Methodology: scope, interpretation and limitations

Half on a Double publishes end-of-day options-positioning reference levels for US stocks and ETFs. This page explains what the public values mean and where their interpretation stops. The calculation implementation, calibration and numerical procedures are proprietary.

Reading the levels

Delta neutral is the model’s reference price associated with balanced aggregate delta across the options chain.
Gamma neutral is the model’s reference price associated with neutral aggregate gamma.
Gamma max is the model’s reference price associated with peak aggregate gamma.
Open interest counts outstanding option contracts. It differs from trading volume, shares outstanding and the number of listed instruments.
Nearest-expiration max pain is the strike that minimizes aggregate intrinsic payout at the nearest listed expiration. All-expiration max pain pools every listed expiration.
Net gamma exposure (GEX) sums modeled gamma weighted by open interest and the contract multiplier at the closing underlying price. Calls are positive and puts negative under the published convention. Call wall and put wall identify the strikes with the largest side-specific GEX magnitude.
GEX flip is published only when the model confirms a sign-changing bracket and a plausible zero crossing; otherwise it is unavailable.

These are model-derived observations, not measurements of any specific market maker’s holdings. The sign convention does not establish who owns the options or the direction of any dealer’s actual hedge. A price near a level, or a recorded cross, does not guarantee support, resistance, a change in direction or a profitable trade.

Dates, coverage and missing values

Use the data-through date on the ticker pages and screens. Data are published after the close and are not live quotes. The publication can lag the latest trading session.

Missing levels are shown as unavailable, never as zero. Thin markets, stale quotes, incomplete chains, model assumptions and processing errors can affect both availability and accuracy. Exercise style, dividends and borrow conditions also limit how closely a simplified model represents actual market behavior.

History and revisions

A methodology revision can change historical values as well as current levels. The active publication and its version are described on the data page. Keep a time series within the same generation; do not join older cached values to a replacement generation without checking its coverage and compatibility.

The September 2026 replacement recomputes the complete history using market-implied carry inputs instead of the legacy fixed 0.25% risk-free-rate assumption. It is one replacement series, not a continuation that should be spliced onto legacy values.

Level distance rankings

The level distance rankings compare gamma max and delta neutral with the close. Signed distance is (level − close) / close × 100; a positive value places the level above the close and a negative value places it below. Absolute distance removes that sign so tickers can be ranked by magnitude.

The historical distance rank compares today with the preceding 252 valid trading sessions, excluding today. It reports the percentage of those sessions whose absolute distance was strictly smaller than the present value. Ties are not counted as closer. Thus 100% means every valid comparison session was closer and today is the most distant observation in that window; 0% means no comparison session was strictly closer. The denominator is displayed because missing levels are excluded rather than treated as zero.

The page publishes both tails of both comparisons: closest and farthest across eligible tickers today, and the lowest and highest prior-session percentages within each ticker’s own history. Secondary distance and open-interest sorting provide a stable order when the primary historical percentages tie.

Public rankings include full-history tickers with at least 100,000 contracts of current open interest and at least 200 valid prior observations. These are descriptive ranks, not probabilities of convergence, reversal, support or resistance.

Open interest change standardization

The unusual-accumulation screen compares today’s absolute change in total option open interest with up to 252 preceding session-to-session changes for the same ticker. Today’s change is excluded from the reference window. The published z-score is (current change − prior mean) / prior sample standard deviation; at least 200 valid prior changes and 100,000 current contracts are required. The screen includes positive changes of at least +2σ and ranks the largest standardized increases first.

This controls for the very different scale and ordinary variability of each ticker’s chain, but it is descriptive rather than a normal-distribution claim. Expiration cycles, corporate actions, symbol changes, data revisions and structural shifts can create extreme observations. An increase means more contracts remained open; it does not reveal direction, owner, opening trade price or whether the activity was accumulation by any particular participant.

Call versus put open interest growth

The call/put growth screens compare percentage changes in outstanding call and put contracts over one or five sessions. A positive call-minus-put gap means calls grew faster; a negative gap means puts grew faster. Growth screens require the leading side to increase, at least 100,000 current total contracts, and at least 1,000 contracts on each side at the start of the comparison.

Separate surprise screens identify unusually large daily call or put contract increases. Relative-shift screens identify unusual changes in the growth gap itself. Each measure uses its own historical reference and requires at least 200 prior comparisons; a difference between two side-specific z-scores is not the relative-shift z-score. Missing or constant reference data cannot yield a standardized surprise. Missing source-session observations are not bridged.

These are whole-chain observations. Expirations, rolls, exercises and corporate actions can change the totals. OI growth does not identify the initiating buyer or seller, and changing call/put OI balance does not establish a change in implied-volatility skew. Neither the scores nor their thresholds are trading probabilities. Results should be interpreted alongside price, liquidity and option premiums.

Advanced options screens and catalyst ranking

Advanced screens use every derived analysis family supplied by the options analytics feed, while keeping the measurements separately inspectable. They cover implied-volatility level and rank, term structure, expected and straddle-implied moves, risk reversal, volatility-smile convexity and surface dispersion, option volume, put/call skew, net and modeled dealer delta, gamma walls, max pain, vanna, charm and related strike/expiration views. The complete arrays are linked from each covered ticker; the screens use normalized scalar projections so tickers can be compared.

The catalyst-intensity score is a cross-sectional rank from 0 to 100. It gives equal weight to four groups: volatility level, front-versus-back term premium, volume/open-interest activity, and skew/smile shape. Correlated measurements are averaged within a group before the groups are averaged, so several views of one volatility surface do not receive several independent votes. The separate directional score summarizes relative delta, modeled dealer delta, risk reversal and put/call positioning on a -100 to +100 scale. Neither score is a probability, return forecast, trade recommendation or proof that a corporate event exists.

Only liquid names with at least 100,000 current open contracts are eligible for these rankings, and only when the advanced feed date exactly matches the core publication. The analytics producer currently computes the complete suite for its high-open-interest universe, so absence from an advanced screen can mean out-of-universe or unavailable inputs rather than an ordinary reading. As with the core fields, missing values are not treated as zero.

Long-option research screens

Eight research screens compare forecast volatility with option-implied volatility, historical payoff scenarios with premium cost, price breakouts with premium value, call and put wing skew with their own histories, IV expansion from low levels, unusual activity with subsequent OI confirmation, forward-volatility windows with their histories, and individual contracts under defined price/time/IV scenarios. Comparisons use corresponding maturities and quote-quality controls. Historical references exclude the publication day; missing or insufficient inputs are withheld, not replaced with zero.

Historical payoff cohorts are descriptive and can be small. Vendor price histories are not independently corporate-action-adjusted; discontinuity checks cannot eliminate all split, dividend or revision effects. Selecting a favorable historical contract does not establish a positive expected return.

Contract scenarios show a seven-calendar-day holding period, favorable and adverse 5% underlying moves, and unchanged or plus/minus five-percentage-point IV. Costs reflect the observed EOD ask, a spread allowance on modeled exits and $0.65 per contract each way. These are European-style model approximations, not executable quotes; American exercise, dividends, borrowing and changing spreads can materially alter outcomes. The scenario inputs are assumptions, not predicted moves. A cheap forward-volatility interval is not the price of a standalone long option.

These screens are experimental research filters, not backtested profitable strategies or buy recommendations. They do not identify buyer-initiated trades or verify an earnings catalyst. Long options can lose the entire premium. “Results unavailable” means the necessary dated inputs have not been supplied; it is distinct from a completed screen with no qualifying matches.

Historical cross outcomes and confidence

Ticker pages group gamma-max crosses by direction. A qualifying event changes from one side of gamma max to the other between published sessions, with gamma max no more than 10% from the close on both crossing sessions. Each path is rebased to 0% at the cross close. “Day +N” is the close-to-close return after N subsequent trading sessions, not N calendar days. Events without enough subsequent sessions are omitted from that horizon rather than treated as zero.

The displayed median is the observed middle return. The accompanying interval uses order statistics and has nominal coverage of at least 90% for the historical population median under independent observations. The confidence floor is the interval bound closest to zero when the entire interval is positive or negative; it is zero when the interval spans zero. It is a conservative companion to the median, not a replacement for the observed result or a probability that the next return will have the same sign.

Crosses can occur close together, so their forward windows can overlap and are not independent observations. The interval and floor are therefore descriptive uncertainty guides, not formal coverage guarantees or tests of statistical significance. Corporate actions, changing volatility, market regimes and revisions to the underlying level series also limit comparability. These summaries describe the available history and are not forecasts or trading recommendations.

Public documentation

The learning guides explain the vocabulary and how to read the published output. The agent directory provides dated text summaries of the same public results. These resources do not publish the calculation recipe or internal inputs.

This site provides educational, descriptive data, not investment advice. Read the disclaimer and data and reuse terms before relying on or redistributing an observation.