The LAB vocabulary

Every column a rule can name: 105 of them, in 9 groups, with the unit, the range, the book it reads and one condition that uses it. The same dictionary the compiler is given, so what you read here is what it understands.

How a rule is written

A rule is one or two legs. A leg has a direction, up or down, the way the trade profits, and a list of conditions on the columns in the groups on the left, all of which must hold at the same minute. The engine then measures the move over the next 5, 15 or 30 minutes of the same session, after a cost, against the core condition alone, out of sample. Say it in words in the LAB chat and it is compiled into this shape; send the shape itself through the API.

{"name": "FOMC last-hour fade",
 "legs": [{"dir": "down", "when": [
   {"col": "ev_fomc", "op": ">=", "value": 1, "core": true},
   {"col": "min", "op": "between", "value": [330, 330]}]}],
 "horizon": 30, "cost": 1.0, "good_move": 3.0}

Operators

operator	meaning
`>= <= > <`	compare with a number
`abs>= abs<=`	compare the magnitude, sign ignored: distances, flows, changes
`between`	inside `[lo, hi]`, inclusive; `[300, 300]` is one exact minute
`top_pctl abs_top_pctl`	in the top share of the column's own distribution over the whole store, 0.5 … 0.99; `0.9` is the top decile, signed or by magnitude

Three things to know first

Distances are signed: spot minus the level, positive above it, so "at the wall" is abs<= a few points and "above the flip" is flip_dist > 0.

Unqualified levels read the measured book; the _volcp and _oi twins read the other two, and a rule should not mix books.

The last minutes have no outcome: a condition on min must leave horizon minutes of session, so keep it at or below 384, 374 or 359 for the 5, 15 and 30-minute horizons.

The same reference as JSON, for scripts and assistants: /api/lab/vocabulary. Every column is computed per minute over every archived session by the Lab's feature-store builder; a new column reaches the compiler, this page and the API in the same deploy.

Time in the session

Where in the day the rule is read. Minute 0 is 09:30, 389 is 15:59; a rule that fires inside the last `horizon` minutes measures nothing, so keep min ≤ 389 − horizon.

column	what it is	unit	range	book	a condition that uses it
`min`	minutes since the 09:30 open (0-389)	minutes since 09:30	0 … 389		`min between [300, 330]`
`mins_left`	minutes remaining to the close	minutes to 15:59	0 … 389		`mins_left <= 60`

Price, VWAP and the move just made

The index itself, the session VWAP and its band, and the last few minutes of tape. Distances are signed index points, positive above.

column	what it is	unit	range	a condition that uses it
`spot`	SPX spot price	index points	the SPX level	`spot >= 5000`
`vwap`	session VWAP	index points	the session VWAP	`vwap > 0`
`vwap_dist`	spot - vwap, points	points, signed	negative below VWAP	`vwap_dist >= 10`
`vwap_sigma`	one sigma of the VWAP band, points	points	one sigma of the band, > 0	`vwap_sigma <= 4`
`vwap_z`	spot distance from VWAP in sigmas (+3 = stretched 3 sigma above)	sigmas, signed	typically −4 … +4	`vwap_z >= 3`
`b2u`	upper 2-sigma VWAP band	index points	upper 2σ band	`spot >= b2u is not allowed; use vwap_z >= 2`
`b2l`	lower 2-sigma VWAP band	index points	lower 2σ band	`vwap_z <= -2`
`candle_delta`	1-minute spot change, points	points, signed	one minute's change	`candle_delta abs>= 5`
`ret15`	spot now minus spot 15 minutes ago, index points (the move just made; positive = drifted up)	points, signed	spot now − 15 min ago	`ret15 <= -3`
`ret30`	spot now minus spot 30 minutes ago, index points	points, signed	spot now − 30 min ago	`ret30 abs>= 10`
`rv_1m`	intra-minute realized vol, annualised -- directly comparable to atmiv (implied); a choppy vs smooth minute	annualised %	intra-minute realised vol	`rv_1m >= 25`

Dealer levels and distances

The zero-gamma flip, the hold band, the call wall and put support, as signed distances from spot. Unqualified names read the MEASURED book; the *_volcp and *_oi twins read the volume and open-interest books.

column	what it is	unit	range	book	a condition that uses it
`flip`	gamma flip level, price	index points	the zero-gamma strike	measured	`flip_dist abs<= 5 (prefer the distance)`
`flip_dist`	spot - gamma flip, points (+ = above the flip / positive-gamma side, - = below)	points, signed	spot − flip; + above	measured	`flip_dist abs<= 5`
`gamma_pos`	positive-gamma regime flag: 1 above the flip (pin / mean-revert), 0 below (trend)	flag	1 above the flip, 0 below	measured	`gamma_pos >= 1`
`hold_lo`	lower dealer hold level	index points	lower edge of the hold band	measured	`hold_lo_dist abs<= 5`
`hold_hi`	upper dealer hold level	index points	upper edge of the hold band	measured	`hold_hi_dist abs<= 5`
`hold_lo_dist`	spot - lower dealer hold, points (- = below it)	points, signed	spot − hold_lo	measured	`hold_lo_dist <= 3`
`hold_hi_dist`	spot - upper dealer hold, points (+ = above it)	points, signed	spot − hold_hi	measured	`hold_hi_dist >= -3`
`cr0_dist`	spot - the major positive-gamma strike (call-side wall / resistance), MEASURED book, points (+ = above it)	points, signed	spot − call wall; negative below it	measured	`cr0_dist abs<= 5`
`ps0_dist`	spot - the major negative-gamma strike (put-side support), MEASURED book, points (+ = above it)	points, signed	spot − put support; positive above it	measured	`ps0_dist abs<= 5`
`cr0_strength`	how much of the near-money book stands on the call-side wall's own strike, 0..1 (1 = the book's fattest strike; the terminal draws line width from this)	share	0 … 1 of the near book on that strike	measured	`cr0_strength >= 0.3`
`ps0_strength`	same for the put-side support strike, 0..1 -- a thin level is drawn like any other but may behave like none	share	0 … 1	measured	`ps0_strength >= 0.3`
`flip_volcp_dist`	spot - the volume-book gamma flip, points -- an alternative flip construction	points, signed	spot − volume-book flip	volume	`flip_volcp_dist abs<= 5`
`cr0_volcp_dist`	spot - the call-side wall of the VOLUME book, points (+ = above it)	points, signed	spot − volume-book call wall	volume	`cr0_volcp_dist abs<= 5`
`ps0_volcp_dist`	spot - the put-side wall of the VOLUME book, points (+ = above it)	points, signed	spot − volume-book put support	volume	`ps0_volcp_dist abs<= 5`
`cr0_oi_dist`	spot - the call-side wall of the OPEN-INTEREST book at 0DTE, points (+ = above it)	points, signed	spot − OI-book call wall	open interest	`cr0_oi_dist abs<= 5`
`ps0_oi_dist`	spot - the put-side wall of the OPEN-INTEREST book at 0DTE, points (+ = above it)	points, signed	spot − OI-book put support	open interest	`ps0_oi_dist abs<= 5`
`flip_struct_dist`	spot - the open-interest (structural) gamma flip, points -- the third flip alongside flip_dist and flip_volcp_dist	points, signed	spot − OI-book (structural) flip	open interest	`flip_struct_dist abs<= 10`
`cw_agree`	how many of the three books name the SAME call-side wall, 1..3. Agreement is rare -- 22% of minutes -- so 3 is a genuinely unusual state, not a default	count	1 … 3 books naming the same call wall	all three	`cw_agree >= 2`
`pw_agree`	same for the put-side wall; the books agree here in only 2% of minutes	count	1 … 3	all three	`pw_agree >= 2`
`cw_agree_dist`	points from spot to the call wall the books agree on (negative = it is above spot). Null when they name different strikes	points, signed	spot − the agreed call wall	all three	`cw_agree_dist abs<= 5`
`pw_agree_dist`	points from spot to the agreed put wall; positive = it is below spot	points, signed	spot − the agreed put support	all three	`pw_agree_dist abs<= 5`

The dealer book and its flow

Net exposures at spot and how fast they are changing. Level columns are the state; *_flow and mc* columns are what moved in the last minutes.

column	what it is	unit	range	book	a condition that uses it
`dex`	dealer delta exposure	$ of index delta	signed; + = dealers long delta	measured	`dex <= 0`
`dealer_dex`	dealer delta exposure (alias)	$ of index delta	alias of dex	measured	`dealer_dex top_pctl 0.9`
`vex`	vanna exposure	$ per vol point	signed	measured	`vex abs_top_pctl 0.9`
`charm`	charm exposure	$ delta per day	signed	measured	`charm <= 0`
`vanna`	vanna pressure (distinct from vex)	$ delta per vol point	signed	measured	`vanna abs_top_pctl 0.9`
`gpct`	gamma percentile vs history	percentile	0 … 100 vs the last 20 sessions	measured	`gpct <= 20`
`ngv_meas`	net gamma value, measured	$M	net gamma at spot, signed	measured	`ngv_meas <= 0`
`ngv_conv`	net gamma value, conventional	$M	net gamma at spot, signed	open interest	`ngv_conv >= 0`
`ngv_volcp`	net gamma value, volume-cp book -- the third book alongside ngv_meas / ngv_conv	$M	net gamma at spot, signed	volume	`ngv_volcp <= 0`
`meas_conv_div`	ngv_meas - ngv_conv: the tape-signed book minus the call-minus-put OPEN-INTEREST book. Not a measured-vs-volume pair -- that is ngv_volcp	$M	ngv_meas − ngv_conv	measured vs OI	`meas_conv_div abs>= 50`
`dgamma_flow`	minute-over-minute change of net gamma (dGamma flow)	$M per minute	signed change of ngv_meas	measured	`dgamma_flow abs_top_pctl 0.9`
`dex_flow`	1-minute change in dealer delta exposure	$ per minute	1-minute change of dex	measured	`dex_flow abs_top_pctl 0.9`
`vex_flow`	1-minute change in vanna exposure	$ per minute	1-minute change of vex	measured	`vex_flow >= 0`
`charm_flow`	1-minute change in charm	$ per minute	1-minute change of charm	measured	`charm_flow <= 0`
`mc5`	max-change GEX over 5 min -- flow velocity: how much the busiest strike's gamma just moved	$M	largest single-strike gamma change over 5 min	measured	`mc5 top_pctl 0.9`
`mc15`	max-change GEX over 15 min (flow velocity)	$M	over 15 min	measured	`mc15 top_pctl 0.9`
`mc30`	max-change GEX over 30 min (flow velocity)	$M	over 30 min	measured	`mc30 top_pctl 0.9`
`concentration`	Herfindahl of near-money gamma: high = the book sits in a few strikes and pins hard, low = spread out	Herfindahl	0 … 1; high = a few strikes hold it	measured	`concentration >= 0.3`

The unsigned book (GROSS)

|gamma| × open interest per strike, calls and puts summed, no side assigned: how much convexity there is and where it sits. The two-sided measured book alongside it.

column	what it is	unit	range	book	a condition that uses it
`gross_total`	sum of \|gamma\| x open interest over the near-money window, $M per index point. Magnitude only	$M/pt	sum over the near-money window	unsigned	`gross_total top_pctl 0.8`
`gross_at_spot`	the unsigned book on the strike nearest spot, $M/pt	$M/pt	the strike nearest spot	unsigned	`gross_at_spot >= 100`
`gross_up`	the heaviest unsigned strike ABOVE spot, $M/pt	$M/pt	heaviest strike above spot	unsigned	`gross_up >= 150`
`gross_dn`	the heaviest unsigned strike BELOW spot, $M/pt	$M/pt	heaviest strike below spot	unsigned	`gross_dn >= 150`
`gross_up_dist`	points from spot up to that heaviest strike above, always positive	points, always positive	distance to gross_up	unsigned	`gross_up_dist <= 5`
`gross_dn_dist`	points from spot down to the heaviest strike below, always positive	points, always positive	distance to gross_dn	unsigned	`gross_dn_dist <= 5`
`gross_near_share`	share of the unsigned book within +-10 pt of spot, 0..1 -- how much convexity sits on top of price	share	0 … 1 within ±10 pt of spot	unsigned	`gross_near_share >= 0.5`
`gross_top3_share`	share of the unsigned book on its three heaviest strikes, 0..1 -- concentration	share	0 … 1 on the three heaviest strikes	unsigned	`gross_top3_share >= 0.6`
`signed_over_gross`	sum \|measured gamma\| / sum unsigned: how much of the convexity the tape-signed book accounts for. Low = the books cancel, the sign is thin	ratio	0 … 1; low = the sides cancel	measured ÷ unsigned	`signed_over_gross <= 0.15`
`meas_below`	sum \|measured gamma\| on strikes within 30 pt BELOW spot, $M/pt	$M/pt	sum \|measured γ\| within 30 pt below spot	measured	`meas_below >= 50`
`meas_above`	same within 30 pt ABOVE spot, $M/pt	$M/pt	sum \|measured γ\| within 30 pt above spot	measured	`meas_above >= 50`
`meas_below_chg15`	change of meas_below against the frame 15 min earlier (negative = the book below thinned)	$M/pt, signed	change of meas_below over 15 min	measured	`meas_below_chg15 <= -20`
`meas_above_chg15`	change of meas_above over 15 min (positive = the book above thickened)	$M/pt, signed	change of meas_above over 15 min	measured	`meas_above_chg15 >= 20`
`meas_below_chg30`	change of meas_below over 30 min	$M/pt, signed	over 30 min	measured	`meas_below_chg30 <= -20`
`meas_above_chg30`	change of meas_above over 30 min	$M/pt, signed	over 30 min	measured	`meas_above_chg30 >= 20`

The expiry split

Today's 0DTE against everything that survives it, per book. Only sessions archived since the wide book carry it; earlier days are NaN and never match.

column	what it is	unit	range	book	a condition that uses it
`gam_0dte`	measured gamma, TODAY's expiry only ($M)	$M	today's expiry, net at spot	measured	`gam_0dte <= 0`
`gam_next`	measured gamma past today -- the book that survives the close ($M)	$M	the book that survives today	measured	`gam_next >= 0`
`gam_next_share`	share of the measured book that outlives today (0..1)	share	0 … 1	measured	`gam_next_share >= 0.6`
`gam_vol_0dte`	volume-book gamma, 0DTE only ($M)	$M	today's expiry	volume	`gam_vol_0dte <= 0`
`gam_vol_next`	volume-book gamma past today ($M)	$M	surviving book	volume	`gam_vol_next >= 0`
`gam_vol_next_share`	share of the volume book that outlives today (0..1)	share	0 … 1	volume	`gam_vol_next_share >= 0.6`
`gam_oi_0dte`	open-interest gamma, 0DTE only ($M)	$M	today's expiry	open interest	`gam_oi_0dte <= 0`
`gam_oi_next`	open-interest gamma past today ($M)	$M	surviving book	open interest	`gam_oi_next >= 0`
`gam_oi_next_share`	share of the OI book that outlives today (0..1) -- the OI book is roughly half next-expiry, so this one really moves	share	0 … 1	open interest	`gam_oi_next_share >= 0.6`
`book_width`	how many post-0DTE expiries the session was ARCHIVED with -- a provenance stamp, not a market state. ~21 on a batch-built day, 3 on a live one (a live payload's flow book really is narrow). Use it ONLY to keep a rule that touches the gam_* columns inside one definition of them; never as a market condition, and never in a rule that does not reference gam_*	count	3 before 2026-08-17, ~21 after		`book_width >= 10`

Implied and realised volatility

The 0DTE ATM surface, the straddle, skew, and realised volatility over three windows, plus the ratios that make them comparable across days.

column	what it is	unit	range	a condition that uses it
`atmiv`	ATM implied-vol level (pair with top_pctl for an 'IV rank')	IV, %	0DTE at-the-money	`atmiv top_pctl 0.9`
`atmiv_chg`	1-minute change in ATM implied vol	IV points per minute	signed	`atmiv_chg >= 0.5`
`straddle`	ATM straddle price	index points	the 0DTE ATM straddle	`straddle >= 30`
`straddle_pct`	ATM straddle as a % of spot -- the expected move, comparable across days	% of spot	the expected move	`straddle_pct >= 0.6`
`exp_range`	expected range implied by the straddle	index points	the implied range for the session	`exp_range >= 40`
`range_used`	\|spot - vwap\| / expected range -- how much of the day's implied range is spent	ratio	\|vwap_dist\| ÷ exp_range	`range_used >= 1`
`div30`	30-min realized-minus-implied divergence -- the VRP proxy	vol points	realised − implied, 30 min	`div30 >= 5`
`vrp_ratio`	realized/implied ratio (rvi30 / atmiv); >1 = realized outrunning what's priced	ratio	realised ÷ implied; > 1 = tape outrunning	`vrp_ratio >= 1.2`
`rvi15`	15-min realized vol	ratio	realised ÷ implied, 15 min	`rvi15 >= 1.5`
`rvi30`	30-min realized vol	ratio	30 min	`rvi30 <= 0.7`
`rvi60`	60-min realized vol	ratio	60 min	`rvi60 >= 1.2`
`skew`	put/call skew	vol points	25Δ put − call IV	`skew top_pctl 0.9`
`skewpct`	25-delta skew as a RATIO: put IV over call IV minus one, in PERCENT (median ~10, 95th pctl ~25) -- NOT a 0-1 percentile; 'rich skew' is skewpct >= 15 or top_pctl on it	percent	put IV over call IV minus one; ≈ −2 … 25, median ≈ 10. NOT a 0…1 percentile	`skewpct >= 15`
`skew_chg`	1-minute change in put/call skew	vol points per minute	signed	`skew_chg >= 0.2`

Volume

Contracts printed across the whole 0DTE chain per minute, and its rank inside the session.

column	what it is	unit	range	a condition that uses it
`vol`	per-minute traded volume	contracts per minute	whole 0DTE chain	`vol_pctl >= 0.9 (rank it, do not threshold contracts)`
`vol_pctl`	traded-VOLUME percentile of this minute vs the same minute of the prior 20 sessions (0-1) -- 'heavy for this time of day'; volume, NOT implied vol	percentile	0 … 1 against the same minute of the prior 20 sessions	`vol_pctl >= 0.9`
`vol_snap`	volume as the snapshot carries it	contracts per minute	the snapshot's own series	`vol_snap >= 1000`

The market calendar

The scheduled-event context of the session, from the same calendar the terminal draws: day flags for FOMC, CPI, payrolls, PCE, ISM, earnings and opex, and two clocks to the nearest in-session event.

column	what it is	unit	range	a condition that uses it
`ev_high`	1 on a day with a high-impact release: FOMC, CPI, payrolls or PCE; else 0	flag	1 on FOMC, CPI, payrolls or PCE days	`ev_high >= 1`
`ev_fomc`	1 on an FOMC decision day (statement 14:00, press conference 14:30)	flag	1 on a decision day	`ev_fomc >= 1`
`ev_cpi`	1 on a CPI day (08:30, before the open)	flag	1 on a CPI day	`ev_cpi >= 1`
`ev_nfp`	1 on a payrolls day (08:30, before the open)	flag	1 on a payrolls day	`ev_nfp >= 1`
`ev_pce`	1 on a PCE inflation day (08:30, before the open)	flag	1 on a PCE day	`ev_pce >= 1`
`ev_ism`	1 on an ISM day, manufacturing or services (10:00, inside the session)	flag	1 on an ISM day	`ev_ism >= 1`
`ev_earn`	1 when a mega-cap (NVDA, AAPL, MSFT, ...) reports today, before the open or after the close	flag	1 when a mega-cap reports	`ev_earn >= 1`
`ev_opex`	1 on a monthly or quarterly options expiration day	flag	1 on monthly / quarterly expiration	`ev_opex >= 1`
`ev_pre_open`	1 if a high/medium release came out before 09:30, so the session already trades AFTER the number	flag	1 if a release came out before 09:30	`ev_pre_open >= 1`
`ev_mins_to`	minutes until the next scheduled event inside the session (FOMC 14:00, ISM 10:00, FOMC minutes); NaN when none is ahead, so a condition on it selects event days by itself	minutes	to the next in-session event; NaN if none	`ev_mins_to between [30, 90]`
`ev_mins_since`	minutes since the last scheduled event inside the session; NaN before the first one	minutes	since the last in-session event; NaN before it	`ev_mins_since <= 15`