Pineify logo

Options and correlation research

Dispersion trading: index vs constituent volatility

Dispersion trading takes opposing volatility positions in an index and a basket of its constituents. A common long-dispersion structure buys constituent options and sells index options. Its exposure depends on the index weights, constituent volatilities, correlations and option sizing. Lower realized correlation can favor that structure relative to the correlation priced at entry, but option premiums, hedging and costs determine the result. A chart of stock returns cannot establish a historical options profit.

See the decision guide

How constituent correlation enters index variance

For a fixed-weight basket with simple return rI = sum(wi × ri), its variance is sigmaI² = sum(wi² × sigmai²) + 2 × sum over i < j of (wi × wj × sigmai × sigmaj × rhoij). Weights sum to one; every volatility and correlation must use the same horizon and observation convention. A live index changes weights, so this fixed-weight identity is a local research model rather than a full historical replication.

  • The squared-weight term measures the contribution without cross-stock covariance. The pair terms explain why constituent moves can reinforce or offset one another. Averaging stock volatilities does not calculate index volatility.
  • An effective average correlation can be inferred as rhoBar = [sigmaI² - sum(wi² × sigmai²)] / [2 × sum over i < j of (wi × wj × sigmai × sigmaj)]. It is a volatility-and-weight-adjusted average, not the ordinary average of all correlation pairs.
  • For realized inputs, use synchronized return observations. For implied inputs, use compatible maturities, quote times and volatility definitions. Mixing an ATM implied volatility with a variance estimate from an entire option strip changes the model.
  • A zero denominator makes the inferred correlation undefined. An implausible result outside the correlation bounds is a reason to inspect coverage, quotes and assumptions, not to clip the value and call it a trading signal.

A two-stock example isolates the correlation effect

Assume a hypothetical fixed-weight basket: A has weight 60% and annualized volatility 20%; B has weight 40% and volatility 30%. Use decimal volatilities and assume the same measurement horizon. The own-variance term is 0.6² × 0.2² + 0.4² × 0.3² = 0.0288. The coefficient on correlation is 2 × 0.6 × 0.4 × 0.2 × 0.3 = 0.0288. No option prices or returns are being predicted.

  • At 20.7846% basket volatility, the reverse calculation gives (0.0432 - 0.0288) / 0.0288 = 0.50. Raising correlation to 0.80 increases index volatility while both constituent volatilities remain unchanged.
  • A related variance-dispersion quantity is D² = sum(wi × sigmai²) - sigmaI². The first term uses weights, not squared weights. Here it is 0.6 × 0.04 + 0.4 × 0.09 = 0.06. At correlation 0.50, D = sqrt(0.06 - 0.0432) = 12.96%; at 0.80, D = 9.03%.
  • D describes dispersion under this model; it is not a portfolio return. DSPX uses its own eligible basket, option-based variance construction and maintenance rules. These two invented stocks do not reproduce that benchmark.
Assumed correlationBasket varianceAnnualized basket volatility
0.000.028816.97%
0.500.0288 + 0.0288 × 0.50 = 0.043220.78%
0.800.0288 + 0.0288 × 0.80 = 0.0518422.77%
1.000.057624.00%

Match the volatility legs before interpreting an edge

A long-dispersion option book often pairs long constituent straddles with short index straddles. First choose the exposure to match: initial vega, cash gamma or variance notional. These choices produce different quantities. Equal premium, equal contracts and equal index weights do not make them equivalent. A vega match only offsets a specified small, parallel change in implied volatility at the measurement instant.

  • These hypothetical Greeks allocate 60% of long vega to A and 40% to B: 1,200 × 0.60 / 24 = 30 straddles and 1,200 × 0.40 / 24 = 20. This is an explicit sizing assumption, not a universal dispersion hedge. Each straddle has one call and one put.
  • If index IV rises by 1 point while both constituent IVs rise by 0.25 points, the frozen-vega approximation is -1,200 × 1 + (720 + 480) × 0.25 = -$900. This excludes delta, gamma, theta, changing Greeks and costs; it is not an exact option repricing.
  • Match expiry horizon and quote time, document strike or delta selection, and inspect skew and term structure. Contracts with different exercise or settlement conventions require separate cash-flow treatment even when their expiry dates match.
  • Record delta hedges and rebalancing as actual transactions. Initial vega neutrality does not guarantee neutrality to correlation shocks, earnings jumps, skew changes, gamma or later volatility moves.
Illustrative initial positionVega per straddleSigned position vega
Short 10 index straddles$120 per 1 percentage-point IV move-$1,200 per IV point
Long 30 A straddles$24 per IV point+$720 per IV point
Long 20 B straddles$24 per IV point+$480 per IV point
TotalAll values include the contract multiplier$0 per parallel IV point

Historical constituents and option chains are required inputs

A credible historical test must reconstruct what was observable and tradable on each decision date. Using the current index members throughout the past introduces survivor and selection bias. Selecting only names with convenient option history also changes the basket; report excluded weight and the remaining basis exposure.

  • Predeclare how to handle a constituent replacement while options remain open: hold the old contracts, close and replace them, or stop new entries. Preserve both the index composition history and the executed book instead of rewriting old trades.
  • Reject new packages with missing or stale required quotes. Retain open-position risk and mark uncertainty when a quote disappears. A missing price is not zero value; a halt is not a free exit.
  • Fix entry selection, hedge schedule, maximum exposure, expiry handling and evaluation dates before testing later data. Include rejected packages and partial fills in the results, with a predefined response to an unmatched leg.
DatasetRequired recordFailure if omitted
Index membership and weightsDated additions, deletions, rebalance effective times, float or share changes, and weights available before the decision.Later constituents or weights enter earlier signals.
Contract historyExpired and adjusted series, underlying identity, strike, expiry timestamp, multiplier, deliverable, exercise style and settlement reference.The model silently trades a contract that did not exist or had different obligations.
Historical option quotesSynchronized bid, ask, quote timestamp and size for every selected leg, including entry, hedge and exit dates.Stale midpoint marks appear as executable relative value.
Model inputs and eventsRates, dividends, corporate actions, earnings dates and the exact volatility/Greek method available then.Model differences or event premiums masquerade as correlation value.
Portfolio accountingLeg fills, hedge trades, cash, fees, financing, margin, assignment, settlement and marks on open positions.A spread chart replaces the cash flows needed to calculate P&L.

Where an apparent dispersion opportunity fails

A rise in common market exposure can make the short index leg lose while the long stock options offer less protection than expected. Separate that risk from paying too much for individual earnings volatility, changing weights, hedge slippage and scarce liquidity. A relative-volatility thesis can be right and still fail to cover the costs of the option book.

  • The sizing example contains 60 straddles, or 120 option contracts. Opening and closing all contracts once creates 240 contract-sides. At an assumed $1 commission per contract-side, commissions alone are $240, before spreads, hedge turnover, exchange charges and financing. This is an arithmetic allowance, not a broker quote.
  • Delta hedging cannot erase an overnight jump at a price that was never tradable. Stress a common market shock, one-stock earnings gap, a correlation increase, wider spreads and an unavailable exit separately.
  • American-style short options may be assigned before expiry. Cash-settled index options and physically settled stock options can have different settlement procedures. Margin demands can force liquidation before the research horizon ends.

Use TradingView for an observable research monitor

Start with a small, explicitly labeled subset of constituents and a standard daily chart. Specify synchronized confirmed closes, the return definition, a common observation window and a documented price-adjustment policy. A monitor can display realized volatility, return correlation and the variance identity above. Label a subset as a subset; do not present it as the complete historical index.

  • A concrete research specification is 60 common daily return observations, sample standard deviations annualized by sqrt(252), and sample correlations from those same dates. The 252-day convention is an assumption. Missing or nonmatching sessions must make the comparison unavailable rather than carrying a stale close forward.
  • Compare 20, 60 and 120 observations as a predeclared sensitivity study, then evaluate later dates without selecting the window from the eventual outcome. Historical price volatility is not implied volatility and cannot reconstruct a missing option chain.
  • Pine Script AI Coding Agent can help draft this monitor from explicit inputs and missing-data rules. Review the generated source and request syntax validation. A successful syntax check would establish syntax only, not the validity of the volatility estimate.
  • TradingView documents a chart-based broker emulator. Requesting another symbol for calculations does not create independent fills in its options. Strategy Optimizer can compare parameters of a supported TradingView strategy, but optimizing a chart proxy cannot validate the multi-leg dispersion book. Keep actual options testing in a separate portfolio simulator with the required history.
Where Pineify fits

Pine Script AI Coding Agent

Draft a realized-volatility and return-correlation monitor with explicit symbols, observation windows and missing-data rules. Keep historical option pricing and portfolio accounting in a separate research workflow.

Also useful: Strategy Optimizer. Compare inputs for a supported TradingView chart strategy. Its results do not supply independent constituent-option fills or a dispersion portfolio backtest.

Draft a research monitor

Related resources

TradingView strategy research guidesBacktest accuracy and fill assumptionsPine Script AI Coding AgentStrategy Optimizer capabilities

Frequently asked questions

Educational research, not investment advice. All example weights, volatilities, Greeks and fees are hypothetical. Dispersion options positions can incur substantial losses, margin demands and execution risk. No historical return, executable trade or complete options backtest is claimed.

Sources and verification

Published by Pineify. Pineify is an independent product and is not affiliated with TradingView. Sources support the technical explanations; examples do not establish trading performance.