6.5 Confidence Intervals & Options Pricing

6.5 Confidence Intervals & Options Pricing

Module 06 · Lesson 6.5 · Estimated read 9 min

The implied volatility printed on an option chain encodes the market’s expected move — the standard deviation of price returns over the life of the option. Translate that to a confidence interval and you have the market’s probabilistic forecast in dollar terms. The 1-sigma expected move is roughly the 68% confidence interval; 2-sigma is roughly 95%. This lesson shows how to compute the expected-move CI from option pricing, where Black-Scholes’ normality assumption breaks down on real markets, and how to size into a trade when your private confidence interval differs from the market’s. The skew and fat-tail corrections are the practical machinery.

1. Expected move as a confidence interval

For an option with implied volatility σ and time-to-expiration T (in years), the 1-sigma expected price move is approximately:

Where S is the current spot price. The factor √T comes from the random-walk assumption — volatility scales with the square root of time. For SPY at $580 with 30-day at-the-money IV of 13.5%:

• T = 30 / 365 ≈ 0.0822
• √T ≈ 0.287
• EM(1σ) ≈ $580 × 0.135 × 0.287 ≈ $22.46

Under the Black-Scholes assumption that returns are normally distributed, this 1-sigma move corresponds to a 68% confidence interval: SPY ends the next 30 days within roughly $557.50 to $602.50 with 68% probability. The 2-sigma CI — double the move — gives roughly $535 to $625 with 95% probability. The 3-sigma CI — triple the move — gives 99.7% probability under normality.

A quicker heuristic for a rough check: the at-the-money straddle price approximates the expected absolute move — about 0.8× the 1-sigma move, since each ATM option is worth ≈0.4·S·σ·√T. If the SPY 30-day ATM call is $9 and the put is $9, the $18 straddle implies a 1-sigma move of roughly $18 / 0.8 ≈ $22.50, matching the formula above. To recover the 1-sigma expected move from a quoted straddle, divide by 0.8. This is the simplest way to extract the expected-move CI from quoted prices without computing IV directly.

The ratio of straddle to spot, called the “straddle yield,” approximates 0.8× the 1-sigma move as a percentage of spot. SPY straddle of $18 / spot of $580 = 3.10% straddle yield; dividing by 0.8 gives a market-implied 1-sigma move of roughly 3.88% over 30 days. Comparing this to historical realized volatility (the 30-day realized 1-sigma move) tells you whether the market is pricing more or less uncertainty than recent history would suggest.

2. Where Black-Scholes assumes normality and breaks

The Black-Scholes-Merton model assumes log-returns are normally distributed with constant volatility. Real equity returns violate both assumptions. Three deviations matter for confidence-interval reading:

Fat tails. Real return distributions have more probability in the tails than the normal distribution predicts. Days that should be 4-sigma events under normality occur far more frequently than the model implies. The August 2024 yen-carry-unwind day was, by 30-day rolling vol, a multi-sigma move; it was not nearly as rare as the 1-in-many-thousand-day count under a normal would suggest. Empirically, equity returns are better described by a t-distribution with low degrees of freedom, or by mixture models, both of which have fat tails.

Skew. Equity returns have negative skew — large down moves are more likely than equally large up moves. The Black-Scholes model is symmetric and prices puts and calls at the same IV at the same delta. The volatility skew (covered in Module 03) is the market’s correction for this asymmetry: OTM puts trade at higher IV than OTM calls because the model under-prices downside tail risk under normality.

Volatility clustering. Realized volatility is not constant; it persists. High-vol days cluster, low-vol days cluster, and the conditional variance evolves predictably. GARCH-family models capture this; Black-Scholes does not. The practical implication: the 1-sigma EM is a forward-looking forecast that conditions on current IV, but if volatility is itself rising or falling, the CI should shift over the option’s life.

The implication for confidence intervals is that the Black-Scholes 1-sigma EM is a useful first cut but understates downside-tail probability and overstates symmetric-tail probability. A true 95% CI for SPY returns over 30 days is wider on the downside and narrower on the upside than the symmetric 2-sigma rule implies. Practitioners often quote “95% expected move” using the actual put and call deltas at 2.5% probability rather than a symmetric 2-sigma move.

3. Pricing options against your own CI estimate

If you have a private estimate of the expected-move distribution that differs from the market’s implied distribution, you have a tradeable view. Two cases:

Your CI is narrower than the market’s. You believe IV is too high — the market is pricing more uncertainty than your model says is justified. The trade is short volatility: sell straddles, sell strangles, or sell tail risk via OTM puts. The expected payoff is positive if you are right, but the loss profile is convex against you (unlimited upside risk on naked short calls, defined but large downside on short puts). Position sizing must account for the convex tail.

Your CI is wider than the market’s. You believe IV is too low — the market is under-pricing uncertainty. The trade is long volatility: buy straddles, buy strangles, buy OTM tail protection. Loss is bounded (premium paid), gain is unlimited if vol expands. The trade pays in expectation if your CI is more accurate, but bleeds time decay if vol stays calm.

The practical question is how to estimate your private CI. Three approaches:

• Realized-vol baseline. Compute trailing 30-day or 60-day realized volatility on the underlying. If realized is consistently below implied, the market is paying you to be short vol; if realized is above implied, you are paying the market to be short vol. The variance risk premium is real but not always positive.
• Regime-conditional adjustment. Same realized vol can imply different forward-looking distributions depending on regime. A 12% realized vol after a stress event is likely to mean-revert higher; the same vol in a sleepy regime may be sticky. Condition your CI on what regime indicators say about the next few weeks.
• Event adjustment. If a known catalyst (earnings, FOMC, NFP) sits inside the option window, the unconditional CI under-prices the catalyst. Decompose the IV into pre-event drift and event-day jump, and check whether the market’s implied jump matches your private estimate.

The discipline is not to have a CI estimate that is precisely calibrated — that is unattainable — but to have one that is reproducible, regime-aware, and explicit. A trader who can articulate why their CI is wider or narrower than the market’s has a real position; a trader who just “feels” vol is too high or too low does not.

4. Trade sizing when your CI is wider than the market’s

Suppose you believe SPY’s 30-day 1-sigma move is $25, but the market is pricing $22 (the SPY straddle is $22). Your private CI is wider by 13.6%. The long-volatility trade pays if realized exceeds $22 over the next 30 days. The Kelly-style sizing principle says position size should scale with your edge (the difference between your CI and the market’s), divided by the variance of the trade outcome.

In practice, several factors limit how aggressively to size:

• The variance risk premium is structurally negative. On average, implied volatility exceeds realized over time. Long-vol positions bleed unless realized vol expands meaningfully or your timing is exceptionally good. Sizing should account for this structural headwind — even a correct view costs you carry while you wait.
• Tail-risk asymmetry. Long-volatility trades pay out asymmetrically: you can lose 100% of premium but gain multiples of premium if realized volatility spikes. The expected payoff can be positive even with a negative carry profile, but the distribution is right-skewed and the trade can sit at a loss for weeks before paying.
• Liquidity and exit considerations. Wide bid-ask spreads on individual-name options can erase edge. Stick to liquid names (SPY, QQQ, IWM, large-cap tech) where the round-trip cost is low relative to the size of the expected move.

A defensive sizing rule: cap any single long-volatility position at a fraction of capital that allows you to absorb the maximum loss (full premium decay) without forced exit. For most trader risk profiles this implies position sizes of 1-3% of equity per trade, which leaves room for multiple concurrent vol bets and avoids the all-eggs-one-basket failure mode that destroys long-vol books in extended low-vol regimes.

5. Practical workflow and pitfalls

The end-to-end workflow for using confidence intervals in options trading:

1. Read the at-the-money straddle price. This is the market’s 1-sigma expected move.
2. Compute the 1-sigma move as a percentage of spot. Compare to recent realized.
3. Adjust for skew: the put wing implies more downside than the symmetric 1-sigma read would suggest.
4. Adjust for known catalysts inside the window: events drive jumps that are not captured by linear vol scaling.
5. Compare to your private CI estimate (realized-vol baseline, regime-conditional, event-adjusted).
6. If the gap is meaningful, structure a long-vol or short-vol trade and size for survivability.

Three pitfalls that consistently trip up CI-based options trading:

Pitfall one: assuming normality literally. The 1-sigma rule under-represents downside tail risk. A trader who sells a 5-delta SPY put thinking “5% probability is essentially zero” underestimates the actual probability of touching that strike, especially in regimes with elevated put skew. The fat-tail correction widens the practical CI on the downside.

Pitfall two: confusing implied with private. The market’s 1-sigma EM is what option pricing implies, not what is going to happen. Treating it as a forecast rather than as a price is the inverse of trading the difference between price and value. Your private CI is your forecast; the market’s is the price you trade against.

Pitfall three: ignoring the variance risk premium. Long-volatility trades face a structural headwind because, on average, implied exceeds realized. A correct view on rising vol still has to overcome the negative carry. Trades sized as if there were no carry are systematically over-sized. Account for it explicitly in expected-payoff calculations and you will size more appropriately.

Key takeaways

• Expected move is a confidence interval. 1-sigma EM ≈ 68% CI under normality; 2-sigma ≈ 95%. The ATM straddle price is the cleanest read on the market’s 1-sigma EM.
• Real markets have fat tails and skew. Black-Scholes normality understates downside tail risk and overstates symmetry. The skew premium in put IV is the market’s correction.
• Trading is implied vs. private CI. If your CI is wider than the market’s, long vol; if narrower, short vol. The trade is the difference.
• The variance risk premium is real. Long-vol carries negative on average. Short-vol carries positive but with convex tail risk. Both have to size for the structural headwind.
• Size for survivability. Long-vol bleeds; short-vol can blow up. Position sizes that survive the worst plausible drawdown for the strategy keep you in the game long enough to collect on a correct view.