iGroww methodology

The primary exchange's session windows and today's holiday row, evaluated in that exchange's own timezone. Pre-open, Open and Closed each carry the reason and the time the state changes.

The last 30 daily closes of each instrument classified as an index. The level is the latest close, the change is that close against the one before it, and the sparkline is the raw close series — the drawing happens in your browser, the numbers are ours.

Advancers, decliners and unchanged are counts over the scanned cross-section: every instrument that is not an index, has at least two daily bars, and has a non-zero prior close. Unchanged means a close exactly equal to the prior close, not a rounding of it.

Reads the precomputed daily cross-section at its latest as-of, not a live scan of the bar table. The ratio is advancers ÷ decliners. Sector rows break ties on sector name, so two identical requests cannot disagree about the order.

Percent change over the chosen window: the latest close against the close 1, 5, 20, 60 or 252 sessions earlier. The thin underlay beneath each bar is the same instrument's move over the window immediately before it. Most active ranks the latest bar's traded volume.

A sector's change is the mean of its members' percent changes, equally weighted, alongside its advance/decline split and its leading name.

The 200 highest-volume instruments, sized by traded volume and tinted by percent change, grouped by sector.

The % change over the last 20 sessions that have a bar — four trading weeks, not a calendar month, so a holiday-shortened stretch is measured as it actually traded — drawn beside the two disjoint 4-week blocks before it, all from our own end-of-day closes.

Wilder's RSI over 14 sessions, from our own daily closes: 100 − 100 ÷ (1 + average gain ÷ average loss), both averages Wilder-smoothed after the first 14 sessions. The distribution buckets are below 30 oversold, 30–45 weak, 45–55 neutral, 55–70 strong, 70 and above overbought; the extremes board uses 30 and 70, and states them on the page so the reading is auditable.

The z-score of the latest daily return against the same instrument's own prior trailing returns: (today's return − their mean) ÷ their standard deviation. A reading of ±2 or beyond is flagged as an outlier.

One-week, one-month and three-month returns side by side, read from the precompute at 5, 20 and 60 sessions. “Aligned” means all three horizons are positive and the shorter ones are strong relative to the longer — momentum building rather than fading.

Relative volume is the session's volume over the mean of the prior 20 or 63 sessions — the baseline deliberately excludes the session being judged, because a baseline containing today drags toward the value under test. At or above 1.5× (adjustable) the move is called confirmed; below it the name lands in quiet movers, which is the list the board exists for.

Two measurements per session: the gap (open against prior close) and the close position, (close − low) ÷ (high − low). A gap counts from 0.5%; a close in the top or bottom third of the range names the pattern. Every pattern definition is printed on the board itself, so the label is auditable rather than a black-box verdict.

The precompute's 252-session extreme flags. An instrument qualifies only once it HAS a full 252-session history, and the board publishes that qualified count beside the scanned one.

Two parallel rails fitted through confirmed pivot highs and lows in LOG price, sharing one slope, admitted only if seven gates hold. The fit length is discovered from the data rather than chosen from a menu of windows, so a three-week channel is reported as three weeks long. Confidence is six weighted terms and every one of them is served, so a score can be read rather than trusted; breakout risk is derived on the read from the health grade, the unclamped position and the current pierce depth, never stored as a fourth score.

Six detectors over ONE zigzag pivot set per instrument, on 400 sessions of split-adjusted daily bars: double top and bottom, head-and-shoulders and its inverse, island reversals and Darvas boxes. Every threshold is a declared, versioned parameter and the resolved set is fingerprinted into the run row, so an edited gate cannot masquerade as the same algorithm. Each shape stays on the board for a multiple of ITS OWN duration — the same clock its projection is stated on — rather than a fixed number of sessions.

Excess return: the instrument's return over the window minus a benchmark index's return over exactly the same sessions. Leaders and laggards split on the SIGN of excess, not on position in one ranking — taking the top and bottom slices of a short list would put the same name in both.

Pearson correlation of daily RETURNS over 60 sessions. Returns, not prices: two rising price series correlate near 1.0 whatever they do day to day, which is the classic way this screen misleads.

Shares × our own latest close — computed here, not repeated from a provider, so it moves with the tape we hold and carries our as-of. FULL and FLOAT capitalisation are kept apart: full answers “which is the biggest company”, float is what an index weight is actually built from, and universe weights are computed on float.

Float is shares outstanding × the free-float factor, the factor derived from the public float the company states in its own annual filing. Short float is the short interest last reported for the name ÷ those float shares. It is a derived reading over two separate observations, and each keeps its own date on the card: the short position's settlement date and the date the float was measured, which can be more than a year older. Where no float was filed the card says the float is assumed (the full share count) rather than passing it as measured. A short position older than its source's declared reporting cadence allows is shown muted and marked delayed. The figure is shown on the Market cap and Monthly movers hover cards. Method short_float/1.

A month's return is its last session's close over the prior month's last session's close — the same ROCP(1) as any chart — on total-return adjusted prices, averaged across the scanned instruments with each instrument weighted by its own year count. The number of observations and the number of calendar years covered are returned beside every average, and the unfinished current year is excluded.

A forward window over our own event records. A CONFIRMED date is on record as published by the exchange or company; an ESTIMATED date is labelled as such wherever it appears.

One fiscal quarter's prints in up to the last five years, each drawn as its close over the close of its reaction session minus one, so every path is 0% on day 0. The reaction session is the first close that reflects the print: the print's own session, or the next one when it was released after the close. The quarter is named one of two ways — by an upcoming earnings date (the quarter that print will report: the one after the quarter the company last reported) or by choosing any quarter of a ticker. The solid line is the events' mean; the dashed line is the newest event. Match % is the Pearson correlation between the newest event and the mean of the OTHER events over the overlapping sessions, mapped to 0–100 as (r + 1) / 2: 100 is the same shape, 50 is unrelated, 0 is the mirror image. The mean absolute gap in percentage points is shown beside it. Beneath the chart, the run-up (day −window to −1), the reaction (day −1 to 0) and the drift (day 0 to +window) are each averaged across the events with the count that rose. Method pattern_match/2; version 1 placed day 0 on the print's dated session regardless of when it was released.

Names printing in the next 30 days, measured on the same fiscal quarter of every earlier year we hold. Sessions count from the REACTION session: an after-close print reacts the next session, anything else the same one (reaction_session/1). Pre-earnings is the close ten sessions before the reaction session over the close the session before it; earnings day is the close before over the close after; post-earnings is the close after over the close ten sessions later. Hit rate is the share of years positive; the median and mean gains are shown beside it. Method earnings_consistency/1.

The same upcoming set, with the most recent completed print and the same-quarter print a year earlier: close before over close after on the reaction-session rule above, the EPS surprise where an estimate exists, the free float as shares outstanding times the free-float factor (or float capitalisation over close where that is what the snapshot holds), and the last close. Method earnings_gainers/1.

Each plan is a stored entry/exit pair around a print. Its action is read against the exchange session, never the wall clock: upcoming, buy today, live, sell today, done. Unrealised is entry close to latest close; realised is entry close to exit close. A plan is low-conviction when fewer than 60% of its held years were positive or fewer than half were at or above the average. The summary tiles judge done plans on realised and the rest on unrealised; payoff is the average win over the average loss. Method plan_action/1.

Confirmed reconstitutions come from our event records. Predicted effective dates are DERIVED from announcement cadence, carry their probability, and are served as a distinct flagged layer drawn hollow.