> ## Documentation Index
> Fetch the complete documentation index at: https://raymondchang.me/llms.txt
> Use this file to discover all available pages before exploring further.

# Vol surface methodology

> Total variance, SVI/SSVI parameterisation, static-arbitrage conditions, and calibration practice.

> [!NOTE]
>
> **Status of this page:** P0 draft. Complements the desk-level pages [Volatility surface](https://raymondchang.me/equities-derivatives/global/volatility-surface.md) and [mock interview](https://raymondchang.me/equities-derivatives/mock/volatility-surface.md). Formulas are plain text; verify parameter constraints against the cited papers before implementing.

## 1. Objects

* Total implied variance w(k, T) = sigma\_imp(k, T)^2 x T, with log-moneyness k = ln(K / F\_T).
* Working in (k, w) removes the explicit time scaling and makes arbitrage conditions easier to state.
* Forward F\_T must be consistent with the dividend and borrow assumptions; an error in F shifts the whole smile in k.

## 2. Static no-arbitrage conditions

1. **Calendar spread:** w must be non-decreasing in T at fixed k (the partial derivative of w in T is non-negative).
2. **Butterfly:** the implied risk-neutral density must be non-negative at each maturity. This is a condition on the first and second derivatives of w in k for every slice.
3. **Wing growth:** Roger Lee's moment formula bounds the large-|k| slope of w by 2 (so total variance grows at most linearly in |k|).

A fitted surface that passes slice-by-slice fits can still fail the calendar condition; check both.

## 3. SVI and SSVI

**Raw SVI** for one maturity: w(k) = a + b (rho (k - m) + sqrt((k - m)^2 + s^2)), five parameters (a, b, rho, m, s). It is linear in the wings, consistent with Lee's bound when b(1 + |rho|) is at most 2 (check the exact form in the source).

**SSVI** (Gatheral and Jacquier) parameterises the whole surface through ATM total variance theta\_t and a function phi(theta), with a correlation parameter rho. It comes with explicit sufficient conditions for no calendar-spread and no butterfly arbitrage, which is why it is popular as a starting point.

## 4. Calibration workflow

1. Clean quotes: drop crossed or stale quotes, use mid or a bid/ask-aware loss, remove deep ITM.
2. Build the forward and discount curve per expiry.
3. Fit slice by slice (SVI) or globally (SSVI) with weights by vega or by bid-ask width.
4. Enforce arbitrage constraints in the optimiser or project afterwards.
5. Check stability day over day: parameter jumps indicate overfitting or data issues.
6. Output: a smooth surface, a local-vol or Greeks-ready representation, and diagnostics (RMSE in vol points, number of constraint violations).

## 5. Applications and pitfalls

* **Pricing and hedging:** use the surface to interpolate quotes for market making; sticky-strike vs sticky-delta assumptions change the hedge ratio. See [Greeks and hedging error](https://raymondchang.me/quantitative-research/derivatives/greeks-hedging-error.md).
* **Warrants and structured products:** quote grid is sparse; extrapolation beyond the quoted range is model-dependent, so use wing constraints.
* **Dividend-driven distortions** appear as apparent smile noise around ex-dates.

## Points to verify

* Exact SVI constraint on b(1 + |rho|) and the SSVI conditions as stated in Gatheral and Jacquier (2014).
* HK single-stock option and warrant quote conventions (strike grids, bid/ask requirements).

## Sources

* Gatheral and Jacquier (2014), Arbitrage-free SVI volatility surfaces, Quantitative Finance.
* Gatheral (2004), A parsimonious arbitrage-free implied volatility parameterization.
* Lee (2004), The moment formula for implied volatility at extreme strikes, Mathematical Finance.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.