Skew Trading Notes: Measuring It, Trading It, and When It Steepens
1. A thirty-second recap: what skew is
Plot implied volatility against strike for one expiry and the smile is asymmetric — OTM puts (left wing) richer than OTM calls (right wing). That asymmetry is skew. The textbook causes: fat tails in returns, the negative price–volatility correlation (leverage effect), and the plainest supply-and-demand fact of all — holders of the underlying are perpetual buyers of crash insurance.
Skew trading trades the slope itself: steepening or flattening, not direction.
2. How to measure it: three conventions
- 25Δ risk reversal: IV(25Δ put) − IV(25Δ call). The market’s standard language, but it requires delta interpolation.
- Fixed moneyness: $IV(K/F=0.95) - IV(K/F=1.05)$ — the 95-105 skew.
- Wide wings: the 90-110 skew, capturing deeper tails.
This note uses conventions 2 and 3 — pure closing-price math, fully reproducible. Chain construction rule: puts for $K < F$, calls for $K \geq F$ (the OTM convention — best liquidity, and ITM inversions are the least stable). $F = S e^{rT}$ with $r = 1.5%$, dividends ignored.
3. A measured snapshot (2026-09-21 close, public data)
Data: public exchange quotes via akshare (IV backed out from closing prices by bisection; monthly contracts only; September was deliberately skipped as it sat on expiry — October and December only).
| Underlying | Expiry | ATM IV | Skew 95-105 | Skew 90-110 |
|---|---|---|---|---|
| 510300 CSI 300 ETF | 2026-10-28 | 15.9% | +3.8 | +3.7 |
| 510300 CSI 300 ETF | 2026-12-23 | 19.3% | +5.6 | +4.8 |
| 510500 CSI 500 ETF | 2026-10-28 | 21.5% | +5.6 | +6.4 |
| 510500 CSI 500 ETF | 2026-12-23 | 23.3% | +9.4 | +8.4 |
Units: vol points (percentage points). Three observations:
- The 500 out-skews the 300: at the same December expiry, the CSI 500’s 95-105 skew (+9.4) is roughly 1.7× the CSI 300’s (+5.6). Richer crash insurance for the 500 is the norm — higher constituent volatility, more rigid hedging demand.
- Far months out-skew near months: December beats October on both underlyings, and the ATM term structure slopes up (510300: 15.9% → 19.3%). The market pays extra slope for distance uncertainty.
- The smile is intact: in the measured December 510500 chain (OTM convention), the left wing holds a 30% plateau against a 19% right wing — the asymmetry is visible to the naked eye:
| Strike | Type | Close | IV |
|---|---|---|---|
| 7.000 | Put | 0.1339 | 29.9% |
| 7.250 | Put | 0.1963 | 29.5% |
| 7.500 | Put | 0.2816 | 29.4% |
| 7.750 | Put | 0.3957 | 29.8% |
| 8.000 | Call | 0.2579 | 19.1% |
| 8.250 | Call | 0.1781 | 19.9% |
| 8.500 | Call | 0.1212 | 20.6% |
| 9.000 | Call | 0.0593 | 22.5% |
(F = 7.902; expiry 2026-12-23; full methodology at the end.)
4. How skew is traded: the textbook expressions
Steepening (long skew): buy the OTM put, sell the OTM call — a risk reversal. The cost: in a gentle rally, the call you sold rallies while your put bleeds — you lose on both legs.
Flattening (short skew): the reverse — sell the put, buy the call, collecting the slope. The catch: skew steepens exactly when markets fall, so you lose the slope, the direction, and the vega all at once.
Three engineering notes:
- Hedge delta to isolate the slope — otherwise you are not trading skew, you are running a leveraged directional bet.
- Match expiries across legs — crossing months trades the term structure of skew, a different strategy.
- Wing liquidity: past 95-105 the quoted spreads widen fast; a closing-price level is not a fill you can get intraday.
5. When skew steepens: a scenario-mechanism table
| Scenario | Mechanism | Skew |
|---|---|---|
| Fast index decline, volatility regime shift | Insurance demand spikes, the put wing is bid | Steepens (the classic) |
| Futures discount deepens | Hedgers buy puts passively | Steepens |
| Prolonged low-vol sideways grind | Sellers compress both wings, insurance goes unwanted | Flattens |
| Gentle bull market | Call-side speculation revives | Flattens (even a call-wing kink) |
The steepeners’ case: the 500-vs-300 gap and the far-vs-near gap both say the market is already paying dearly for tails — the question is whether that premium keeps rising, or is rich enough to trigger mean reversion. The flatteners’ case: slope this high invites sellers. Both sides have a point; that is precisely why the market clears.
6. Three ways skew trades die
- Naked vega: if the legs’ vega is not balanced, a parallel vol shift kills you even when your slope call is right.
- Roll timing: near-month skew behaves strangely in the last two weeks; roll one day wrong and the slope convention changes completely.
- Deciding on closes, executing intraday: a closing-snapshot deviation can be half-eaten by the next open’s spread.
7. Methodology (reproducible)
Data from public exchange quotes (fetched via akshare); IV by bisection against BSM closing prices; OTM chain split at $K/F$; skew as the difference of linear interpolations in $K/F$. Assumptions: $r=1.5%$, no dividends, monthly contracts, T on calendar days/365. A different assumption (say $r=2%$) moves the levels by tens of basis points, but the relative structure stands — which is why this note stresses relative comparisons (500 vs 300, December vs October) rather than absolute truths.
Data shown for research and educational purposes; nothing here is investment advice.
Next note, back on the main line: from single curves to the full surface — term-structure interpolation and cross-expiry no-arbitrage.