping point is what is most worth carrying away from the table below. The procedure is not a machine for producing a single label. The procedure is a machine for producing the shortest honest list.

Why three marks and not two?

Every name, at every step, carries exactly one of three marks, and the third one is the reason this procedure is worth running rather than guessing.

RULED OUT means an answer exists, and that answer is inconsistent with this name. A ruling out is a conclusion, it rests on something written down, and the line it rests on can be shown to somebody.

STILL OPEN means nothing found so far removes this name. Staying open is not agreement and it is not evidence in the name's favour. An open name is simply one that nothing has yet given a reason to strike.

NOT ESTABLISHED means nothing in the material supplied answers the question. Not a small answer, not an ambiguous answer, no answer. Collapsing "the record does not say" into "no" is how a reader turns missing information into a conclusion, and refusing that collapse is the whole discipline of this reading. Two marks are not enough because two marks force every unanswered question into one side or the other, and the side it gets forced into is always the comfortable one.

The reading of this record finished with three questions marked not established: whether the decision rule is written in advance or exercised each time, whether any derivative carries part of the book, and what the manager's own rule for closing a position is. Each of those goes into the output as a question, next to the seven lines that do have answers.

The third mark buys something in practice. Suppose step 1 had come back with a market rather than a company. Names 1 and 2 would both have stayed open, separated only by the decision rule, and the decision rule is exactly what this record does not answer. The reading would have finished with two names standing and a mark beside them, and that would have been a completed reading rather than a failed one. A reader with two marks instead of three would have picked one of the two names, and would have picked it on nothing.

How does somebody actually use this on a Tuesday afternoon?

An analyst covering a set of funds is not usually asked whether a fund is any good. The job does not look like that from inside. The job looks like this: eleven documents arrived, three of them describe themselves in the same two words, and somebody upstairs needs to know by Thursday whether those three are actually the same kind of thing or whether the shared phrase is a coincidence. The question is one of classification, and classification done badly is worse than no classification at all: it produces groups that then get compared to each other.

Run against eleven documents, this procedure does three useful things and one that matters more than the other three. First, it produces the same seven lines for every fund, so the eleven descriptions are comparable in a way that eleven summaries written in eleven house styles never are. Second, it makes the exposure arithmetic mandatory, so nobody ends up with a gross figure for one fund and a net figure for another and then puts them in the same column. Third, it forces the liquidity comparison to be attempted, so the second document actually gets opened.

The fourth thing is the one worth the whole exercise: it produces a written list of what each document does not say, and that list is what an analyst actually sends back. A reader who marks three questions not established on this fund has produced three specific things to ask. Each one names the exact gap it fills, and that makes it a far better question than a general request for more about the strategy. Naming the gaps is what a structural descriptionA statement of what a book is, with no claim about what it will do. is for. The description states what is there, marks what is not, and asks nothing of the future.

The same shape works for somebody who is not an analyst at all. A person sitting in a room where a fund is being described can ask, quietly and in order, what is long, what is short, against what net assets, and what the two exposure figures are. The four questions are answerable from any honest description in about a minute, and asking them in that order rather than asking for the strategy label changes what comes back.

When is the reading finished, and what has not been concluded?

The reading is finished when the seven lines are written, the questions with no answer are marked as questions with no answer, and the name has been placed against the description rather than in front of it. The stopping condition is a real one: there is no eighth step, no scoring, no summary judgement, and no point at which the description turns into a view.

For Nilgiri Absolute Return Fund at the record date, the finished reading is a directional equity book, run at 180.0 per cent gross and 80.0 per cent net against net assets of Rs 5,00,00,00,000, carried in shares long and shares borrowed short, through one prime broker holding four roles, dealing quarterly on notice with a lock-up, a gate and a side pocket named. Three questions came back with no answer. The name that fits that description is number 6 on the list, long-short equity, and the record fixes it for this fund.

The procedure refuses to produce a verdict, and the refusal is not politeness or hedging: it is the boundary of what the seven questions can support. Nothing in a description of positions, exposures, instruments and dealing terms says whether an approach works. Nothing in it says whether the approach suits anybody, whose money it should be near, or what it might return. Nothing in it settles whether the concentration at the prime broker is a problem, and that judgement is one about somebody else's arrangements made from outside them. All of them are real questions. Not one of them is a question about the book, and this procedure only reads books.