Module 12 of 12 · Running it for real

Tail risk and hard limits

Everything in this track has been building to one day: the ETF gaps down and implied volatility jumps at the same moment. A hedged short straddle book has no delta, so the first dollar of the move does nothing. The second does: gamma turns the hedge wrong, and the further the market goes the wronger it gets. On top of that, vega: every option in the book reprices higher as implied vol jumps, whether or not the ETF ends the day far from where it started.

The lab sizes a book to a risk target at a starting implied vol, then applies one day's shock. It separates the gamma loss from the vega loss, expresses the total in units of the daily risk the target implied, and shows the same day at six other risk targets. The loss scales one-for-one with τ. The strategy does not decide whether you survive; the risk target does.

It also shows a kill switch: a threshold loss, as a fraction of capital, at which the book is flattened and the system halted until a human looks at it. That rule has to be written down, automated and tested before the first trade, because the day it is needed is not a day for judgement.

🎛 Crash lab

20%
16%
8%
15 pts
25% of capital

Loss that day

-4001.6%

of capital

Of which from the vol jump

-7.4%

vega

In daily-risk units

3176.1σ

versus the target's typical day

Kill switch

FLATTEN

threshold 25%

τ = 5%τ = 10%τ = 15%τ = 20%τ = 30%τ = 40%
-1000.4%-2000.8%-3001.2%-4001.6%-6002.3%-8003.1%

A delta-neutral short straddle book sized to the risk target at the starting implied vol. The shock is one day: the ETF moves by the chosen amount and implied vol jumps by the chosen number of points. The hedge removes the first-order loss; what remains is gamma (the curve) and vega (the vol jump), and on the way down both hit at once. The ladder shows the same day at other risk targets: every loss scales one-for-one with τ, which is why the risk target, not the strategy, decides whether a book survives its first real crash.

Size for the shock, not for the income

At a 20% target, a 16% starting vol, an 8% drop and a 15-point vol jump, the book loses about a quarter of its capital in one day, roughly 20 daily standard deviations. That combination is not hypothetical: it describes several days in the last two decades on the main equity ETF. At a 10% target the same day costs half as much and the book survives to collect premium next month; at 40% it is gone.

The research behind this track settled on a 15% to 20% target for a diversified, hedged book and a 25% flatten threshold, and noted that the backtest's own worst day exceeded that threshold. A short-volatility strategy that has never tripped its kill switch has not been running long enough.

Things to try

  • • Find the largest risk target at which the day in the default settings stays under the flatten threshold.
  • • Set the vol jump to 0 and see how much of the loss was vega: on the worst real days it is roughly half.
  • • Raise the starting implied vol to 35%: the book is smaller for the same target, and the same 8% drop hurts far less. High vol is safer to sell than low vol, per dollar of risk.

Original interactive lesson. Labs run on a seeded synthetic ETF in your browser, built to illustrate mechanics, not to predict any real market. Option prices use a textbook model with zero interest rates. Educational analysis, not investment advice.

How are we doing?