× what this screen can and cannot tell you
or take the guided tour — one spotlight per card, in the order the rail reads, about ninety seconds.
It does not tell you direction. That was tested three independent ways on this data and is not there: momentum on negative gamma does not replicate 2021-26, the sign of net gex carries nothing (t between −0.51 and +0.15 across horizons), and a random entry time beats the ±100k crossing (p 0.92). There are no buy/sell arrows here because there is nothing honest to put in them.
It tells you how far. Amplitude is real and measurable, and every number below is about size of movement, not its sign. Read it for position size, stop width and whether today's geometry suits your model — never for which way to face.
the panel, in the order it reads
The rail is three zones. Controls at the top decide what is drawn. Readouts below answer, in order, the questions you actually ask: what regime are we in, where is price, where are the levels book by book, what movement was priced and what was realised, and what the dealer is holding. Experiments at the foot are exactly that — rules under test, not readings.
| regime | The one verdict this panel exists to produce: positive gamma means dealer hedging DAMPS moves, negative means the same hedging AMPLIFIES them. Each book decides it on its own flip, so they can split — and the split is the reading, not an error to hide behind one book's answer. In LAB: `gamma_pos`. |
| flips in | The regime card's second line: how far price must travel, and which way, to leave that regime by the CLOSEST book's flip. Below three points it says at the flip instead of a number: the flip is a root recomputed every minute from a book that moves, so a 1.3-point gap is the same number twice, and the verdict above it is about to change. |
| expected move | Two forecasts and the band they describe. The band is anchored at the opening print, not at spot — drawn around spot the marker would sit at its midpoint by construction and could never move. Anchored at the open it says how much of the move the market priced this morning the day has already spent, and which way. In LAB: `range_used`. |
the levels on the chart
Every level is drawn twice over, in two independent renderings you switch on separately — LINES puts them across the chart with their price chips, BANDS paints each wall over the strike it actually owns. A band with no line is the weight of the book with no numbers on the chart at all. Each book has its own switch in its readout box, so one, two or three books can be on the chart at once; where two agree on a price their chips sit side by side on the same line.
| colour | What kind of level it is, never which book it came from: green is the major positive-gamma strike, red the major negative, amber the zero-gamma flip. The book is named on its own chip. |
| width | How much of the near-money book stands on that strike — its share of the gamma in the window. A fat line is a heavy wall; a thin one is a level the book barely carries. |
| fade | Distance from spot. A level forty points away draws at about a quarter strength: legible as "there is a wall out there", never competing with the ones price can reach this hour. |
| bands | Half a strike step either side — the price interval the strike genuinely owns, read off the frame rather than assumed. Opacity carries the same share the width does. The flip never gets one: it is a root, not a strike, and no gamma sits on it. |
| levels · filter | Hides levels holding less than a share of the strongest drawn one. The line under it says how many survived and how many were hidden, so a quiet chart is never a silently truncated one. 8–20% is the useful range; 0 shows everything the book knows. |
What none of it says is direction. A strong level is where hedging concentrates, and the same concentration holds price on a positive-gamma strike and accelerates it away from a negative one. Whether price is drawn to strength has not been measured here — it is an open idea in the LAB feed, not a claim on this screen.
the two forecasts, side by side
| market | The ATM 0DTE straddle, doubled to an expected range. This is what the market is charging for the rest of the session. Measured against realised moves it beats our own forecast at every entry (its t +3.54 against our +0.03). |
| implied · ours | Our gamma-based forecast. Shown for comparison and for honesty: against the straddle it adds nothing (ΔR² = +0.000). When the two disagree, the market is the one to believe. |
realised / implied — the one thing that predicts
Trailing realised range against the implied range over the same window, at 15 / 30 / 60 minutes. 1.00 means the market priced the move exactly. Above 1 it underpriced, below 1 it overpriced. This is the strongest relationship in the whole study (coefficient +0.49, day-clustered t +12.8, deciles running 0.70 to 1.35).
What it does not do is pay. The naked round trip fails — and not on the spread, which is only 8-22% of the move. It fails because implied vol moves against you by almost exactly as much as you gain: at the 60-minute horizon the underlying leg runs −$213 to +$324 across deciles while the IV leg runs +$115 to −$130. Delta-hedging does not fix it either, because hedging removes delta, not vega. Break-even hedging cost came out at 0.141 index points against a real 0.149.
ATM implied vol and Δ IV
The vol the market is charging, and how it changed over the last 30 minutes. This is the row that explains why a correct call loses. If realised/implied is above 1 but Δ IV is negative, you were right about the movement and still paid for it. In the last five minutes both are greyed: implied vol is divided by a vanishing √T so it explodes, and the straddle is up to four minutes stale on its 5-minute grid.
the book
| volume | The cp-signed volume book at 0DTE — the conventional public gamma object, built here independently from the tape. It is the read most screens show. In LAB: `ngv_volcp` (the volume book). |
| open interest | Same construction on open interest. Its net gex sums four expiries at once, so that single number is not comparable with the 0DTE books. |
| measured | Signed from the actual trade tape with the prevailing quote — a read no public feed offers. It disagrees with the convention book about the regime in 39.5% of all minutes — measured over 1,083 archived sessions, and at least once in 96.8% of them — which is why regime at the top of the panel names the books and the side each one takes rather than answering for all three. In LAB: `ngv_meas` (the level) and `meas_conv_div` (the disagreement). |
| gamma pctl | Where today's book sits against up to 20 prior sessions. 0 = thin book and a wide session; 100 = dense and quiet. (Nd) means the pool holds only N sessions and the number is not rankable. In LAB: `gpct`. |
| concentration | Herfindahl share of the near-money gamma. Low = spread across many strikes, high = a handful hold it all. In LAB: `concentration`. |
| flow · 15m | How much gamma actually moved in the last 15 minutes, summed over strikes — whether the book is being rebuilt or merely drifting. max change gex shows the single biggest strike over each lookback -- five independent differences, not one growing envelope; this is the total. In LAB: `mc5` / `mc15` / `mc30`. |
the LAB vocabulary
LAB is the research side: describe a setup in plain English and it compiles your words into conditions, then measures the intraday move that follows — at 5, 15 and 30 minutes — across every archived session. It reads magnitude and behaviour, the same honest scope as this screen; it does not price multi-day spreads and holds nothing overnight. These are the words it understands.
| spot · VWAP · VWAP-z | Price, the session VWAP, and the stretch from it in sigmas (+3 = three sigma above); the 2σ bands are its edges. "Pinned", "stretched" and "extended" all land here. |
| flip · hold · gamma regime | Distance to the zero-gamma flip and to the upper / lower dealer hold, in points — "near the flip", "above the flip", "into the hold". The gamma regime flag splits positive gamma (above the flip: pin / mean-revert) from negative (below: trend). |
| walls · vanna · flow velocity | Distance to the major positive-gamma strike (the call-side wall) and the major negative-gamma strike (put support) — "at the wall", "at support". Vanna pressure on its own, and flow velocity — how much the busiest strike's gamma moved over 5 / 15 / 30 minutes — for "the book is being rebuilt". |
| realized vol · concentration | Intra-minute realized vol — annualised, so it reads straight against ATM IV: a choppy minute vs a smooth one. Concentration — Herfindahl of the near-money gamma: high = the book sits in a few strikes and pins hard, low = spread out. |
| gamma pctl · net gamma · flow | Where the book sits against history (0 = thin and wide, 100 = dense and quiet), net gamma — measured off the tape or conventional — and their disagreement. Flow is the minute-over-minute change of gamma / delta / vanna / charm: "vanna building", "charm accelerating". |
| ATM IV · IV rank · move | The vol the market charges; "IV high / IV rank" pairs the level with its top-percentile. IV change catches "IV spiking / fading"; the expected move is the straddle as a % of spot, comparable across days; range used is how much of it price has spent. |
| VRP · realized vol · skew | "VRP / realized minus implied" is the 30-minute divergence or its ratio (>1 = realized outrunning); realized vol at 15 / 30 / 60 min. Skew is put/call, with its percentile and its steepening rate. |
| volume · volume pctl | Per-minute traded volume and its percentile within the session — volume, not implied vol. The 1-minute candle is the last price change. |
| reading a threshold | "High / low" becomes a percentile or a sigma; "rank > 50%" becomes a top-percentile; "a big move either way" becomes a magnitude. The window it scores is the next 5, 15 or 30 minutes — never a multi-day hold. |
moving around a session
The scrubber is half a pixel per minute in a 235 px column, so a specific time is a keyboard job rather than a mouse one:
| ← → | one minute |
| shift + ← → | ten minutes |
| Home / End | the open, the close |
| space | play / pause |
Stepping by hand pauses playback. Having picked a moment, you did not ask to be moved off it a tick later.
how to use it in practice
Take the implied move as the day's expected span and size against it: if the cone is twice as wide as usual, the same dollar risk buys half the position, or the stop has to be twice as far. A high gamma percentile with a tight cone says mean-reversion geometry — smaller targets, work the edges. A low percentile with a wide cone says the opposite: do not stand against the move and widen the stop.
What none of it licenses is a directional entry. Every number here is a statement about magnitude. The moment you find yourself reading a side out of it, that is the reading the data does not support.