Percent of Max Is a Fact About Your Entry Price
Percent of maximum captured is the most common exit trigger on an options book and one of the few numbers on the page that contains no information about the future. It is a fact about your entry price. Sorting a book by it is a directional bet, placed without anyone deciding to place one.
The ask was administrative. A client was leaving a derivatives counterparty. There was a fixed account fee — five thousand dollars a month, charged at the account level rather than per position, running roughly $46,700 to the last expiry. Sixteen paired structures were still open against that counterparty: a capped call spread and a smaller uncapped call on the same index, same maturity, held side by side. Somebody needed a list of what to sell.
A list was produced. Each pair carried a decision and a one-line rationale, and the rationale leaned on a single number: percent of maximum captured. Positions reading above 85% were marked sell and close — minimal remaining upside. Positions reading in the middle were sell and re-buy. Three positions reading deeply negative were flagged client discussion — re-buy is an active rally bet.
That is a defensible-looking document. It is also, read carefully, a recommendation to sell the convex positions and keep the bond-like ones, made by a metric that cannot tell the difference and was never asked to.
What the number actually measures
Percent of maximum captured is current profit divided by maximum achievable profit:
(value today − premium paid) ÷ (maximum value − premium paid)
Premium paid appears twice. It is a sunk cost. It is the one term in the expression that describes something that already happened and cannot be recovered by any decision available today.
The consequence is arithmetic, not opinion. Take one position — one strike, one cap, one index level, one number of days remaining — and vary only what was paid for it. The position is identical. Its value is identical. The gap between its value and the most it can pay is identical. The metric moves anyway, and moves a lot.
That is easy to assert and easy to check, so here it is checked against a real book.
Twenty-eight positions at the same moneyness
A single insurance book, one valuation date, 524 open call spreads across six counterparties. Filter to the ones sitting at comparable moneyness — index between 10% and 20% above the cap, so the payoff is nearly fixed in every case. Twenty-eight positions survive that filter.
Across those twenty-eight, percent of maximum captured ranges from 23.6% to 99.3%.
The forward-looking measure — the gap between today's value and the maximum, expressed as a return on the capital still committed — ranges from 0.32% to 9.31%.
One of those two ranges is describing the positions. The other is describing the trades that created them. The correlations say which is which:
| Within the 28 equally-deep positions | Correlation with % of max |
|---|---|
| Premium paid, as a share of maximum value | −0.90 |
| Days remaining | −0.20 |
| Cushion (how far the index sits above the cap) | −0.13 |
Percent of maximum captured is almost entirely a function of the entry price. Once moneyness is held roughly constant, it is close to uninformed by anything about the position itself. The forward measure behaves the opposite way: within the same twenty-eight, the gap to maximum correlates +0.95 with days remaining, which is exactly what it should do, because what is left in a position that is nearly certain to pay its maximum is discounting and nothing else.
The clusters are visible without any statistics. Positions entered at 50–67% of maximum value read 76–99% captured. Positions entered at 91–97% read 24–73%. Same book, same day, same moneyness band. Different trades.
Across all 524 the picture is more flattering — percent of max correlates +0.88 with cushion and −0.88 with days remaining — and that is how the metric survives. At the level of a whole book it looks like it is tracking moneyness and time, because deep positions do tend to read high. It is only when you compare positions you might actually choose between that it stops working.
What happened next
The migration list was drawn up on a Tuesday. Seven days later the same positions were re-marked. The index had risen 5.3%.
One pair had expired. The remaining fifteen went from roughly $975,000 to roughly $1,188,000 — +21.85%, about $213,000.
The list had measured the total remaining upside in those fifteen pairs at about $326,000, spread over maturities running out more than nine months. 65.4% of it arrived in seven days.
That is not a forecasting triumph and I am not presenting it as one. A levered options book in a 5.3% index rally does that. The rally is the reason the week is informative, not the reason the analysis was wrong. What the week did was make three specific defects legible at once.
The metric scored 41% of the gain at zero. Every pair held a capped spread alongside an uncapped call, and remaining upside was computed on the spread alone — maximum value less current value. An uncapped call has no maximum value, so it contributed nothing to the measure. It contributed $86,600, 40.6% of the total gain. The one leg in each pair with unbounded payoff was invisible to the number driving the decision.
The two positions marked "minimal remaining upside" delivered 4.9× their measured upside. Both read above 85% captured. Between them the list credited about $5,500 of remaining upside. They gained about $27,000.
The three positions flagged as an active rally bet were the best performers in the book. They were underwater, deeply out of the money, and carried the most convexity of anything on the list — which is why they were flagged, and why they gained 31.1% in a week. They were 17.9% of the pair value on Tuesday and produced 25.4% of the gain, about $54,200. That is more, on its own, than the entire $46,700 fee that motivated the migration.
Across the fifteen surviving pairs, percent of maximum captured correlated −0.97 with the realized seven-day return. The metric ranked the book almost perfectly backwards.
The part that cuts the other way
That −0.97 was guaranteed to be negative in a rally. Positions with low percent-of-max readings are the ones with convexity left; convexity pays in a rally; the correlation follows mechanically. In a 5.3% selloff the sign flips and the same three flagged positions are the worst performers on the list.
I want to be exact about this, because the inverted version of the error is just as bad. Percent of maximum captured is not a bad ranking that a better ranking would beat. It is a pure proxy for how much convexity a position has left, wearing a label that suggests it measures progress toward a goal.
Which means sorting a book by it and selling from the top is a coherent trade with a name. It is selling convexity and keeping duration. There are real books where that is exactly right — a hedging program that has met its objective and wants the risk off, a portfolio that needs the collateral. The problem is not the trade. The problem is that on this list nobody chose it, nobody wrote it down, and the document said minimal remaining upside instead of we are reducing convexity in a rising market, which is what it meant.
An exit metric is a governance rule. This one had never been written down, so it was never reviewed, so nobody noticed it was expressing a market view.
What the exit decision actually is
Strip the metric out and the arithmetic is unforgiving in a useful way.
At mid-market, with proceeds reinvested at the risk-free rate, holding and selling are identical by construction. Today's value carried forward at r is the risk-neutral expected payoff. The option price already contains that discounting. The workbook accompanying this piece has a cell for the difference, and with zero transaction cost and cash earning r it reads exactly zero — not approximately, exactly.
So a liquidation decision is never about how much profit is already banked. It cannot be. It is about three things:
- What execution costs. The one certain number in the exercise, and the one most often omitted. A position whose remaining return is 30 basis points does not survive a 40 basis point round trip.
- What the cash earns. Sell only if your funding rate genuinely beats the rate embedded in the option price.
- Whether you hold a view the market does not. This is legitimate and it is the only one of the three that can justify a large exit. It must be stated as a view, by someone accountable for it.
Percent of maximum captured answers none of the three.
The migration case is cleaner still, and it is the situation most books actually face. Selling a position and re-establishing the same structure with a different counterparty leaves the exposure unchanged. The market view drops out entirely. What remains is the round-trip cost against the benefit of moving — a fee avoided, collateral released, a limit freed. That is a subtraction. It needed no view on the index and it never needed percent of maximum captured, which is how a metric that encodes a directional bet ended up ranking a decision that had nothing to do with direction.
What to write down
Before the next exit list, not during it. Four lines, in language somebody can check afterward.
- The trigger. Which number initiates a review, and — stated explicitly — what it measures. If it contains the entry price, say so in the same sentence.
- The unit. Whether the rule applies to a leg or to a structure. A measure computed on one leg of a paired position will score the other leg at zero, and the leg it ignores is usually the convex one.
- The three costs. Execution, funding differential, and the view. The first two are numbers. The third is a person's name.
- The view, if there is one. Reducing convexity is a legitimate decision. It is not a legitimate side effect. If the exit rule sells convexity first, write that down as the policy it is, and let somebody approve it.
A book with those four lines produces exit lists that survive being read six months later. A book without them produces a defensible-looking document, a set of trades nobody explicitly chose, and a meeting after the fact in which somebody explains what the metric was actually measuring.
The positions were never the problem. The number was.
The calculator
Every mechanism in this piece is reproduced in a Black-Scholes-Merton workbook you can download and check. It prices a call spread and an optional uncapped call, shows percent of max beside the forward-looking measures for the same position, and reproduces the paired-structure blind spot on your own inputs. The Metric Trap tab holds one position at one index level and varies only the entry price, so you can watch the metric swing while the position does not.
Every price in it is a mid-market theoretical value. It assumes you can trade in and out at mid, which you cannot — put your own round-trip cost into the Exit Decision tab, because it is the input most likely to change the answer.
It is on the Tools page.
Scarborough Road works with insurers and asset managers on derivatives governance, hedging operations, and the investment data infrastructure that demonstrates the governance was followed.
Client data, used with permission and de-identified. Entity names, counterparty names and position identifiers have been removed; figures are expressed as percentages and rounded notionals, and structures are described generically. General commentary on measurement and governance practice. Not legal, actuarial or investment advice, and not a recommendation of any strategy, position or counterparty. Model figures are outputs of a Black-Scholes-Merton implementation under stated assumptions — constant volatility, constant rates, continuous trading, lognormal returns, European exercise, a single flat volatility — none of which hold exactly. Real quotes reflect a full surface, bid-offer and financing terms, and will differ. A seven-day observation window in a rising market is an illustration of a measurement failure, not evidence about returns.
