Option Delta: What Can Be Approximated, and What Cannot

Delta measures how much an option premium moves when the reference asset moves by one rupee. Producing a delta for either contract below needs an estimate of how far the reference asset might yet travel, and the working example holds no estimate of that kind. Two things survive that absence: the bounds each contract imposes, and that the call delta less the put delta, at one strike and one end date, is one.

Delta is a slopeThe rate at which one quantity changes as another changes, measured at a point., and a slope is a localTrue near where the measurement was taken and not away from it. statement: it says what happens to the premium for a small move from where the reference asset stands, and nothing about a large one. Everything useful about delta and everything misleading about it follow from that: it can be multiplied by a quantity, added across contracts, and it stops being true the moment the move stops being small. The instrument below computes every part of that which these figures honestly support, and prints a refusal where the rest would have gone.

The pair of contracts it opens on has been running through this sequence already. A reference asset priced at Rs 2,000.00/-. A call and a put, both struck at Rs 2,000.00/-, both running to one date a year out. Financing at 6.50 per cent a year. The reference asset pays nothing while it is held. A payout would change every carried figure below. The call premium of Rs 180.00/- is given by the working record; the put premium of Rs 57.93/- is derived from it by the parity relationshipThe fixed link between a call, a put, the reference asset and the strike at one end date. Built and proved separately., and deriving it is arithmetic on a rate. Given, derived, refused: three different statuses, and keeping them apart is the whole discipline.

On those figures it returns a present value of the strike of Rs 1,877.9343/-, a parity difference of Rs 122.0657/-, and a put premium of Rs 57.9343/-, carried through this sequence as Rs 57.93/-. The instrument then reads a slope at the end date at Rs 2,200.00/-, where a bought call is worth Rs 200.00/- and the slope is one, and asks it to cover a fall of 400 rupees to Rs 1,800.00/-. The estimate is minus Rs 200.00/- against a true value of nil. So it understates the position by Rs 200.00/- on every unit, prints to two places while doing so, and warns nobody.

What is delta meant to measure?

Ask it in the plainest form. The reference asset costs Rs 2,000.00/- today. Suppose it goes to Rs 2,001.00/- and nothing else in the world changes at the same moment. How much of that one rupee does the call premium pick up? The share the premium picks up is the call delta, and the call delta is not an amount of money anybody has: it is an exchange rate between two movements, rupees of premium per rupee of reference asset.

The phrase doing the heavy lifting is nothing else changes at the same moment. The strike, the end date, the financing rate and whatever anybody believes about travel all stay exactly as they were. A delta read in a world where two of those moved at once is not a delta at all, and most of the confusion readers carry into this subject comes from watching two things move and attributing all of it to one.

The everyday version: a small tiffin service buys rice wholesale. A one rupee rise in the rice rate adds exactly the kilos bought to the monthly bill, with the tiffins, the recipe and the customers all held still. At twenty rupees the same question stops being clean. By then the recipe changes and customers are lost. Delta fails in exactly that way.

Now the bounds, the only numerical statement about delta available without asking anybody for anything. A call delta sits between nil and one and a put delta between minus one and nil, and both bounds come from the contracts rather than from any calculation. A rupee on the reference asset cannot make the right to buy at Rs 2,000.00/- less attractive, so the call delta cannot fall below nil; and that right can never be worth more than the reference asset itself, so a rupee cannot add more than a rupee. The put runs the mirror argument to minus one and nil.

Why can no delta for the contracts in this example be produced here?

No delta for the call at Rs 180.00/- or the put at Rs 57.93/- can be produced here, today, with a year still to run, and the reason is one missing input rather than an oversight.

Naming that input in full is worth the trouble. Half the readers who think they know what is missing name the wrong thing. The missing input is how far the reference asset might travel between now and the end date, and how likely each of those journeys is: not what it did last year and not what somebody expects, but the range of movement still available to it with the relative weight of each part. The working record behind this sequence carries no figure of that kind anywhere.

Everything else is present. The price and the strike at Rs 2,000.00/-, one year remaining, financing at 6.50 per cent, a reference asset that pays nothing while it is held, and both premiums. Five of the six inputs are present, and five is still not enough. The sixth is not a decoration on the calculation. The travel estimate is the calculation. Draw only what is honestly known and the picture is a single point: Rs 2,000.00/- against Rs 180.00/-. A point has no steepness, and the fan of lines through it is narrowed by the contracts only as far as the bounds allow.

The reason for refusing rather than picking a plausible line inside the wedge is worth stating outright: a delta produced from an invented travel estimate would look exactly as authoritative on the screen as a correct one. It would carry two decimal places, multiply cleanly by a quantity, and sit in a report beside figures that were checked. A refusal is visible and an invention is not.

Which delta statement holds whatever anybody assumes?

One statement about these two deltas is exact whatever figure anybody eventually supplies for the missing input. At one strike and one end date, the call delta less the put delta is one. The locked distance comes from the parity relationship already established here, plus one observation about which of its terms move.

Check that arithmetic rather than taking it. The present value of the strike is Rs 2,000.00/- divided by 1.065, or Rs 1,877.9343/-, and taking it off Rs 2,000.00/- leaves Rs 122.0657/-. On the other side, Rs 180.00/- less the rounded Rs 57.93/- comes to Rs 122.07/-. The two are not identical. Rounding the put premium to two places for the rest of the sequence stands them Rs 0.0043/- apart. The relationship holds to the paisa and is never called exact. A reader taught to accept an equality the printed figures do not produce has been taught to stop checking.

Now move the price by one rupee and read the right hand side term by term. The strike is written into the contract and the rate and the date are held still, so the present value of the strike does not move: Rs 1,877.9343/- before and after. So the whole change on the right is that one rupee, the left hand side must move by one rupee too, and the left hand side is the call premium less the put premium. The call delta less the put delta is therefore one.

The missing input enters both deltas, so changing the estimate of travel changes both, but they move in lockstep and the distance between them never budges. That distance is the one delta statement here no assumption can disturb, and every slope the instrument returns is checked against it rather than against either figure alone.

Is there any moment when a delta is known outright?

There is exactly one: the last day. At the end the contract is about to settle and the premium simply is the payoff, not an estimate of anything. That payoff line has been drawn twice already in this sequence, and the slope of a line already drawn is arithmetic on a diagram rather than a model result. The end date slope is the one slope the instrument above will compute.

Read the four slopes straight off the drawing below: above Rs 2,000.00/- they are one on the call and nil on the put, and below it nil on the call and minus one on the put. Then test the locked relationship on both arms without leaving the diagram. One less nil is one, and nil less minus one is one. The relationship has just been checked against a drawing rather than against the instrument that uses it. Checking an instrument only against itself is checking it against nothing, and the drawing therefore comes first.

Then the honest gap, said in the same breath rather than in a footnote. At the strike itself, exactly at Rs 2,000.00/-, neither slope is defined at all. The payoff line has a cornerA point where two straight arms of a line meet at an angle, so no single steepness belongs to the point. there: from the left the slope reads nil, from the right one, with no reason to prefer either. The corner is a point where the question has no answer rather than a rounding problem, and the working example sits at exactly that price. Set to read a slope at Rs 2,000.00/-, the instrument refuses both figures rather than choosing.

Does what was paid for the contract change the steepness of the line?

No, and this is the mistake a careful reader makes rather than a careless one. Having been told earlier that the payoff is one thing and the profit another, a reader reaches a slope and reasonably asks which of the two it belongs to. For the call, the premium of Rs 180.00/- was paid at the start. Carried at 6.50 per cent for the year, it becomes Rs 191.70/- and sits beside an amount arriving at the end. The profit at any price at the end is the payoff there less Rs 191.70/-. Subtracting the same fixed amount at every single price moves the whole line down without tilting any part of it, so the profit line and the payoff line have the same steepness everywhere.

One steepness for both lines is why no figure below is a profit, and why the instrument above works on values. A reader who carries the premium into a slope adds an amount to a rate of change, adding rupees to rupees per rupee. The premium shows up in one place only: the break even at Rs 2,191.70/-, the strike plus the carried premium, and a break even is a place on the horizontal scale rather than a steepness.

The put side carries one extra piece of care. The unrounded put premium of Rs 57.9343/- carried forward at 6.50 per cent lands on Rs 61.70/- with no remainder. Watch it land rather than trust it. Parity makes the put premium the call premium, less the price of the reference asset, plus the present value of the strike. Carried forward a year, those three terms become Rs 191.70/-, less Rs 2,130.00/-, plus Rs 2,000.00/-, so Rs 130.00/- comes off Rs 191.70/- and Rs 61.70/- is left. The rounded Rs 57.93/- carried forward instead gives Rs 61.69545/-. That gap is small enough to ignore and large enough to identify which of the two figures was carried, and it was the unrounded one.

What does this tool compute, and where does each figure come from?

Every field in the instrument carries one line saying where its figure is found, and none says what it means, because a field note that starts explaining is a lesson wearing a form. The card below sorts the instrument into four groups, and the last is worth reading twice.

Everything the instrument returns is either arithmetic on a rate, as parity is, or the slope of a line already drawn, as the payoff at the end date is. Nothing it returns comes out of a pricing model, and nothing it returns is a delta for these contracts today. The refusal is the design rather than a disclaimer bolted to the bottom of a calculator.

Which leaves the question the card cannot answer: if the slopes are readable only at the end date, what is the instrument for before then? For the two things that hold at every moment. Parity fixes the put premium against the call premium, and the distance between the two deltas stays one. Entering a strike, a rate and a premium returns both to the paisa while the slope stays refused.

What happens to the put delta when the call delta moves?

The instrument at the top proves the locked distance at every reading price. The control below proves it the other way round: supply a call delta and watch the bar between the two markers refuse to change height.

The control opens at the one setting this example can justify: a call delta of 1.00, the slope of the call payoff above the strike on the last day, returning a put delta of 0.00 that can be checked against the flat upper arm of the put beside it. The other end is worth running too: nil returns minus 1.00, and the put payoff below the strike does fall a rupee for every rupee. Where to leave the control depends on how far the reference asset might travel, and that estimate is missing.

What does a delta convert a holding of contracts into?

One practical use drives almost every delta anybody computes. A delta turns a holding of contracts into an equivalent quantity of the reference asset, for a small move and only for a small move. A contract picking up some fraction of every rupee behaves, over a small move, like that fraction of a unit. Multiplied by the contracts held and by what one contract covers, the delta gives a quantity in units rather than in contracts. Units are the entire point: delta equivalentThe quantity of the reference asset a contract currently behaves like over a small move. figures add together and contracts of different kinds do not.

Three quantities get mixed up at exactly this point. The premium of Rs 180.00/- is money that has already moved: the buyer paid it, the writer has it, it is gone. The Rs 2,000.00/- of reference asset the contract is written on is exposureThe value of the reference asset a contract is written on. Nobody has paid it and nobody has received it; it is a size.. The delta equivalent is a third quantity again, the fraction of that exposure the position behaves like. All three are different numbers about the same contract, and a report naming only one has told the reader almost nothing.

Which is why a careful reader should notice that three figures here all read Rs 2,000.00/- and are not the same quantity. The price and the strike agree because this pair is struck at the moneySaid of a contract whose strike stands at the current price of the reference asset.; the exposure agrees because exposure on one unit is the price of that unit. Change the strike in the instrument above and only one of the three moves.

Where does an estimate built from a delta go wrong, and in which direction?

The direction of the error is settled by the shape of the payoff alone, with nothing assumed. Draw a straight line touching the payoff at any single point at the end, and the payoff line never sits below it anywhere. Work the instance the instrument opens on: a straight line through the call payoff at Rs 2,200.00/-, where the payoff is Rs 200.00/- and the slope is one. Follow it leftwards and at Rs 2,000.00/- it reads nil, as the payoff does; at Rs 1,900.00/- minus Rs 100.00/- against a payoff of nil; at Rs 1,800.00/- minus Rs 200.00/- against a payoff of nil. Follow the same line rightwards and it agrees at every price. Above the strike the payoff is that very line.

So the honest statement of the direction is this: an estimate built from a slope never overstates a bought position at the end, and it understates one once the move carries the price across the corner. The sharper version some readers will have met, that the estimate is always too low after any large move, is not right at the end: on the arm the line was drawn along it is right rather than low. The error runs one way only, and a written position inherits the mirror of it. Set the instrument to a written call and the same gap appears on the other side of the truth.

Then the limit, and it matters as much as the result. The size of the gap is settled only at the end date. Away from the end date the size depends on the shape of the premium line, and describing that shape needs the input the working example does not have. Before the last day the line has no corner at all: it has curvatureThe way a slope itself changes as the reference asset moves, so one slope describes the line only near where it was read. along its whole length. Only the direction survives away from the end date.

Where would a real delta come from?

From a pricing modelA method that turns an estimate of how far a reference asset might travel into a premium, with the sensitivities coming out alongside it.: machinery that takes a description of how far the reference asset might travel over the life of the contract, together with everything this guide already has, and returns a premium. The delta falls out as a by-product. A way of pricing at Rs 2,000.00/- is also a way of pricing at Rs 2,001.00/-.

The plain consequence is that a delta is only ever as good as the travel estimate behind it. Two people looking at the same contract on the same day, both careful, put different estimates in and hold different deltas without either being wrong. A delta quoted without that estimate has been separated from its assumption, and somebody will read it as measured rather than modelled. How such an estimate is built and what it is called are covered separately, as are pricing models themselves. The last group on the card above is exactly the group the instrument refuses.

How does anybody read this outside a classroom?

Four readers meet a delta outside a classroom and each wants something different from it. In all four rows what matters is not the figure but the estimate of travel that produced it. A delta separated from its assumption will keep printing to two decimal places long after that assumption has changed.

The lender row carries one extra piece: for a bought contract nothing further can be called for, while for a written one that amount belongs to the collateral row named above. The household row is the tiffin service grown up. Ask the small movement question and the large movement question separately, every time.

Does a delta tell anybody what to do?

No, and the reasons for declining are worth stating out loud rather than dodging. Three refusals, each a different kind, and the third is the one that swallows the other two.

Everything drawn here describes an obligation and a rate of change. Not one line says which price will arrive. A sensitivity is not a signal.

Which figures follow from the four inputs, and which cannot be produced at all?

Every figure below comes from four inputs and nothing else: a price of Rs 2,000.00/-, a strike of Rs 2,000.00/-, financing at 6.50 per cent for one year, and a call premium of Rs 180.00/- the working record supplies. The put premium of Rs 57.93/- follows by parity. Enter those four in the instrument at the top and every row below comes back.

The last row read against the one above it holds the argument of this guide in a pair of lines. The distance between the two deltas costs nothing; the deltas themselves are unavailable at any price. An instrument that returned a plausible figure for the last row, printed to the same two decimal places as the rows above, would be more satisfying and worth less than nothing. Nobody could tell by looking which of the two had been given.

The reference asset, the strike, both premiums and every figure worked from them are invented for teaching.
Educational material. Not advice on any investment, tax, budget or market position.