Volatility Surface Trading: Skew, Convexity and the Price of Protection

An option’s implied volatility is meaningful only in context: its strike, expiry, executable price and relationship to the rest of the surface. A low absolute volatility can still be expensive relative to the risk expected over that option’s life. A high downside volatility can reflect a premium for a real concentration of jump risk rather than an obvious selling opportunity.

This reference report consolidates the basic surface lessons, advanced readings and companion exercises into one framework for interpreting level, skew, curvature and term structure. It preserves the teaching structures and numerical conventions while separating a descriptive surface signal from a trade whose expected payoff has actually been tested.

Start with a surface of usable prices

The underlying object is a set of option prices across strikes and expiries observed at a coherent time. Implied volatility converts each price into a common quoting language through a chosen vanilla model. If the inputs are inconsistent, a smooth three-dimensional picture can hide poor data.

The minimum useful chain contains strike, exact expiry, option type, bid, ask, quote timestamp, underlying or forward reference and the carry inputs used in inversion. Last-traded price is often unsuitable because the trade may have occurred before the current spot move. Retain bid and ask IVs so the apparent relative-value difference can be compared with execution uncertainty.

The retained synthetic exercise has 252 history observations and a 62-row option chain: two expiries with 31 strikes each. The maturities are approximately 0.0192 and 0.0767 years, and strikes run from 26,300 to 27,800. These are sample inputs designed to show skew and a term premium, not an independently sourced live option surface.

Level, skew and curvature answer different questions

At a given expiry, three quantities give a compact summary.

Three different descriptions of one expiry

Under the put-minus-call convention used here:

$$RR_{25}=\sigma_{25\Delta\,\mathrm{put}}-\sigma_{25\Delta\,\mathrm{call}},$$

$$BF_{25}=\tfrac12(\sigma_{25\Delta\,\mathrm{put}}+\sigma_{25\Delta\,\mathrm{call}})-\sigma_{ATM}.$$

In words: The butterfly subtracts central implied volatility from the average implied volatility of the two wings.

Each $\sigma$ is an implied volatility at the specified delta and expiry. A positive risk reversal under this convention means that the put wing is more expensive in volatility terms. Some markets quote the opposite sign, so the definition must travel with the number.

The source uses risk-reversal bands of two to four volatility points as an ordinary regime, four to six as elevated and above six as stressed. It uses a butterfly range of roughly 0.8–1.6 points as a representative baseline. These are retained historical teaching ranges, not universal thresholds. A new market, sampling method or structural regime can change the distribution.

To compute a 25-delta point, interpolate locally between suitable nearby strikes using a stated delta convention. Regressing volatility against every strike’s delta and calling the result “interpolation” can distort the wings. In FX, spot versus forward delta and premium adjustment add further conventions. The same symbol does not ensure the same economic point.

Figure 01 · Conceptual framework

A credible surface begins with coherent prices

How the parts connect

1. AOption chainvalidate market observationsQuote validation
2. BReference inputscheck input consistencyQuote validation
3. CQuote validationbuild the surfaceImplied-volatility surface
4. DImplied-volatility surfacetest the relative-value claimExecutable comparison

A skew is a price of asymmetry

An equity-like downside skew can reflect demand for crash protection, an asymmetric return distribution, leverage effects and the cost of supplying insurance. A steep skew says that downside optionality commands a different price from upside optionality. It does not isolate the contribution of each explanation.

The retained asymmetric volatility teaching ladder

The numerical asymmetry is the exhibit: using one flat 15% volatility would ignore substantially different wing prices.

A put-minus-call volatility spread can appear mean reverting historically while remaining elevated through a developing crisis. Selling the rich wing introduces the loss exposure for which the premium may be compensation. If the purpose is to trade relative skew, neutralising initial vega helps separate a parallel-level move, but does not remove delta, gamma, jump or changing-smile exposure.

“Cheap calls and expensive puts” also does not by itself justify a bullish trade. The difference can be structurally persistent, and bid-ask costs can exceed the expected normalisation. The required forecast concerns the relative surface change over the holding period, not simply the current rank of a number.

Read term structure in variance-time units

Comparing annualised implied volatilities at two maturities is useful, but forward variance requires total variance. If $\sigma_1$ and $\sigma_2$ are consistent implied-volatility measures for maturities $T_1<T_2$, a simple forward-variance diagnostic is:

$$v_{\mathrm{forward}}=\frac{\sigma_2^2T_2-\sigma_1^2T_1}{T_2-T_1}.$$

The interpretation requires comparable option definitions and appropriate assumptions; it is not obtained by subtracting the two volatility percentages. A sharp event inside the short interval can lift short-dated annualised volatility without implying negative total variance later.

The materials use a calendar example with a near option priced at 165 and a farther option priced at 315, producing a debit of 150. Their quoted volatilities are 11.8% and 13.2%. These inputs show why the trade cannot be understood from the upfront debit alone. At the near expiry, the far option still has time value that depends on its own future implied volatility.

A long calendar can benefit from favourable relative decay or a differential volatility repricing. It is not guaranteed to be delta-neutral, positive theta or profitable in a particular slope regime. The source gives conflicting simple rules about whether backwardation favours or disfavours it; the resolution is to price both legs under explicit post-event scenarios. A slope label alone is insufficient.

Figure 02 · Qualitative comparison

Read four distinct dimensions of the surface

Read four distinct dimensions of the surface

Dimension	Comparison	Economic question
Level	ComparisonCentral implied volatility at one expiry	Economic questionHow is aggregate movement priced over this horizon?
Skew	ComparisonRelative downside and upside wing volatility	Economic questionHow is asymmetric protection priced?
Wing curvature	ComparisonAverage wing volatility relative to the centre	Economic questionHow expensive are tails relative to central outcomes?
Term structure	ComparisonTotal variance across consistent maturities	Economic questionHow is risk distributed across time and events?

Rank and percentile must not share one label

The source’s IV rank is a range position:

$$IVR=100\frac{\sigma_{\mathrm{today}}-\sigma_{\mathrm{min}}}{\sigma_{\mathrm{max}}-\sigma_{\mathrm{min}}}.$$

In words: Implied-volatility rank places today’s volatility within the observed minimum-to-maximum range.

An empirical percentile instead counts the share of past observations at or below today’s level. One extreme historical spike can depress range rank while leaving the percentile high. Treating the two as synonyms can change which trades a rule selects.

The exercise thresholds of 70 and 30 describe the upper and lower parts of a range, not automatically the top or bottom 30% of observed days. A constant historical series makes the denominator zero and requires an explicit handling rule. A shorter initial sample is not a full 252-day rank simply because the spreadsheet column has that name.

Neither statistic establishes that volatility will mean-revert during the chosen holding period. A useful signal compares the market-implied distribution with a forecast and an event calendar, then asks whether an executable structure benefits if that forecast proves right.

Translate the view into a structure and a scenario set

Surface views, candidate structures and remaining exposure

A volatility butterfly statistic is not the same object as a conventional price-payoff butterfly. Buying wing volatility and selling the middle needs explicit quantities and vega matching; the name alone does not define a payoff. Likewise, an unmatched ratio spread is not a bounded-risk structure merely because it contains several legs.

The teaching case uses a hypothetical instrument with spot 1,000; it does not represent any actual security. The short call and put strikes are 1,035 and 965, and the protective long call and put strikes are 1,070 and 930, using a five-point strike grid. Both short strikes are 3.5% from spot, and both protective strikes are 7% from spot.

Synthetic leg premiums and their combined credit

Each wing is 35 points wide, so the maximum terminal loss is 24 points per unit. The breakevens are 954 and 1,046. Maximum terminal profit is 11 points per unit between the two short strikes; the 24-point loss is reached at or beyond either protective strike. These premiums and contract specifications are synthetic assumptions, not observed quotes.

At the example’s assumed 100-unit multiplier, maximum loss is 2,400 per package. A strict 7,000 risk budget permits two packages, not three: three risk 7,200. Calling the excess “within tolerance” would change the stated rule. The quantitative lesson is to define whether the budget is binding before the attractive trade appears.

Figure 03 · Conceptual framework

A surface view needs a trade and a dynamics test

How the parts connect

1. ASpecific surface thesisexpress the viewCandidate structure
2. BCandidate structurestress remaining risksDynamics scenarios
3. CCandidate structuremake implementation explicitExecution and event rules
4. DDynamics scenariostest robustnessEvidence-based judgement
5. EExecution and event rulestest attainabilityEvidence-based judgement

A good surface fit does not prove realistic dynamics

A local-volatility model can reproduce an appropriately cleaned vanilla price surface while implying a particular pattern of future smile movement. A stochastic-volatility model provides another dynamic interpretation through mean reversion, vol-of-vol and spot-variance correlation. Neither is automatically the correct hedge model for every payoff.

For a vanilla relative-value strategy, it is often more useful to test plausible sticky-strike, sticky-delta and empirically estimated surface scenarios than to assume one calibrated model is truth. These scenarios describe what the implied volatility assigned to a moving option will do as spot changes. The resulting “total delta” can differ from the partial derivative that holds that option’s IV fixed.

The surface also needs no-arbitrage checks in price space. Calls should behave appropriately across strike, and the implied probability structure should not become negative because a spline oscillates between sparse quotes. A clean visual mesh is not a substitute for those checks.

Preserve the evidence behind performance claims

The advanced reading reports a comparison of flat-volatility and surface-aware premium-selling approaches:

Source-reported strategy results and their evidence limits

These are retained historical backtest claims whose full trade-level evidence has not been reproduced. The source also describes an event filter that could put one strategy in cash. Consequently, the table cannot isolate the value of a better pricing model from the value of changing exposure. A fair comparison must disclose the entire rule difference.

The same discipline applies to the proposed 50%-of-credit profit exit, volatility-rank filters and timing stops. They are research parameters. Compare net results across reasonable alternatives, include missed trades and adverse fills, and reserve an untouched period for final evaluation.

Quote uncertainty is part of the relative-value threshold

Suppose a surface comparison uses mid-prices for both legs. The resulting theoretical difference is not necessarily obtainable: the investor may have to buy at the ask and sell at the bid. The amount of spread crossed should be evaluated on the whole package, including the possibility that only some legs fill before the market changes.

This becomes particularly important for wing trades. A far out-of-the-money option can have a large implied-volatility interval because a small absolute premium difference translates into a large volatility change when vega is low. A visually extreme wing quote may therefore indicate low information quality rather than a large expected profit.

Open interest and volume provide context but do not guarantee an executable size at the displayed price. Retain a quote-quality flag, age, spread and available size. A useful research filter can exclude observations with insufficient information, provided the exclusion rule is declared before the performance test and the omitted share of the sample is reported.

Surface changes should also be measured at consistent coordinates. Comparing yesterday’s fixed strike with today’s 25-delta strike mixes a market movement with a change in which option is selected. Either can be useful, but they answer different questions. A desk can maintain both a fixed-strike view for held positions and a constant-delta view for the evolving market’s relative shape. Keeping both makes it easier to distinguish a true wing repricing from the mechanical effect of spot moving through the strike grid.

A surface-based investment decision should state what is expected to change: level, a particular wing, a maturity interval or the joint spot-volatility relationship. It should show the position under at least a contained move, an adverse gap, a parallel volatility change and a twist between expiries. Record the premium and spread at which the view becomes worth expressing.

The resulting discipline is precise: read the surface as several related prices of risk, then choose a structure that expresses one clear relative-value proposition. The aim is not to produce the most colourful surface chart, but to understand which change must occur for the position to earn more than its funding, execution and risk costs.