9.12 Case studies

9.12.1February 2018: "Volmageddon" and the end of XIV#

Context XIV (VelocityShares Daily Inverse VIX Short-Term ETN, issued by Credit Suisse) delivered the daily inverse return of an index of short-dated VIX futures. Holders were short volatility, rebalanced daily. Its prospectus allowed the issuer to accelerate (redeem) the notes if the intraday indicative value fell to 20% or less of the prior day's closing value.

Timeline

• Friday 2 Feb 2018: VIX closed at 17.31. XIV's closing indicative value was $108.3681 (Credit Suisse press release, 6 Feb 2018).
• Monday 5 Feb 2018: VIX closed at 37.32, a rise of 115.6%, reported as its largest one-day percentage increase (reported by Invest In Vol via PRWeb and by an SLCG Economic Consulting working paper). XIV's closing indicative value was $4.22 (reported by Benzinga citing issuer data).
• 6 Feb 2018: Credit Suisse announced an acceleration event, because the intraday indicative value on 5 Feb was at or below 20% of the prior day's closing value. Expected accelerated valuation date 15 Feb, last trading day 20 Feb 2018; holders would receive the closing indicative value on the valuation date.
• Reported redemption value: $5.99 per note, based on the 15 Feb closing indicative value (reported by Schaeffer's and others).

Mechanism From 108.37 to 4.22 is a fall of 96.1%. For a −1× daily product, that implies the short-term VIX futures index rose about 96% during 5 Feb. Two pieces of this module's theory explain it.

Rebalancing feedback. A daily-rebalanced product with leverage $L$ and assets $A$ holds exposure $LA$. After an index return $R$, its exposure is $LA(1+R)$ but its target is $LA(1+LR)$, so it must trade $LA\,R(L-1)$ of the underlying at the close (Problem 14). For $L=-1$ this is $+2AR$: after VIX futures rise, inverse products must buy VIX futures, in the same direction as the move, at the same time of day. Leveraged long VIX products ($L=+2$) must also buy ($+2AR$). The flows add up to a large, predictable bid for VIX futures into the close on a volatile day.

Negative skew of short variance. XIV was a systematically sized short-variance position. For years it collected the roll yield of upward-sloping VIX futures (the futures version of the variance premium), with a return distribution like the table in 9.9: frequent small gains and one loss of almost everything. The 20% acceleration trigger turned a large loss into a permanent one.

There is a disputed question of whether trading in the minutes before the 4:15 pm VIX futures settlement on 5 Feb was engineered to trigger the termination (one working paper by SLCG Economic Consulting argues so). That is an allegation from one analysis, not an established finding.

9.12.2The Jane Street interim order: the expiry-day gamma angle#

The facts, timeline and first-order (delta) mechanism of SEBI's ex-parte interim order of 3 July 2025 are in M08 · 8.12 Case B, which also records where the sources conflict. They are allegations, disputed by Jane Street; the Securities Appellate Tribunal reserved its order on Jane Street's document-disclosure plea on 21 Oct 2026, and no ruling on the merits or final SEBI order had been reported as of 6 Nov 2026. Here we add only what this module's theory says about why the alleged pattern concentrated on expiry days of weekly Bank Nifty options (Wednesday expiries, discontinued for Bank Nifty from 20 Nov 2024).

One additional figure from the order SEBI's interim order (paras 15.27–15.49) states that on 17 Jan 2024 the group's Bank Nifty options delta exposure moved from −₹7,311.19 crore (about 09:30) to −₹39,426.15 crore by 11:47 during the morning "Patch I", while it bought ₹4,370.03 crore of constituents and futures, and peaked at −₹46,620.79 crore around 13:53; in the rest of the day ("Patch II") it net sold ₹5,372.12 crore of stocks and stock and index futures. Its profit that day in Bank Nifty options was ₹734.93 crore.

Delta per rupee of premium explodes near expiry For an ATM option, by (9.26) and $V_{ATM}\approx\sigma S\sqrt{\tau/2\pi}$, the delta notional bought per rupee of premium is $\Delta S/V\approx0.5\sqrt{2\pi}/(\sigma\sqrt\tau)$, which grows like $\tau^{-1/2}$. With Bank Nifty at a hypothetical 46,600 and 15% implied vol (Black–Scholes, $r=0$):

Time to expiry	ATM call (points)	Delta notional per ₹1 of premium
7 days	386.2	61
1 day	146.0	160
3 hours	51.6	452
30 minutes	21.1	1,106

A short-call, long-put book on expiry morning carries several times the index exposure, per rupee at risk in premium, that the same book carries a week earlier. This is how a delta of −₹39,426 crore can sit beside a cash position of a few thousand crore.

Gamma turns a temporary index move into delta The growth of the reported option delta from −₹7,311 crore to −₹39,426 crore during the morning is what high gamma does: as the index is pushed up through strikes, the calls sold become deeper in the money and their delta grows faster than at any other time in the option's life (9.8: on expiry day the "gamma zone" is a few tens of points wide). In this reading, the morning buying both raised the price at which options were sold and, through gamma, enlarged the short delta those options carried into the afternoon.

Back of the envelope Ignoring gamma in the afternoon, a fall of $x\%$ earns about $39{,}426\times x\%$ crore on the option delta; the reported ₹734.93 crore corresponds to $x\approx1.86\%$. Against that the cash leg loses its round-trip impact cost on a position roughly one-seventh the size of the option delta. The "extended marking the close" allegation targets the settlement level directly, which on expiry day is the only number the option payoff depends on.

Implied vol is not the lever, the index is Nothing in the alleged pattern needs implied volatility to be wrong. Options were priced off the index by market makers' models (M08 Case B, point 1); the profit comes from moving the index while holding a very large, gamma-amplified delta. For a market maker who sold those options and hedged them, the morning move and afternoon reversal show up in (9.12) as two large squared moves on a short-gamma book: a gamma loss that is a transfer to whoever caused the moves.

What changed Several later SEBI and exchange measures address exactly these channels: intraday monitoring of index-option positions with futures-equivalent limits (circular of 1 Sep 2025), one weekly expiry per exchange (from 20 Nov 2024, before the order), and the return to VWAP-based expiry settlement from 7 Dec 2026 (9.10). The settlement window makes moving the settlement print costlier.

9.12.3The SEBI retail-loss studies#

The numbers The studies, their primary press releases and the caveats on how the FY26 figures were obtained are in M01 · 1.11 Case 1, the course's single full treatment. This case uses only the FY26 figures that bear on volatility: options were about 92% of individual losses; about 97% of individual traders were option buyers and about 2% sellers; 59% of index-option turnover was in contracts on their expiry day and 75% within one day of expiry; 87.7% of option traders lost money against 66% of futures traders; and about 99% of the gross profit of prop traders and FPIs came from algorithmic entities (SEBI FY26 study; M01 · 1.11 Case 1).

Mechanism The module's results line up with these numbers.

• Who pays the premium. Retail is overwhelmingly long options. If implied vol exceeds realised on average (9.4), the long side loses on average before costs.
• Zero-days-to-expiry options. By (9.26) and (9.27), a 0DTE option's value is almost all gamma and theta. Its payoff distribution is the mirror image of 9.9's table: frequent total losses, rare large wins, strongly positive skew. A buyer with lottery preferences (Barberis and Huang 2008) overpays for that skew.
• Costs. At 0.15% STT plus exchange charges and flat brokerage, costs are a large fraction of cheap-option premiums, and a 0DTE trader trades often.
• The other side. Prop and FPI algorithms are, in aggregate, net sellers of this volatility and hedgers of its gamma, operating with the scale, cost structure and risk systems that 9.9 says short-vol needs.

9.12.4March 2020: India VIX and the failure of short volatility#

Timeline (as reported)

• 23 Mar 2020: Nifty closed at 7,610.25, down 1,135.20 points (−12.98%), reported as its largest one-day percentage fall (Tribune/Business Standard coverage of the close). Sensex fell 13.15%.
• 24 Mar 2020: India VIX touched 86.63 intraday, above its 2008 peaks (Business Standard, 24 Mar 2020); its highest close in that episode is reported at about 83.6 on 24 Mar 2020 (secondary sources). Business Standard gave the all-time intraday high as 92.53 in November 2008.
• Nifty's intraday low of that episode was 7,511 on 24 Mar 2020 (reported).

Mechanism Express the 23 Mar fall in units of the implied daily standard deviation a seller would have sold at. At an implied vol of 20% (a hypothetical level, for scale), the daily standard deviation is $0.20/\sqrt{252}=1.26\%$, and $\ln(1-0.1298)=-0.139$ is about 11 such standard deviations. Under the lognormal model its probability is about $\Phi(-11)\approx2\times10^{-28}$; under 9.9's GARCH-t-jump world it is rare but present, which is why that world's worst week was 18% of notional. A short strangle seller faced three losses at once: the put leg's intrinsic value, the rise in implied vol on every remaining option (negative vega), and the margin increase that forces reduction at the worst prices. In variance-swap terms the realised premium was hugely negative: realised variance far exceeded what had been implied.