What a Derivatives Post-Mortem Does for the Next One

A reconstruction is written for the decision after it, not the one it describes. A reconstruction records what was in front of the person taking the decision, the assumptions the reasoning leaned on, and which of the four checks were actually run. Written often enough, the leaning assumptions start repeating, and that repetition is the only thing anybody carries forward. No result is needed at any point.

Nothing new is defined from here on. The forward price, the notional, the payoff and the margin were all settled in earlier work and are used below rather than explained. The new thing is a habit, and the habit is the last thing this subject has to hand over.

Here is the situation it is built for. An analyst takes a position this morning. By Thursday the reason is remembered roughly. By next month what is remembered is a story about why. And by the time a position shaped like that one turns up again, almost nothing is left that would actually help. A post-mortemA reconstruction of a decision as it stood at the moment it was taken, written from what was in front of the person taking it. is what gets written against that drift. The reconstruction sets the decision down as it stood when it was taken: what was on the screen, what was assumed without stopping to argue about it, and which of the four checks were run rather than intended to run.

One of those is worth a little to the person who wrote it. A hundred of them are worth something else entirely, and the difference is not that the hundredth is better written. A single reconstruction is a record of one Tuesday. A long run of them is a list of the assumptions the analyst keeps making without noticing, and a list is the only thing anybody actually carries from one decision to the next. Everything below is about what accumulates.

What does a post-mortem do across a hundred decisions that it cannot do across one?

One reconstruction gives less than the practice is usually sold on, and the honest version is worth stating first. One reconstruction records what happened inside one person's head on one particular morning. A record of one morning is useful to have and it is not a skill. Nothing about writing it makes anybody better at reading a screen the next day, and if the practice stopped at one, it would be a diary with arithmetic in it.

Run it thirty times and something appears that was not available at any point along the way. On the thirtieth reconstruction the assumption the decision leaned on gets written down, and it is recognised. Almost the same sentence was written on the fourth and has not been thought about since. The recognition is the whole product. Thinking harder on the thirtieth morning would not have produced it. Writing the fourth morning down in a form that could still be read did.

Here is the everyday version, and it is worth holding on to because it is the same mechanism at a smaller scale. A cook writes down the oven setting, the resting time and what the dough felt like, every single time, for four months. Somewhere in the fourth month she notices that every disappointing batch came out of a Thursday, and Thursday is the day the flour arrives from the other shop. No amount of remembering would have produced that. The remembering was working perfectly. The problem is that memory does not sort. A written record of reasoning can be searched and a memory of reasoning cannot, and that single difference is the whole of the advantage this practice has.

The modest version of the claim is the true one and the grand version does damage, so the claim stays modest. The practice does not make anybody a better judge of a contract, a price or a moment. The practice does not sharpen instinct. It makes a searchable record of reasoning, and a searchable record is a different kind of object from a sharper instinct: it can be handed to somebody else, it can be disagreed with, and it can yield the sentence written nine months ago and entirely forgotten since.

What sits on the standing agenda of every reconstruction?

Four questions, in the same order, whatever the position was. Four questions are the whole of it, and the shortness is not a compromise. The standing agendaThe fixed list of questions that every reconstruction answers, kept the same from one write-up to the next so it never has to be invented. is the four checks already established, written down as the fixed contents of a document rather than as things to remember at the right moment. Nothing on it is new. The one new thing is that the four are printed rather than recalled.

The shape matters more than any one of them, so read the four before the arithmetic. Was the forward price treated as arithmetic. Was the headline quantity divided by what actually moved. Did every figure say what kind of figure it was. Was every move restated against what was posted. Four questions, and a reconstruction is finished when all four have an answer written beside them and not before.

A fixed agenda beats a thoughtful one, and the reason is nothing to do with the quality of the questions. A thoughtful agenda has to be invented each time, and inventing it is work, and work that has to be done before starting is work that gets skipped on the third bad morning in a row. A list that does not have to be invented is a list still running in eighteen months. The four below are not the four best questions anybody could ask about a position. The four are questions that can be answered without deciding anything first.

The first line: was the forward price multiplied, or was it read?

A reference asset, invented for teaching, has a spot price of Rs 2,000.00/-. Financing costs 6.50 per cent a year. The asset pays nothing at all while it is held. Paying nothing is the fact that matters most here, and it comes back twice below. Multiply: Rs 2,000.00/- by 6.50 per cent a year for one year adds Rs 130.00/-, and the forward price is Rs 2,130.00/-. The carryThe cost of holding the underlying from today until the agreed date, which here is financing and nothing else because the asset pays out nothing along the way. of Rs 130.00/- accounts for the entire distance between the two numbers, and nothing is left over.

Now the question the reconstruction actually asks, and it is narrower than it looks. The question is not what the forward price was. The question is whether that multiplication was performed at the time, or whether the number was read off a screen and written down. The two answers produce the same figure in the write-up and completely different states of mind behind it. A reconstruction that records the forward price of Rs 2,130.00/- has recorded a number. One that records the carry of Rs 130.00/- beside it has recorded that the multiplication was done.

Why the distinction is worth a whole line of a fixed agenda: a forward price that has been multiplied cannot be mistaken for an opinion about direction, and a forward price that has only been read very easily can. Once Rs 130.00/- is written next to it, there is no room left in the figure for a view. The whole gap is spoken for.

The second line: was the headline divided by whatever actually moved?

The arrangement has two sides. One pays a fixed 7.20 per cent a year and receives a floating benchmark. The other does the reverse. The notionalThe amount that the payments under an arrangement are multiplied by. The amount is a multiplier and it does not change hands. is Rs 1,000 crore, and the floating benchmark reads 6.00 per cent a year for the first period. A floating benchmark is only known one period at a time, so the first period is the only one that can be worked.

Multiply both legs out. The fixed leg over one full first period is Rs 72.00 crore gross. The floating leg is Rs 60.00 crore gross. The two differ by a net of Rs 12.00 crore. The fixed payer hands that net to the floating payer, and the net is the only money that moves. The two quotients say different things, so divide now in both directions. Rs 12.00 crore into Rs 1,000 crore is 1.2 per cent, so 98.8 per cent of the headline stays exactly where it was. And Rs 1,000 crore divided by Rs 12.00 crore is two hundred and fifty over three, rounding to 83.33 times.

A figure nobody has checked is exactly what the agenda exists to catch, so the rounding is worth one careful sentence. Two hundred and fifty over three does not terminate. Multiplying the rounded 83.33 back by Rs 12.00 crore gives Rs 999.96 crore, and that falls Rs 4,00,000/- short of the headline. So the notional is about 83.33 times the first period net rather than exactly that, and a reconstruction writes about.

The second line of the agenda catches a multiplier being written up as an amount at risk. Writing a multiplier up that way is the single most effective way to make an arrangement look nearly two orders of magnitude bigger than the cash it produces. The check is one division and it takes four seconds. Which sum would actually change hands in the period, and what is the headline divided by that sum.

The third line: did every figure say which kind of figure it was?

Take the option pair. A call is struck at Rs 2,000.00/- and cost a premium of Rs 180.00/- to buy. Put the price at Rs 2,130.00/- and ask what the position delivers. The payoffWhat a contract delivers at expiry, ignoring altogether what was paid to get into it. is Rs 2,130.00/- less the strike of Rs 2,000.00/-, or Rs 130.00/-. The profitThe payoff net of what was paid for the position, which is a different number and usually a smaller one. is that Rs 130.00/- less the Rs 180.00/- paid, or minus Rs 50.00/-. The profit does not reach nil until the price gets to Rs 2,180.00/-.

Two numbers, one position, one price, and they differ by exactly what was paid. Neither is wrong. Writing one down under the heading of the other would be wrong, and the whole of the third line of the agenda is a question about column headings rather than about arithmetic. Does this number know what was paid for the position. Rs 130.00/- does not know. Rs 50.00/- does.

The cost of getting the third line wrong is precise rather than vague: it writes a gain of Rs 130.00/- into the record of a position that was at that moment down Rs 50.00/-. Nobody has lied. Somebody put a payoff in a column marked profit, and the record now says the opposite of what happened.

The fourth line: was the move restated against whatever had been posted?

An initial margin of 8.0 per cent of the exposure puts Rs 160.00/- against Rs 2,000.00/- of exposure. No authority publishes an initial margin of 8.0 per cent, and no rule requires that figure. The 8.0 per cent stands as given so the arithmetic can be worked, and who sets what is actually posted appears below.

Now move the referenced price 4.0 per cent the wrong way. On Rs 2,000.00/- of exposure that is Rs 80.00/-. Restate the same Rs 80.00/- against the Rs 160.00/- posted and it is 50.0 per cent of it. And the two figures together make the exposure 12.50 times what was posted. Three sentences, one movement, and they sound nothing alike: a small move, half of everything, and twelve and a half times.

A ratio without its base is not yet a number, and the fourth line of the agenda exists to make the base appear in the same sentence as the figure rather than in the paragraph above it. Every one of the three sentences in the paragraph before this one is true. Only the second one conveys what a bad afternoon feels like from the inside.

What does the agenda look like once it has been run all the way through?

Here is the whole of it on one document, run once, on the positions used throughout this guide. The last row is worth reading before any of the others. Most reviews would refuse to print it.

The blank result row is not a gap in the document. Leaving the result out is a claim about what a reconstruction needs, and the claim is defended below.

What actually changes in somebody who has run the agenda a hundred times?

The tempting answer is judgement, and it should be refused. Judgement is not observable. Nobody can check it, including the person claiming to have it, and a claim that the practice improves judgement would be a claim no reader could ever verify and no reader could ever be argued out of. The narrower and duller claim is the one that can actually be looked at.

The order in which four ordinary steps get done is what changes. Nothing else changes, and every one of the four steps is visible in the write-up somebody produces. Such a writer names the base before writing the ratio, so a percentage never arrives without the quantity it was struck out of. The writer asks what actually moves before quoting a headline, so a multiplier never gets reported as a size. The writer labels the column before filling it, so a payoff never turns into a profit while it is being copied across. And they carry the price at the rate before they say anything at all about the forward, so a financing cost never gets read out as a view.

The four share the least glamorous property available, and that property is the reason the practice works. Not one of the four is an insight, and every one of them is an ordinary step done in a particular order. Nobody has become cleverer. Somebody has swapped two steps around, four times, and now writes sentences that cannot carry a certain kind of error because the error would have had to be introduced before the sentence began.

Here is the same thing outside finance. A joiner who measures twice and cuts once is not a better joiner in any way that could be tested. He has one ordering, and the ordering removes a category of ruined timber from his life entirely. Asked what improved, the honest answer is that nothing improved: a step moved.

Why does the practice run when there is no result anywhere in it?

The empty result line is worth naming plainly before it is used. A reconstruction does not record what the referenced price did next, whether the call finished above Rs 2,180.00/-, or what the floating benchmark read in the second period, and it attaches no likelihood to any price. Every reconstruction above was written with that line empty, and every one of them was finished.

The empty line is not a limitation being apologised for. Emptiness is the property that makes the practice available at all, and here is why. A review that waits for a result can only be written after the result arrives, and that means late. Such a review gets written on the positions somebody still remembers, and those are the dramatic ones. And by the time it is written, the reasoning it is supposed to reconstruct has been quietly rewritten by knowing how things went. A review that needs only the reasoning can be written on the afternoon of the decision, and it can be written on every single position rather than on the ones that were memorable.

There is a second reason, and it is sharper. Judging a decision by how it turned out is called resultingJudging a decision by how it turned out rather than by what was known and reasoned at the time it was taken., a term Duke sets out in Thinking in Bets, 2018. A decision taken on thin reasoning that happens to work out gets filed as a good one. A decision taken carefully that happens to go badly gets filed as a mistake, and the careful reasoning inside it is thrown out with the outcome. So a review that waits for the result is not neutral while it waits. The waiting imports exactly the error the review exists to prevent, and it imports that error into the only record anybody will ever have of what was actually thought.

The everyday version is a driving one. Somebody checks the mirror, indicates, and pulls out safely. Somebody else does none of that and also arrives home fine. Graded by the outcome, the two evenings would have to be called equal, and anybody watching would say that is nonsense. The reasoning was different and the reasoning was the part that was under anybody's control.

So the result line stays empty on purpose, and it stays labelled rather than deleted. A blank labelled result is a document saying it did not need one. A missing result line is a document that forgot.

What does a reconstruction record that a later one can actually use?

The difference between a file somebody reopens and a file nobody ever does comes down to the last line rather than to anything above it. The property that makes a written line useful later is transferThe property of a recorded line that lets a later decision use it. The line has to be about this decision rather than about decisions in general.: a line transfers when a later reconstruction can recognise it, disagree with it, or find it false.

A line that transfers has two parts and takes one sentence. The line names a load-bearing assumptionAn assumption the decision would have changed without. Take it away and the reasoning has to be redone rather than merely adjusted., and then it says what would have changed without it. The two parts are the entire form, and writing them out on the positions used above makes the shape concrete rather than described.

The assumption everything above rests on is that the reference asset pays nothing at all while it is held. Now the second half. Had it paid something out, that payout would have come off the carry of Rs 130.00/-, the forward price would have been lower than Rs 2,130.00/-, and with a large enough payout the forward price could have sat below the spot price of Rs 2,000.00/- altogether. A forward price below the spot price reverses the shape of everything downstream of it. The sentence is worth keeping because a later position on an asset that does pay something out will collide with it directly.

Now the test, and it is one question that can be asked of any closing line ever written. Would this line read exactly the same after any position on any asset? If it would, it has recorded nothing about this one. A line that fits every decision equally well has described no decision at all, and the whole of the difference between a useful file and a dead one is whether its closing lines pass that single question.

Two examples, and read them slowly because they look far more similar than they are. The assumptions should have been examined more carefully. Against: the asset paying nothing while it is held was the assumption everything rested on. The first would sit unchanged at the foot of a write-up about a currency arrangement, a property purchase or a decision to change jobs. The second could not be moved anywhere without being rewritten, and that immovability is exactly what makes it worth keeping.

Where does the practice go wrong once it has become routine?

Everything above is about a practice working. The same practice eight months later is another matter, and it is the one failure that cannot show up early. Nothing goes wrong on day one. Decay is slow, and it happens in a document that still looks completely correct.

The agenda turns into a template, and a template gets filled the same way every time. All four lines still get answered. The answers get shorter. One assumption sits in the front of the writer's mind, so the closing line names it again, and then names it again, and by the twentieth consecutive position nobody has noticed that the four questions have stopped catching anything at all. The document is still being produced. The document has quietly stopped being a reconstruction and become a habit with a form attached.

So here is the diagnostic. Staying alert has already failed by this point, so the diagnostic is something to do on a particular afternoon rather than something to watch for. The last ten closing lines get read together, in one sitting, one after the other, with nothing else consulted. Two minutes. If those ten lines name ten different assumptions, the agenda is still doing its work. If they are the same line ten times over, the write-ups are recording habits rather than decisions.

One point looks like a contradiction of the opening claim, and it has to be said carefully. Recognising at entry thirty the same assumption written at entry four is the value of the practice. Writing the same assumption ten times in a row is its decay. The two are not in tension, and the difference is not subtle once it is seen: recognition happens across a run in which the lines mostly differ, so one repeat stands out. Decay is when nothing differs, so there is nothing for a repeat to stand out against. A repeated line is only information when the surrounding lines are varied.

Now the repair, and it is worth being exact about because the instinct is wrong. The instinct says try harder: ask the four questions more carefully, sit with them longer, be more honest. Trying harder does nothing. Effort was never what failed. The four questions were answered perfectly on every one of those twenty positions. A fixed list is a strength right up until it becomes the only question anybody puts, so change what the agenda asks rather than how hard it is asked.

In practice that means adding a fifth line for a while, and it can be almost anything as long as it is not one of the four. Which fact would I have needed for the sentence I wrote to be a statement rather than a guess. Which part of the position did I not look at. Which figure did I take on trust. Retire it when it goes quiet, and put a different one in. The four stay; the fifth rotates.

What does somebody who will never write a review do with any of this?

Most people who could use the agenda will never publish a line and have no file to keep. The agenda still applies. The four questions are questions about writing, and everybody reads. Here is what the practice looks like for four people who are not writing anything at all.

The household case matters most and gets written about least, so take it first. A message arrives from somebody persuasive, about an arrangement with a headline of Rs 1,000 crore, and it sounds enormous because it is meant to. Line two of the agenda, run on a message rather than on a position, asks what would actually change hands in the period, and the answer here is Rs 12.00 crore, or 1.2 per cent of the headline. The household has not learned anything about the arrangement. The household has learned that the number in the first sentence was a multiplier, and it now knows to ask what the second number is before feeling anything about the first.

A lender looking at a borrower who has open positions runs line four and nothing else. The exposure gets volunteered. The posted amount counts more, along with what a move does to it. Rs 80.00/- against Rs 2,000.00/- of exposure sounds like nothing, and the same Rs 80.00/- against Rs 160.00/- posted is half of it. The lender is not judging the position. The lender is working out how quickly the borrower could be asked for cash it does not have.

An analyst uses all four as a filter on the writing rather than on the position, and that is a different job from either of the two above. A note cannot answer whether the position is a good one. A note can show whether it is describing something or arguing for something. A note that carries the carry of Rs 130.00/- beside the forward price of Rs 2,130.00/- has done its arithmetic in the open. A note that carries only the forward price may have done it and has not shown it, and the analyst now knows which questions to send back.

And somebody who has already taken a position, and is looking at a screen full of numbers about it, runs line three hardest. Which of these numbers knows what I paid. Every one of the four questions works just as well on somebody else's writing as on one's own, and the agenda is therefore worth having even to a reader who never keeps a file at all.

What does none of this do?

One claim is more likely to be misread than anything else above, and it belongs at the close of the subject.

A reconstruction improves the account that can be given of a decision. A better account is the entire claim, and it is worth taking seriously. A better account can be handed to somebody, checked, argued with and searched. Here is what a reconstruction is not. Suiting a person is a fact about that person, and no document about a contract contains one, so a reconstruction cannot say whether the decision suited the person who took it. A reconstruction does not change what any position obliges, at any price, by any amount. The Rs 80.00/- still moves. The profit at a price of Rs 2,130.00/- is still minus Rs 50.00/-. And running the agenda a hundred times does not make any contract a different thing to hold than it was before.

A method taught in writing sits one clause away from breaking its own rule here, so the last point deserves its own sentence. Running a check catches a misreading. A misreading was never part of what the position obliged in the first place, so nothing about catching it reduces what is owed, changes what is posted, or makes anything safer. A reader who runs all four questions is better at reading and no better protected, and those are two entirely different things that careful writing must never let merge.

And one more thing belongs here, at the widest point of the subject, rather than tucked into a note. Somebody who lost money on one of these contracts did not necessarily fail at any of this. A well reconstructed decision and a decision that went well are different objects, this subject has said so throughout, and it would be a poor close to imply in the last few paragraphs that anybody harmed had simply not been careful enough. Care is not a shield. Care is a way of knowing what was done.

Who sets the requirements no written account can fill in?

Four requirements are touched by everything written above. The authority for each is named here and the value for none of them, and the sheet below is drawn with every value box empty.

Each row is drawn empty for a reason that sits in the reader's interest. Each of those four is set by the authority named inside its row. Each of them changes. A printed value would not be merely out of date on the day it moved; it would be wrong, presented with the confidence of a printed figure, and nothing in the text would distinguish the two.

The one percentage printed in this guide belongs in this block too, and it is not in the sheet because it is not a requirement. The initial margin of 8.0 per cent of the exposure that produced Rs 160.00/- against Rs 2,000.00/- is assumed. No authority set it. Nobody publishes it. The figure was chosen so that the arithmetic in the fourth line of the agenda could be worked through rather than merely described, and it carries that label every single time it is used. The amount actually posted against a position, and how that amount is arrived at, sits with SEBI at sebi.gov.in and with the clearing corporations working under its framework, it differs by contract and by day, and it moves.

After all of this, should a reader be inside one of these contracts?

The question arrives here, at the end of everything, and it gets answered here rather than in small type at the foot. The question is a reasonable one and it has been sitting under the whole of this subject.

The answer is that this platform does not answer it. Everything taught across these contracts describes what an arrangement obliges: what each shape commits each side to, where a forward price comes from, what a premium is and is not, what a notional multiplies, what is posted and what stands behind it. Not one line of any of it says whether a particular person belongs inside one.

Naming what somebody would have to know before that question could even be attempted is the useful part, and the list explains why no written account could hold the answer. Anybody answering would need to know what the reader already holds. Only then can the position be seen against everything else in that reader's life rather than on its own. They would need to know what a wrong move would take from the reader, and that is a fact about the reader's circumstances rather than about the contract. They would need a likelihood attached to every price rather than a set of prices, and no contract specification carries a likelihood. And they would need to know what SEBI at sebi.gov.in permits that reader in particular to do.

So three of the four sit outside anything that could be printed for a stranger, and the fourth has an address rather than a value. The shape of the gap is the actual reason the question cannot be answered here, and the three missing facts can be gathered by whoever wants them.

Understanding a mechanism completely is not a reason to use it, and a reader who leaves holding that one sentence has taken the most useful thing on offer here.

The reference asset, the option pair and the two-legged arrangement are invented.
Educational material. Not advice on any investment, tax, budget or market position.