Signal Glossary

Counted per contract by the options clearinghouse and published once a day, overnight. Any open interest figure you see during a session therefore reflects the close of the prior session. Our morning checks and the Level Map's morning refresh wait until the new figures are actually published before using them. A contract missing from a day's data is treated as unobserved, never as zero, so an expiry newly entering our data cannot look like a sudden build.

It is a net count. A rise means more contracts were opened than closed, not who opened them or whether they were bought or sold to open. Exercise and assignment also change it, and today's activity only shows up in tomorrow morning's figure.

Volume resets every session; open interest does not. Premium uses the option's own price. Notional uses the stock price instead (contracts x 100 x share price) and describes the size of the exposure, not money paid, so it is always far larger for the same trade. If a contract did not trade today, our data provider still reports its last trading day's figures; in most of our calculations we count today's volume as zero rather than carry the old number forward.

A large premium total says money changed hands, not who paid it or why. One trade can be a leg of a spread or a hedge against stock, so a big number on its own is not a directional view.

Backed out of each option's market price with a standard pricing model; our data provider supplies a value per contract. Our at-the-money (ATM) reading averages the call and put IV at the strike nearest the stock price, on the nearest expiry at least two weeks out, so very short-dated contracts do not dominate it. IV Rank, IV Percentile, skew and term structure, in the options section, are all built from these values.

It is the market's price for uncertainty, not a forecast of direction. Implied moves often run larger than the moves that follow, and IV usually drops once a known event such as earnings has passed.

The usual form is a z-score: today's value minus the stock's recent average, divided by its typical variation over the same sessions. Some signals use a percentile within the stock's own history instead, or a median-based version that a single extreme day cannot distort. Until a stock has enough history, the reading is left blank or marked Limited History or warm-up rather than estimated. Under a normal distribution a reading above +2σ happens on about 2% of days.

A high reading means unusual for this stock, not large in absolute terms, and neither good nor bad. Market data has fatter tails than a bell curve, so big readings appear more often than the textbook rate. A new listing or a data change can shift the baseline.

Computed per contract by our data provider from a standard option-pricing model. Delta also works as a rough moneyness scale, which is why skew is read at the 25-delta options on each side. Gamma matters for hedging: a market maker who hedges with stock must buy or sell more shares as gamma changes delta, and that is the basis of the dealer gamma estimates on the Level Map. Contracts whose Greeks cannot be computed, mostly deep in the money, are left out of our stored option data.

They are model estimates that change constantly with price, time and volatility, and they inherit the pricing model's simplifications. A Greek describes sensitivity, not what any trader intends to do.

Intraday groups contracts as Today / this week (0-7 DTE), Swing (8-30 DTE) and Longer-dated (31-90 DTE). The Flow Log hides contracts under 7 days by default and keeps LEAPS, which in our product means over 120 days, behind their own toggle. Level Map calculations use the stored chain out to about four months. Standard monthly expiries fall on the third Friday; weekly and daily expiries fill the gaps on many tickers.

Expiry groups are a convenience, not a statement about intent: a 10-day option can be part of a long-term strategy. The textbook definition of LEAPS is more than a year out; ours is shorter.

The sum of price times size for every trade in the session, divided by the total size. Our data provider reports a session VWAP for each stock and each option contract. We use it as a reference point: where a large print sat against the day's average, and as the price in estimated option-volume dollar totals where exact trade prices are not needed.

A print above VWAP is not proof of a buyer, and one below it is not proof of a seller. Where we have it, we show where a print landed against the quote at that moment instead, and label it as such.

Off-exchange prints are identified by their reporting venue on the consolidated trade feed. We measure the share of each stock's volume that traded off-exchange, compared with that stock's own history, and track the largest individual prints by dollar size.

A report shows price, size and time. It never shows the buyer or the seller, or whether the trade opened or closed a position. Retail orders filled internally by brokers also count as off-exchange, so a higher share is not automatically institutional.

One computation feeds both versions; only the wording and the amount of technical detail change. On the dashboard a toggle switches between them. For email, your edition is set in your profile, and the daily, weekly and free emails each come in both.

Regular view rounds and simplifies. When a precise value matters, such as an exact level or a z-score, Advanced view shows it.

After the close our pipeline computes every metric, then gives Claude the results, the stock's recent history, our prior notes, and the day's news and calendar. It writes the market read, each stock's note and the Key levels rows (Higher, Lower, Read is wrong), and assigns the direction, conviction and change tags. Automated checks then test the text against the data, such as level prices against the close, and fix or remove failing sentences. Next session, code checks the prior read's line against the close.

The direction tag and, usually, the levels picked for Key levels are the AI's judgment, not a model output, and can be wrong. When it names no levels, the card shows computed support and resistance instead. One AI-assigned label also feeds a single Reversal radar condition. The commentary never knows who traded.

Four pillars, scored daily from price and moving averages: Maturity (how long the trend has held), Stretch (how far price sits from its 21-day average), Slope (whether that average is rising or falling) and Stack (whether short, medium and long timeframes agree). They combine into a signed score and one of seven labels. The read flips sides only after three straight closes beyond a 0.25% band around the 21-day average, so a one- or two-day dip reads as a trend holding through a pullback. The market reading is a weighted blend of six index funds, including SPY, QQQ, IWM and RSP.

It describes the trend; it is not a forecast or a buy signal. By design it is slow at absolute tops and bottoms, so it confirms a turn after it starts. It is unrelated to the Low, Medium or High conviction attached to a written read.

Claude, the AI model we use, writes it after every number is computed, working only from that day's metrics: index trends, cross-market reads, sentiment, positioning, key levels and notable prints. It computes no score of its own. Written rules constrain the wording, keep unexplained jargon out of the Standard edition, and require it to say how the read compares with the last one: New read, Strengthening, Weakening, Unchanged, Reversing or Conflicted. Signal Lab shows the same narrative with a count of recent regime days.

It is interpretation, not a forecast or advice, and it can be wrong. The level to watch is a reference point, not a target. Its conviction label is the writer's confidence in the read, not the Conviction Score.

Six factor blocks: Credit, Volatility & Dealer, Institutional Flow, Breadth & Structural, Cross-Asset, and Curve & Rates. Each input is normalized against its own recent history and averaged within its block, and the blocks combine with fixed weights. Pulse is the fast read (1 to 3 days), smoothed asymmetrically: it moves quickly toward fear and slowly toward greed. Bands use hysteresis so a score near a boundary does not flicker. Despite its name, the flow block uses options and trade-flow proxies, not identified institutions.

It describes sentiment and has never forecast returns. A fear reading is not a buy signal and greed is not a sell signal. As a fast read, it can swing on a single day of stress.

Per contract, opened sides equal the session's volume plus the overnight change in open interest, and closed sides equal volume minus that change. This is exact arithmetic on published data, not an estimate. Contracts whose open interest moved by more than their volume (exercise or assignment) are left out. Opened puts and calls are summed across tracked single stocks, and the put share is compared with the previous 15 qualifying sessions. Nothing shows until enough stocks report and enough history exists.

It does not show who opened the positions or whether they were bought or sold. It always describes the previous session, because open interest publishes the next morning. Index funds are not included.

Built from the same block scores as Pulse, with slower, symmetric smoothing (a 10 to 15 day read) and several sessions of confirmation before its band changes. The two gauges differ in speed, not in inputs.

Because it uses the same inputs as Pulse, agreement between the two is not independent confirmation. It lags turns by design and has never forecast returns.

A simple agreement count across same-day moves: the S&P 500 fund's direction, credit (a high-yield bond fund against stocks), gold against stocks, and a bitcoin fund against stocks, with the yield curve's shape feeding the stagflation check. A named regime needs several inputs to agree, and moving from one named regime to another takes more agreement than holding the current one, which limits day-to-day flips.

It reads one day's moves, so it can change quickly. When markets disagree, it does not tell you which one will turn out right. It is not a forecast. The US dollar is not currently an input.

The label compares the day's move in a high-yield corporate bond fund with the S&P 500 fund. The Advanced figure is the standard option-adjusted spread (OAS) of a broad high-yield bond index over Treasuries, in basis points; a wider spread means lenders demand more to hold risky debt. The small chip counts consecutive sessions in the same state.

It is a one-day comparison and can flip often. Credit can lead or lag stocks; a Warning is a prompt to look closer, not a sell signal. The spread series is published with a short delay.

2s10s is the 10-year Treasury yield minus the 2-year yield, in basis points. Steepening means the gap is widening and flattening means it is narrowing; in the textbook terms, bull means it happens with yields falling and bear means with yields rising. The Curve label reads the shape across 3-month, 2-, 10- and 30-year yields: Inverted when the 3-month yield sits above the 10-year, Flattening when the gaps across maturities are all small, otherwise Neutral. Steepening, bull and bear are the textbook terms; the label does not show them.

Inversion has a long but imperfect record as an early sign of slowdowns, with very uncertain timing. The curve moves slowly and says little about the next few sessions.

Built from families we already compute: directional options proxies for the S&P 500 fund and for the largest stocks, options-implied demand for volatility, and how broadly unusual activity is spread across stocks and which way it tilts. Market context adjusts the result. The Index and Stock lines are sub-scores of the same composite, and the four named patterns are rules that combine the composite, its sub-scores and price.

Despite the product labels, it does not identify institutions or see whether anyone bought or sold. Its inputs are proxies built from price, volume and options data, so Buying and Selling describe the proxies. Its outcomes are graded; no lead over price has been established.

Cboe computes VIX, VIX9D and VIX3M from S&P 500 index option prices at those horizons. The Signal Lab chart labels the curve Inverted when VIX9D is above VIX, Contango when VIX3M sits clearly above VIX, and Flat otherwise. The email's 9D/3M figure is VIX9D divided by VIX3M.

It describes how options are priced right now. An inversion can resolve within days or persist through a selloff, and the curve does not say which way the index will move.

Each flag is a rule on standard building blocks, checked on the index funds. RSI diverging from price drives Momentum bottoming and Momentum fading. The put/call ratio at an extreme of its own history drives Extreme fear in options and Complacency in options. VIX rising alongside the index drives Insurance buying up. Up-day and down-day volume drives the other four. Flags fire only once a trend label has held for a few sessions.

They are context, not signals with a proven record: no flag has shown a measured edge. The names describe patterns in price, volume and options data; Quiet selling at highs and Insurance buying up do not identify who traded.

Three families of stress conditions: how the S&P 500 itself is trading, strain between stocks and other markets such as credit and the dollar, and the volatility market. Advanced view shows a 0-100 score. It rises most when stress shows up in more than one family at once, fades gradually, and needs confirmation before its band changes.

It is experimental. Its outcomes are recorded and graded, and no edge has been established yet. Elevated is a reason to look closer, not a sell signal, and Calm does not mean low risk.

Eight conditions per side, each a rule on data we already compute, drawn from market-wide positioning, options pricing and estimated dealer gamma, credit and other markets measured against stocks, volatility, and how far the index sits from recent highs and lows. The label reflects how many conditions are active, with extra weight when they persist across sessions.

So far it has run close to chance, with no measured edge yet. Treat it as a checklist, not a forecast. Setup confirmed means many conditions are active at once, not that a reversal is under way.

Each pattern compares one block of the Pulse composite with the overall reading, for example the credit block reading fearful while overall sentiment reads greedy. An alert needs two or more blocks to disagree, and broader disagreement raises it to Strong divergence. Patterns that test poorly can be switched off.

It is a watch signal, not a buy or sell signal. Alerts are rare and none has a track record yet. Dealer regime flip imminent refers to a modeled estimate of dealer positioning, not a known event.

Based on each fund's estimated dealer gamma exposure, compared with that fund's own history. When dealers are estimated to be long gamma, their hedging tends to damp moves (Coil); when they are short gamma, hedging can amplify moves (Expand). Readings close to normal show No clear read.

Dealer positioning is modeled from open interest, not observed. It says nothing about direction, covers only these three index funds, and single stocks can behave differently.

For each fund and expiry window, the short-term change in total open interest across the whole option chain is compared with that fund and window's own history, and only unusual changes show. Call-led or put-led appears when one side accounts for most of the change. Open interest is reported the morning after, so each reading reflects the prior session.

It measures inventory change, not direction: a put build can be a hedge, and either side of a trade can open a position. Near-term changes are often event or hedge flow. It does not show who opened or closed the contracts.

Each closing-auction print is ranked against that stock's own closing prints in our records, which reach back up to a decade; a record close is the largest on file. The day's count of top-ranked closes against a trailing baseline sets the Tier, and notable prints are grouped by sector to measure concentration. The Flow label (Accumulation, Distribution or Two-sided) describes how those stocks' prices behaved into the close.

Accumulation and Distribution describe price action around the closing print, not who traded or which side they took; every closing print matches buyers and sellers in equal size. Rebalance-day volume is largely mechanical.

VIX is Cboe's 30-day implied volatility index for the S&P 500. HY OAS is the option-adjusted spread of a broad high-yield bond index over Treasuries. 2s10s is the 10-year minus the 2-year Treasury yield. Gold shows the price of a large gold ETF, not spot gold. WTI is the published daily crude oil spot price. USD is the Federal Reserve's broad trade-weighted dollar index.

These are reference values, not signals. Some series arrive late: the dollar index is published weekly, and crude and the credit spread can be a day behind.

Four families of inputs, each measured against the stock's own recent history: share volume, how options are priced for downside versus upside (skew), how busy the options market is relative to the stock, and how rare the day's largest trades are. Unusual activity that persists for several days adds to the score. The inputs are weighted from outcome testing, then calibrated so a given score means the same thing over time. Earnings windows and mechanical volume days, such as option expirations and quarter ends, are dampened.

It is non-directional: a high score can come before a drop as easily as a rally. It does not identify who is trading. Index ETFs use an older recipe, so an index score is not directly comparable with a stock's.

A display rule on the current FlowScore formula for individual stocks. Because FlowScore is calibrated to a fixed scale, the badge means the same thing from one day to the next. Index ETFs and stocks still on an older score version do not get the badge. Every flagged stock's move over the following sessions is recorded and compared with same-day peers.

Non-directional, and the current formula's record at this level is still short. It flags attention, not a trade. A stock can carry Move Watch for days without a large move.

The change in FlowScore over exactly five trading sessions, shown only when the rise is large and today's score is already meaningful. Falling scores are tracked as well but are not badged. On Today's Flow Leaders, a fast-rising score can lift a stock above one with a slightly higher but flat score.

It describes a change in activity, not direction. A surge can come from one unusual day that fades quickly. It applies to individual stocks only.

Individual stocks only; index and leveraged ETFs are left out. Stocks are ordered by FlowScore, with credit for scores rising quickly over five sessions. The reason line names the FlowScore input contributing most (volume surge, skew shift, options activity, rare tape or persistent flow) or the five-session change. Badges show Move Watch, Flow Surging or Flow Building when they apply.

A ranking of attention, not a list of trades. A stock can lead because of one rare print or a short burst of options activity. It says nothing about direction.

Single prints only, ranked by dollar value (shares times price). Opening and closing auction crosses are identified by their condition codes and timing and kept out of this list; closing-auction prints have a ranking of their own. Clusters of smaller trades are not ranked. The list is updated by scans during the session and again after the close. Headline callouts also require a minimum dollar size, so a thinly traded stock cannot top the list with a small trade. Each print is labeled off-exchange or exchange.

A large print shows size, not intent. We cannot see who traded or whether it was a buy or a sell. A newly tracked stock's prints are ranked only once its ten years of history has been loaded.

Four scans through the session each cover their own time window, with no overlap, so a trade is counted once. Single prints and closing-auction prints are ranked in separate lists. On Signal Lab, Ranked watchlist trades shows the week's standouts: the top 25 in 10-year history for watchlist names and the top 50 for SPY, QQQ and IWM. On the Intraday tape, a link icon marks tickers our options scanner also surfaced that day.

Rows show size and venue (dark or lit), never the buyer or seller. Late-reported trades can surface after a scan. Large closing-auction trades are routine for many stocks, which is why they are ranked separately.

Share is off-exchange volume divided by total volume, from full-day data. Elevated compares today's share with the stock's own history after removing the part explained by market-wide volatility, since off-exchange share tends to shift when volatility does. Persistent means elevation on most recent sessions. VWAP bias compares off-exchange dollars printed above versus below the day's volume-weighted average price. We also measure how much of the activity came in block-size prints.

Off-exchange reports do not show the buyer or the seller, and a print above VWAP is not a purchase. Much off-exchange volume is routine order routing. After a data improvement in September 2026, stocks whose off-exchange share stepped up are compared with readings taken since then.

Relative volume = today's share volume / average daily share volume over a recent window. FlowScore uses its own version of this baseline, weighted toward recent days. The RVOL figure on Flow Radar cards uses a plain trailing average.

Volume says how much traded, not who traded or which way. Earnings, option expirations, quarter ends and index rebalances can lift it with no change in the story. It means less in a stock's first days of trading.

Options-to-stock ratio = option contracts traded x 100 (shares per contract) / shares traded. Today's ratio is divided by its trailing average, which is the "20d" in the label. It is one of the readings behind FlowScore's options-activity family.

It counts contracts, not dollars or direction, and cannot separate opening from closing trades. Expiration days and earnings weeks raise it for mechanical reasons.

Days to cover = shares sold short / average daily share volume. The data are the exchange-reported short interest for the mid-month and month-end settlement dates, published with a lag of about a week or more. We compare days to cover with the stock's own history and track the change from the prior report. Individual stocks only. Short interest also feeds the Squeeze setup market state and Signal Agreement.

The figure is stale when published and misses shorts opened since. High short interest is not a squeeze forecast. It does not include short exposure held through options or swaps.

Up to six indicators can vote: options skew, the options price proxy, off-exchange price location, open interest change, activity in related leveraged and inverse ETFs, and short interest. Each votes only when its own reading clears its rule. Votes are weighted, and the score is the share of weight on the majority side.

Agreement is not accuracy: our grading has not shown a measurable edge for strong agreement so far. It says nothing about the size or timing of a move.

A rule-based classifier over readings described elsewhere in this glossary: the options price proxy, Volatility Demand Score, skew, the estimated dealer gamma regime and how concentrated it is by expiry, off-exchange price location, open interest trends and short interest. Each label has its own conditions, checked in a fixed priority order. A Leaning label marks a partial match, and Neutral means none applies.

Labels rest on price-location proxies and estimates: Accumulation and Distribution do not mean buying or selling was observed. Breakout Fuel can resolve in either direction. Labels can flip from day to day near their edges.

Today's FlowScore is compared with the stock's own FlowScore history as a standard score. When it is unusually high, the card adds context from the recent price trend and the options price proxy, and notes nearby earnings.

A description of unusual activity in context, not a forecast. It does not show who traded or which side: the card says "Buyer or seller unknown".

Each contract's volume is signed by its day close versus its day VWAP, then weighted by how much stock exposure the contract carries. The day's total is compared with the stock's own history as a standard score. The watchlist detail's Options price model row reads a closely related premium-weighted version.

It does not identify customer side: a close above VWAP is not proof that anyone bought. Contracts that trade little give weak readings. It is a one-day read and can reverse the next session.

Uses the same close-versus-average signing as the options price proxy, but weights each contract by its sensitivity to volatility (vega) rather than its stock exposure. The total is compared with the stock's own history as a standard score. It is one input to the Fear bid market state.

Like the price proxy, it does not show customer side or whether positions opened or closed. Demand for volatility can come from hedgers as easily as from speculators.

The same four parts as the market Conviction Score, applied to the stock's own prices: maturity (how long price has held one side of its trend average), stretch (how far price is from its 21-day average), slope (whether that average is rising or falling) and stack (whether shorter and longer averages agree). A change of side must be confirmed before the badge flips, so it moves slowly by design.

Trend-following by design: it describes the trend in place, not a turn ahead. It lags at reversals, and a test of the trend is not a prediction that it will turn.

Many of our measures compare a stock with its own history, and options history cannot be bought after the fact, because historical options-chain snapshots are not sold. A new ticker's options history therefore starts on the day we begin tracking it. The notice shows until the stock has about five trading days of our own analysis.

Readings during warm-up are less reliable, and measures that need a longer history, such as off-exchange elevation, take longer than five days to settle.

Earnings dates come from a public earnings calendar; expiration is the standard monthly expiry on the third Friday, counted in trading days. Several of our models adjust for these windows so a mechanical spike is not read as unusual activity. Flow Radar cards show whether a report comes before the open or after the close.

Dates can move, and some companies confirm late. A chip is context only: it does not say how the stock will react.

Contract volume divided by the open interest reported for the prior session's close. Open interest is published once a day, the morning after, so the denominator is always the previous session's figure. Intraday readings use the volume of the scan window they come from, so readings from different windows are not added together. Closing trades can only remove contracts that already exist, which is why a ratio well above 1 points to opening activity.

It is evidence about a contract, not about any one trade. Contracts bought and sold again within the same session add volume without changing open interest. It says nothing about who traded or which side opened.

Built from the intraday scans. A contract qualifies when it is 8 to 90 days from expiry and its volume within one scan window is at least 2× the open interest before the session. Each contract is counted once, at its strongest scan, then ranked by that multiple and by premium. The next evening we read the new open interest and record whether the position opened. Puts open at about the same rate as calls.

Opening is measured for the contract, not for any single trade. It does not say who traded, whether they bought or sold, or where the price goes next. A small sample so far.

Compares each contract's open interest with an earlier snapshot. The size of the change is judged against this stock's own normal range of changes for contracts with similar time to expiry, using an outlier-resistant score, and ranked within today's option chain. Changes that persist score higher, and quality checks discount known distortions such as earnings and expirations. A contract missing from an earlier snapshot is never read as zero open interest.

Open interest cannot say who opened the contracts or whether they bought or sold them. It reflects the prior session's close. A rise does not establish that any particular print opened a position.

By volume: total put contracts traded divided by total call contracts traded, across the stored chain out to about 120 days. By open interest: total put open interest divided by total call open interest. We also compare the volume ratio with the stock's own recent average, because what counts as high differs a lot between names.

A put can protect shares as easily as it can express a downside view, and a call can be sold as easily as bought. The ratio alone does not say which. Open interest reflects the prior session.

IV Rank = (today's IV - lowest IV) / (highest IV - lowest IV) × 100, over the history we have recorded for that ticker, up to about a year. Today's IV is the average of the call and put implied volatility at the strike nearest the stock price, on the nearest expiry at least about two weeks out. There is no smoothing.

It does not establish whether an option is cheap or expensive, only where IV sits in its own recorded range. For many tickers that history is still shorter than a full year, so the range is narrower than the textbook version. It does not adjust for earnings.

Count of recorded days with at-the-money implied volatility at or below today's value, divided by the number of recorded days, times 100. Uses the same IV series and history as IV Rank.

Same history limits as IV Rank: for many tickers it covers less than a year. It describes IV's position in its own history, not whether options are fairly priced.

Skew = implied volatility of the 25-delta put minus implied volatility of the 25-delta call. Both are read from a cubic spline through the delta/IV curve on the nearest expiry at least about two weeks out, fitted separately for calls and puts. Skew Z compares today's skew with the stock's own recent history as a z-score.

Skew shows how options are priced, not who is trading them. A wide skew can reflect hedging, supply from sellers, or an event, and it does not forecast direction.

Takes the recent change in 25-delta skew over several sessions and compares it with the distribution of past changes of the same length for that stock, as a z-score. A reading beyond a fixed band is labelled put or call acceleration; otherwise stable.

It describes a change in option pricing, not a trade. It needs enough history to judge what is unusual, and it is not a directional forecast.

Compares at-the-money implied volatility on the stock's first two monthly (third-Friday) expiries at least a week out. Back-month minus front-month IV gives the slope: positive is contango, negative is backwardation, near zero is flat. We also flag the first day the regime changes and count how many days it has lasted.

Backwardation says near-term movement is priced at a premium, not why. Earnings and scheduled events often cause it. It does not say which direction a move might take.

For one contract over one session: opening sides = volume + change in open interest; closing sides = volume - change in open interest. Both are exact. The session's volume is paired with the open interest published the following morning, which reflects that session's close. Contracts whose open interest moved by more than their volume changed through exercise or assignment, not trading, and are left out. Totals are summed by calls, puts and strike. Published each morning once the new figures have landed.

It counts trade sides, not people, and cannot say who bought or sold or which side of a trade opened. It describes the previous session, not today. It is descriptive and makes no forecast.

A row passes the default view when the contract has at least 7 days to expiry and its volume is at least 2× the open interest before the session. Contracts under 7 days sit behind their own toggle. Open interest is read from the ticker's own stored snapshots for the session before and the session after. A contract missing from the prior session's stored chain is never treated as zero open interest: it fails every view except Everything.

The opening check is about the contract, not proof that any one trade opened. It does not show who traded or which side they took. Intraday rows are provisional until the session's own open interest is on record.

Each contract from the intraday scans is assigned a band by its days to expiry on the scan date. Longer-dated contracts come from a separate chain fetch so they never change what the shorter bands show. Ranking within a band uses the same rules as the rest of Notable options flow.

Opening rates differ by band, but none of them says which way the price will move. Contracts more than 90 days out are not in these chips; see LEAPS flow.

For each contract on the best trades list we sample its largest single-leg prints in its strongest scan and compare each with the national best bid and offer in force when the print was reported. The two numbers are the shares of matched volume at or near the ask and at or near the bid. Quotes older than a short limit are not used. Multi-leg and stock-tied trades are left out, because they have no side of their own.

Location is not who initiated the trade, and it is never a bought or sold label. It covers a sample of prints, at report time rather than execution time. When too little volume can be matched, it says so instead of guessing.

Scans cover the first hour, midmorning, midday and last hour. The market's most active stocks are screened, then each candidate's option chain passes contract-level checks on size, liquidity and distance from the money. Survivors are ranked on premium relative to that stock's own recent notable premium, how deep qualifying activity runs across the chain, sweep activity, recent volume acceleration, Vol/OI and early open-interest build. Prints that look like routine hedges are filtered out. There is no directional input.

This is an attention ranking, not a forecast. Customer side and whether trades opened positions are unverified. The underlying price shown may lag option activity.

For each stock and each type, we count contracts that individually pass checks on size, premium, open interest and distance from the money and trade heavily against their open interest. A stock is listed when those contracts span several strikes and expirations and one type takes most of the qualifying premium. Checks tighten around earnings, when option prices are inflated. Windows build cumulatively from the open.

It shows where contract activity concentrated, not who bought or sold or whether positions opened. Several contracts can belong to one spread or roll.

Same-day-expiry volume divided by volume across SPY's complete options chain, every expiration including LEAPS. Measured at several checkpoints through the day; the post-close value is the final one. Each checkpoint is compared only with the same checkpoint on prior clean sessions, never with an end-of-day figure. Data-quality tripwires check that cumulative volume behaves as it should between checkpoints and grey out a reading when the capture looks incomplete.

Intraday readings are provisional, because late-reported trades keep arriving. A high share often lines up with event days but does not say which way anyone is positioned. Tiers wait until enough history exists.

Each print is assigned by rules on trade size, premium, delta and whether it was part of a sweep. Per group, the call/put mix score is shrunk toward zero when the sample is thin, so a handful of prints cannot swing it wildly. The ratio of contracts between the two groups is compared with each ticker's own recent norm, and the heatmap shades how far today sits from it.

These are size proxies. They do not identify institutions or retail traders, and the size rules have not been validated as identifying who traded. A call-heavy score is not a bullish signal, and no group's side is known.

Reported intermarket sweep prints carry trade condition 219. We group nearby condition 219 prints in the same contract across venues into one sweep event, which feeds several of our flow screens. Flow Radar's Last: ISO tag only marks that a contract's latest print carried the sweep condition. Blocks are the largest single-leg prints in a contract. Both are drawn from the trade tape for the most active contracts on each ticker.

Close timing does not prove that one customer placed every print in a group, and there is no shared order ID. Neither label says who bought or sold. A sweep can open, close or hedge a position.

Combines four families: how rare the premium is against that stock's own history, whether the strike sits near the structural levels shown on the Level Map, structural features of the activity itself, and repeat activity in the same name over recent sessions. The result is scaled 0-100. The Flow Log's NFS column shows it.

It is an attention ranking with no directional claim. It does not identify who traded or which side, and it does not prove a position opened.

It starts from an unusual build-up of new open interest, either at a single strike or across a group of nearby strikes on the same side. Several independent checks, drawn from options pricing, options activity and off-exchange trading, must then point the same way before a stock is flagged. Filters hold it back around scheduled events and when the data behind it is thin or distorted.

No stable measured edge yet, so it is graded but not proven. Customer side and whether prints opened or closed positions are unverified. Read it as a reason to look closer, not a signal to act.

A separate nightly pass reads every tracked ticker's chain beyond about 120 days. Only rows with a confirmed bar for that session and large size or premium are kept; the full long-dated chain is not stored. The next evening's pass fills in next-session open interest for those rows, and a contract missing from a chain that was read is marked absent, never zero.

We have not yet measured how often these rows go on to open positions, so the Flow Log's opening rates do not apply to them. It does not identify who traded or which side.

For each strike above the current price we add up gamma times open interest across every call expiry in our stored chain, out to about four months, and take the largest. Weighting by gamma favors contracts near the money and near expiry, where hedging activity concentrates, over far-away strikes with large but inactive positions. Open interest is the prior session's; on paid accounts the level is recomputed each morning once new figures publish.

It is not a ceiling. Price can trade straight through it, and the wall can move day to day as positions open and close. There is no distance limit, so on a thin chain it can sit far from price. It says nothing about who holds the calls.

For each strike below the current price we add up gamma times open interest across every put expiry in our stored chain, out to about four months, and take the largest. Weighting by gamma favors contracts near the money and near expiry over far-away strikes with large but inactive positions. Open interest is the prior session's; on paid accounts the level is recomputed each morning once new figures publish.

It is not a floor. Price can fall straight through it, and the wall can move day to day. There is no distance limit, so it can sit far from price. Heavy put open interest does not say whether those puts are protection, income trades or directional trades.

Standard gamma exposure (GEX): for every contract, gamma x open interest x 100 x stock price squared x 0.01, positive for calls and negative for puts, summed over the stored chain. It reads as the dollars of stock dealers would trade for a 1% move, under the common assumption that dealers are net long calls and net short puts. Advanced view adds the day-over-day change; both views flag when the reading is unusual against the stock's own past 20 sessions.

Dealer positions are assumed, not observed, so the sign can be wrong. It describes conditions, not direction, and uses the prior session's open interest. The badge's second half (Exposed, Cushioned, Fragile, Protected) comes from a separate net put delta estimate that is being recalibrated; treat it as provisional.

Using the per-strike gamma exposure described under Dealer gamma regime, we sort strikes from lowest to highest, keep a running total, and find where it crosses zero, interpolating linearly between the strikes on either side. With several crossings we use the one nearest the current price, and we do not show a flip that sits implausibly far from it. This is a strike-by-strike method; some vendors instead reprice the whole chain at hypothetical stock prices, which can give a different level.

It is an estimate built on an assumption about who holds which options, which nobody observes. Price is not drawn to it, and crossing it does not predict direction.

Textbook calculation on a single expiry: the nearest one at least two weeks out, so it is often but not always the monthly. Each listed strike is tried as a closing price; for each, we add up every call's and put's intrinsic value at that price times its open interest, and take the strike with the smallest total. Open interest is the prior session's.

It is a reference point, not a target: evidence that prices drift toward max pain is weak and mixed, so our email commentary does not treat it as support or resistance. It covers one expiry and ignores what each position cost.

Range = last close x ATM implied volatility x square root of (days / 252), using 1 day for today and 5 trading days for the week. This is the standard one-standard-deviation move implied by option prices, centered on the last close and drawn as shaded bands on the level stack.

It is the market's pricing, not a forecast, and implied moves often run larger than realized ones. Bigger moves happen more often than a bell curve suggests, especially around earnings. It does not lean up or down.

Early each morning, once new open interest is published and a check confirms it has rolled over for that stock, we recompute the walls, gamma flip, max pain and dealer gamma estimate from the live option chain. Three more passes during the session refresh gamma and the stock price; open interest is fixed for the day, so those changes come from gamma and price alone. Z-scores and open interest signals keep their close values and are marked at close.

The stock price used may be delayed (spot may lag), and the open interest is always the prior session's close. A refresh is a recomputation, not new information about who traded. Free accounts and the sample page show end-of-day levels.

We split the dealer gamma estimate by expiry, take the monthly third-Friday expiry's share of gamma across all expiries, counting each expiry's size regardless of sign, and subtract its signed amount to project the regime after it expires. The panel appears only in the last sessions before a monthly expiry, and only when that expiry carries a large share of the total. A regime flip means the projected total changes sign.

It assumes nothing else changes: new positions, rolls and price moves after expiry can all change the picture. It inherits every assumption of the dealer gamma estimate and is not a forecast of next week's moves.

Built mechanically by the page, not by the AI. The candidate levels are the walls, gamma flip, max pain and the leading open interest signals, sorted by price. When price sits above or below all of them, or a side has only one level, the current price becomes one end and a fixed percentage, named in the row label, stands in for the missing level. Reward vs Risk is target distance over stop distance.

These are structural reference points, not recommendations, and the page says so. The levels themselves ignore your own entry, position size and how often each level has held before. Levels shift daily.

We compare today's computed values with the prior session's: gamma flip, call and put walls, the dealer gamma sign and net put delta sign, FlowScore, the DFS and VDS z-scores, and the Market State label. A level is listed only when its move is large relative to the stock price, so rounding noise stays out. Once the morning refresh has run, moves are measured since the last close.

A level moving up is not bullish and moving down is not bearish; it means open interest or gamma shifted. Changes rest on the prior session's open interest, so they can lag what is happening today.

Both groups run after the close over thousands of listed stocks, first on price, volume and liquidity, then on each survivor's options chain; Big Movers also runs after the open and at midday. Every hit is recorded and its forward returns graded against controls, including hits we chose not to show. Reversal Watch shows a hit only when the stock's options are deep and actively traded that session, and at most a handful a day. In emails a stock that qualifies for both groups appears once, under Reversal Watch; the Flow Radar page shows both cards.

A screen, not a recommendation. Hits are often volatile names with rough paths, and only Reversal Watch has a public record so far. Options activity on a card never tells you who traded or which side.

A full-market screen of liquid, optionable stocks with an extended recent move, leaving out leveraged and inverse ETFs. It looks for several exhaustion patterns in price bars and in options volume versus open interest and chain positioning, and suppresses patterns that tend to mark continuation. A recurrence count tracks how often the pattern has fired lately. Since October 14, 2026 it also flags blow-offs: a large one-day jump in a common stock or ADR that had already run up over the previous month, on heavy dollar volume. That rule was written down before it was tested on years it had not seen, and it held there. A hit is published only when the stock's options are deep and actively traded that session. The setup's direction (top or bottom reversal) is the call of record, even when the day's options flow disagrees.

Paths are often rough: picks frequently move against the call before resolving, so tight stops fare poorly. The Strong and Building tiers and the pattern chips have not been shown to separate outcomes. Blow-off picks join the record from October 14, 2026 and have no live record of their own yet. Past picks do not guarantee future results.

Stage one screens every listed stock for liquidity, price, an outsized move (very large jumps are left out to drop binary events) and confirming dollar volume. Stage two checks the options chain of the top survivors for unusual contracts, chain put/call open interest and technical context such as RSI and distance from key moving averages. Hits are ranked by Setup Score. Intraday scans use partial-session versions of the same screen.

In a replay of every trading day since 2018, a big one-day move alone did not predict the next 1 to 12 weeks in either direction, and the day's options lean did no better than chance. Big movers are volatile both ways. Until October 14, 2026 this group was called Momentum and showed an outlook from the options lean; it no longer does.

Points come from three families: the size and consistency of the day's unusual options activity, technical context such as momentum and where price sits against longer-term averages, and the time to expiry of the contracts involved. Names that keep reappearing earn some credit. Hits that score too low are not shown.

It ranks fit to the screen, and our grading has not found that higher scores lead to larger or better moves. It does not reflect trade side or whether positions opened.

The flow read compares call and put premium for the day, then checks the lean against implied-volatility skew and the chain's put/call open interest, with context for large up or down days. It returns bullish, bearish, mixed or volatility magnet. On Reversal Watch the pattern's direction stays the call of record because, measured on the same picks, it has held up better than the flow read.

Premium share is not trade side: heavy call premium can come from sellers as easily as buyers. When pattern and flow disagree, the card keeps the pattern's direction. The flow read has no demonstrated edge on its own.

Relative strength = sector close / SPY close x 100. RS-Ratio = 100 plus a standard score of that ratio against its trailing 65-session window, drawn on a fixed axis. The trend is the least-squares slope of RS-Ratio over 20 sessions, which orders the leaderboard. This follows the widely used relative-rotation approach, built on split-adjusted daily closes. A separate options-based positioning axis is still under validation and is not shown. Rank change counts places moved in the RS-Ratio ranking itself, a different order from the trend that sorts the table.

Descriptive, not a forecast: our grading has found no predictive edge in the ranking. A sector can be improving while still trailing SPY. It is price only, so it says nothing about money flowing into or out of the funds.

Options balance: a model-weighted difference between call and put contract-day activity, signed by each contract's closing price versus its whole-day VWAP. Small balances, and days when too little premium can be assigned, are withheld rather than shown. Dark-pool dollars: off-exchange share times closing price times share volume, an approximation. Large single prints lists each ETF's biggest trade of the session.

The balance is a model result, not executed net premium, and does not show who bought or sold. ETF activity includes hedging and overlay trades, so it is not money moving into a sector. One session only, with no trend.

Open interest comes from the figure published each morning, which reflects the prior session's close; the morning check waits until the new figures have actually landed. Volume and premium rules read the live option chain at each intraday check. Premium traded is contracts traded times the session's average price times 100. A nightly end-of-day record keeps the history complete and delivers any open-interest alert the morning check could not. A contract missing from our stored data is reported as not observed, never read as zero.

A monitor reports that your rule was met. It does not say who traded, whether a trade opened or closed a position, or what the price will do. Open interest is a day old by design, and the underlying stock price may be delayed.

The rule is replayed against stored end-of-day data for recent sessions, using the same evaluation logic as the live checks. The verdict comes from the share of judged sessions on which it would have fired. With too few sessions to judge, no verdict is shown. Suggestions are tighter versions of the same rule replayed on the same sessions, kept only when they cut the firing rate substantially without dropping many sessions from the count.

It is a replay, not a promise. End-of-day data is an approximation of the intraday checks, so a volume rule can fire differently live. A Noisy rule is not wrong, only frequent.

The observed value divided by the threshold you set, shown to one decimal below 10x and as a whole number above. The bar count rises with the multiple. It is not shown for "at most" rules, for a threshold of zero, or for triggers recorded under an earlier version of the rule.

It is not a market reading: a 26× trigger is not more bullish, bearish or important than a 2× one. It depends entirely on where you set the threshold.

Counts the monitor's most recent sessions on which at least one check could judge the rule, and how many of those it fired on. A fire counts whether or not it was emailed, so muted fires are included; that is why the copy says fired rather than alerted. The card and the email apply the same rule.

It describes how often your rule triggers, not anything about the contract. The card and the email can occasionally disagree when many recent sessions could not be checked.

Uses the same arithmetic as Overnight positioning: for each contract, sides opened = volume + change in open interest, and sides closed = volume - change in open interest. Totals are rounded to two significant figures, with ≈ when rounding changed the number. Contracts whose open interest moved by more than their volume changed through exercise or assignment and are left out and noted.

It counts trade sides, not people, and cannot say who bought or sold or which side opened. It describes the prior session only.

Compares open interest between two consecutive complete option-chain observations, after checking that each session's published figures belong to the session they claim to. Your contract comes first, then contracts in the same and nearby expirations, then the whole chain over a comparable set of contracts. Open interest reflects the prior session's close. When a session cannot be verified, no comparison is shown rather than a guess.

Open interest counts outstanding contracts; it does not identify who changed exposure or why. A rise is not proof of new buying, and a fall is not proof of selling. Comparisons appear only for sessions that pass a publication check; while that check is being renewed, no new comparisons appear.

Each mark carries its source and timestamp. A mark drives headline P&L only after checks on the contract's identity, its contract multiplier, your recorded trades and any split or contract adjustment since entry. A price that fails a check is still shown as an observed price, labelled not eligible, and kept out of P&L. Recorded fees are included; taxes and unrecorded costs are not.

Marks are observations, not quotes you can trade at. A reported option trade can be older than the check that read it, and stock prices are delayed. The Weekly Lookback values tracked trades at the latest close.