Trading wiki

Futures — what actually trades

A future is a contract to deliver a fixed quantity at a fixed date. Everything is standardised except the price: quantity, quality, delivery month, smallest price step. It is agreed on an exchange; the counterparty afterwards is the clearing house.

What trades is the contract, not the underlying — the index, commodity or exchange rate it refers to. Buying NQ delivers no basket of shares. Most positions are closed before expiry, and equity index futures settle in cash rather than delivery.

Contract size, point, tick

Contract size says how much underlying sits in one contract: 1,000 barrels for CL, 100 ounces for GC, for an index a fixed amount per index point. From it follows the point value: what one price point is worth per contract. The tick is the smallest price change the exchange allows; the tick value is tick times point value.

Contract	Tick	Point value	Tick value
NQ — E-mini Nasdaq-100	0.25	$20	$5
ES — E-mini S&P 500	0.25	$50	$12.50
MNQ — Micro Nasdaq-100	0.25	$2	$0.50
MES — Micro S&P 500	0.25	$5	$1.25
GC — Gold	0.10	$100	$10
CL — Crude Oil WTI	0.01	$1,000	$10
FDAX — DAX	0.5	€25	€12.50

Confusing point and tick is the most common arithmetic error at the start; in NQ and ES the two are a factor of four apart. The Trading Station derives every money figure from the point value. A wrong number there is off by the same factor in every profit and loss line. So a test does the arithmetic on the table above: tick times point value has to give the tick value, and the German version has to name the same figures.

Expiry and the roll

Every contract carries a contract month: NQ 09-26 is September 2026. Equity indices and currencies run the quarterly cycle — March, June, September, December. The station resolves gold to February, April, June, August and December, crude oil monthly.

For the equity index quarters the date that counts is the third Friday of the contract month, not its first day; take the month marker for the expiry and you roll three weeks early. Trading moves into the next contract before that, the station's roll calendar eight days ahead of it. Metals and energy expire before their contract month and follow their own rule.

The chart jumps at the change, and that is not a fault. Two contracts on the same underlying trade at different prices in the same second. The gap is the cost of holding until each expiry: financing, storage for commodities, expected distributions for indices. How large it is depends on the market.

Glue two series together and a step appears at the seam. Back-adjust the older one and the seam disappears, but no older bar shows a price that traded. The station follows the front month and back-adjusts nothing.

Margin and leverage

Margin is a performance bond, not a purchase price. Initial margin has to be there on entry, maintenance margin at all times; below it the broker asks for more or closes out. A broker may require less for positions closed the same day. That changes the deposit, not the risk.

Margin is not a cap on losses either: a contract is worth price times point value, a multiple of the collateral demanded. That is where the leverage sits, and it works both ways — one NQ point is $20 per contract, whatever was deposited. A loss can exceed the deposit.

Position size therefore follows from the distance to the stop and what a mistake may cost, not from the margin still free. The station computes profit and loss from the point value; the margin requirement sits with the broker.

E-mini and micro

Each large contract has a smaller sibling on the same underlying and contract month. In the equity indices the micro is a tenth of the mini at the same tick size: NQ $20 against MNQ $2, ES $50 against MES $5. That does not hold everywhere — Micro Silver is a fifth of SI, and Micro Euro FX has a coarser tick than 6E.

It matters at the start: the smallest possible position gets smaller. Ten points against you cost $20 in MNQ instead of $200. Ten micros are still one mini, though — the step shrinks, the leverage does not. The books are separate too, less rests in the micro, and the DOM ladder shows the book of the subscribed contract.

Why futures suit order flow

Footprint, Delta, Volume Profile and Speed of Tape count how much traded at which price level against which side. A count needs a complete denominator. A future has one order book per contract: every trade in NQ 09-26 goes through the same matching engine of the same exchange.

Market	Book	What a volume figure means
Futures	one central book per contract	turnover of the contract
Equities	many venues, plus off-exchange	a share of it, or a delayed consolidation
FX spot	no central exchange	turnover of one bank or platform
Crypto	one book per exchange, many exchanges	turnover of that one exchange

Three limits remain. The DOM shows what is reported — a snapshot of the book the feed delivers, not what lies behind it. In the roll week volume splits across two contracts, so the front month profile is thinner than the trading overall. And a complete denominator still says nothing about the next price. It says what happened.

Trading futures carries the risk of total loss. Nothing here is investment advice or a recommendation to buy or sell a contract.

The Market Is an Auction

A market does not have one price. It has two, side by side.

Bid is the highest price anyone is currently willing to buy at, ask the lowest price anyone is willing to sell at. The spread is the distance between them. Market depth is the size resting at those two prices and behind them; the Station shows it as the DOM ladder in the right-hand panel column, fifteen levels per side by default.

What sits in the book is not a promise. A resting order can be pulled before anyone takes it, and the DOM shows only what the feed reports.

Two ways to send an order

	Limit order	Market order
What it does	joins the book and waits	takes what is in the book
Price	set by the order	decided by the book
Fill	uncertain	immediate, at whatever prices the book holds
To the spread	forms it	crosses it

Whoever waits provides liquidity; whoever takes consumes it.

Limit equals passive is still too coarse. A buy limit priced at or above the ask fills at once and is then the taking side. What decides it is not the order type but who crosses the spread.

The aggressor

A trade happens when one side crosses the spread. That side is the aggressor.

Filled at the ask, the buyer was the aggressor. Filled at the bid, the seller was. The Station records that side for every print in Time & Sales; the footprint sorts each price level into a bid and an ask column, and delta is ask volume minus bid volume.

Where the assignment fails: live, the standing quote arrives with every print, so the side is determined. For historical ticks NinjaTrader stores no quote, and the side is reconstructed with the Lee-Ready tick rule — uptick counts as ask, downtick as bid, an unchanged price inherits the previous classification. That last case is a carry-forward, not a measurement. Bars filled in from one-minute data carry no order flow at all and are drawn dimmed. On a CFD or FX account through MetaTrader 5 there is no tape; prints carry no side.

There is no "more buyers than sellers"

Every trade has a buyer and a seller, the same size, the same price. Contracts bought and sold are always equal in number. A surplus of buyers is arithmetically impossible, however far the market rises.

What is meant is always aggression: not how many wanted to buy, but how many were willing to take the worse of the two prices. That is what delta counts.

Aggression is not a statement about direction either. Aggressive buying can meet a resting seller who absorbs the size without the price moving: delta climbs, the price stands still. That configuration is called absorption.

Where its reading fails: that the size was absorbed is in the data. Why, is not. The same numbers appear when a large order happened to sit there and then disappears. A large delta stays an observation about the past, not about what follows.

What a price actually is

The last price is the place where someone was last willing to trade. It is not a value, not an opinion, and not a commitment for the next contract. Between two fills there is no price at all, only bid and ask.

One property follows, and it spoils analyses. The last price always sits on one of the two sides and bounces between bid and ask without anything about the situation having changed. This bid-ask bounce looks like reversion; on the last traded price it is what gets measured first. WTS Analytics therefore computes on the midpoint.

Why everything else rests on this

Footprint, delta, CVD, imbalance and absorption are bookkeeping on the same auction. All of them hang on a single field: the aggressor side. The diagonal imbalance compares a bid cell with the diagonally offset ask cell — both numbers from that same assignment. If it is wrong, everything above it is wrong, quietly, because the output still looks plausible.

The volume profile is the exception: POC and value area need only price and size, the side only in the Ask/Bid and Delta types.

That is why the origin of the side is marked. Bars without order flow are drawn dimmed and tagged OHLCV only in the tooltip; WTS Analytics states above every study how many trades stayed unclear. Without that figure, what is drawn on top cannot be judged.

And what the auction does not deliver: any statement about what happens next. It shows who was willing to pay up to a moment ago. Who is willing in the next moment is in no book.

The Order Book and Market Depth

The order book lists the orders resting in the market, none of them filled. The Station shows it as the DOM ladder (level 2) in the right panel column: price levels stacked, bid size and ask size per level, the last price marked, each level backed by the session profile of that price. Fifteen levels per side are visible by default.

One sentence first, because everything else hangs on it: what stands there are intentions. Nothing in the book is a trade. It becomes one when someone takes it, and only then does it appear in Time & Sales.

	Order book (DOM)	Time & Sales
Shows	orders that wait	trades that happened
Tense	the state right now	the past
Revocable	any time, at no cost	no
Says	what someone would do	what someone did

Liquidity

Liquidity is the size that can be traded without price running far. Only part of it sits in the book: the resting limit orders, shown as size per level. The Heatmap window draws those same sizes over time, so you see how long something sat at a price; the Station can also take resting size from Bookmap.

Liquidity is a property of the moment, not of a price. It is there as long as it stays.

What the book does not contain

Market orders are never in it. They arrive and take what is there. So the demand that moves a price is the demand nobody could see beforehand. Stop orders are not in it either. They wait as triggers and turn into market orders only when they fire. An area can look thin and still carry a lot of fills.

Beyond that, the feed delivers the reported depth. What sits behind the last level shown is written nowhere.

Three terms

Term	What happens	What it can mean	Where the reading breaks
Iceberg order	Only a fraction of the size is displayed. Take it, and size is back at once.	Someone wants to trade a lot without showing it.	The exchange does not report it. The Station infers it from the print stream: many small fills on the same side, at one price level, in a short window. Several independent small participants look the same.
Stacking	Several levels on one side are filled with size.	Willingness to trade into depth.	The same picture appears when someone only wants a picture. Cancelling is free.
Pulling	The size disappears before price reaches it.	The intention hung on conditions that no longer hold.	Cancelled or filled: the book does not say. Only the comparison with the tape separates them.

The Station labels conspicuous fills as stop run, initiative or absorption, and marks two patterns separately: the sweep across consecutive price levels, and the iceberg as an attribute on top. That is a reading of the print stream, not an exchange message. It can be wrong.

Why a large bid is not a floor

Twenty ticks below the market, one level carries a strikingly large bid size. The obvious thought: someone is holding price there.

An order in the book is revocable until it fills. Cancelling costs nothing and is faster than a person can look. So the large size says one thing only: someone would stand there if price were there now. It is not.

Two resolutions are possible, distinguishable only afterwards. Either the size gets taken, and then it stands as volume in the tape and as a cell in the footprint. Or it is gone before price arrives, and price falls through a level that looked occupied a moment ago.

A price with a lot of size is a place where something can happen. What happens is decided in the tape.

Speed of Tape: what the tape adds

Speed of Tape (Instant) measures the volume flow of the last few seconds. By default that is three windows of ten seconds side by side, coloured by delta direction, with vol/s and trades/s. Side by side they show whether the flow is building or breaking off, and on which side. The measurement runs against the most recent print, not against the machine clock.

That is the quantity the book does not have: pace. The book is a state and has no before. The tape is a sequence, and it carries the aggressor side of every print: who took, not who would have offered.

Live, that side is certain, because the prevailing quote arrives with every print. In history the quote is missing and the side is reconstructed by the tick rule, a good but not a perfect reproduction. A reading built on the aggressor side of old prints rests partly on an estimate.

The order of things

The DOM carries no decision on its own. It is the counter-check on what has already traded.

The sequence therefore runs from filled to offered. An observation forms in the tape, the footprint, the delta or the profile; then comes the question whether the book fits it. A book that supports an observation does not make it safe. A book that contradicts it makes it questionable. Reverse the order and you lean on statements that can be withdrawn while you read them.

The Station treats the book the same way: in the Forecast, book imbalance is one voice among nine, and it only votes where there is real level 2 data.

None of this is investment advice. Trading futures carries the risk of total loss.

Order Flow: Delta and Cumulative Delta

Every trade has two sides: one rested in the book, the other reached across. Delta counts only the side that reached: volume traded at the ask minus volume traded at the bid. +240 means 240 more contracts were lifted out of the book than sold into it.

What it counts

Delta says something about	Delta says nothing about
which side paid the spread	how much traded in total
whether that came in small or large orders	who stands behind an order
how far aggression ran inside the bar	where price goes next

The reading that breaks most often is in neither column. A patient buyer resting on the bid produces negative delta. Size that waits is invisible here; only impatience is counted.

240 on 300 contracts is not 240 on 12,000. The Stats panel therefore carries: Delta % is delta over the bar's total volume, Delta Effort is delta over the bar's range in ticks. Max/Min Delta shows how far delta ran inside the bar; Multi-Range splits it by order size.

From one value to a curve

A single bar delta is noise. Cumulative delta adds bar deltas as they come: in the Station as CVD in its own pane and as the Delta Dly row in Stats. It restarts at the session change, a value for the day. What the line offers is not its level but its slope against price.

Four readings

Reading	What it can mean	What argues against it
Price and CVD move together	market orders carry the move	price moves by taking liquidity out of the book, so the agreement is largely built in
Divergence: new price high, no new CVD high	the second push carried less aggression	they are common, and size matters more than presence
Absorption: delta runs, price stands	size on the other side takes everything	"price stands" is a tolerance, not a state; a still price can also mean a quiet market
Exhaustion: delta no longer turns into price	the move is running out	in the delta alone it looks identical to absorption

On divergence. The Station's CVD Divergence engine produced 68 signals per day across 18 trading days. A sweep on the same data brings that to about seven only with a minimum divergence size and a longer cooldown. The engine was left unchanged and the signal is off by default. With a different pivot pair the same spot looks different.

On absorption and exhaustion. The Station labels a large print Absorption when it moved price at most two ticks from the previous print while the delta of the running minute pushes the other way. Two ticks on NQ is not two ticks on GC. Two pivots at the same price in the same session are a double tap; if CVD stands still across them, the Station calls it absorption. For exhaustion there is no label. Instead there are the COT rows in the Stats panel: the delta that traded after the bar set its high, and after its low. Heavy buying after the high bought nothing above the high. That is a measurement; reading intent into it is not.

Initiative and response

Initiative means price moved with the aggressor, by default at least three ticks, and the delta of the running minute agrees. Absorption is the counter-case, one print at a time.

Both come from a cascade over every large print: stop run at the session extreme first and only with a volume spike, then initiative, then stop run near that extreme, then absorption. Match nothing and the print stays unlabelled. That order is a choice: without initiative first, every fast run near the high would read as a stop run. A label also records which rule fired first.

Where the delta comes from

Delta is not reported by the exchange. It comes from assigning each trade to bid or ask. Live, NinjaTrader sends the standing quote with every print, so the aggressor is known. For historical ticks no quote is stored, and the tick rule reconstructs it: uptick counts as ask, downtick as bid, unchanged as before. Good, not perfect. Where tick history runs out, one-minute bars fill the gap and carry no delta. On a CFD or FX account through MetaTrader there is no exchange tape: prints come from bid and ask, the side stays unknown, delta and CVD are an approximation.

A delta whose sides were guessed is not a measurement. Which source delivers what is in the chapter on limits. None of the four readings is a trading rule.

The Footprint

What a cell is

A bar reports four numbers: open, high, low, close. It says nothing about how volume was spread between them. The footprint splits the same bar into price levels and writes into each row how many contracts traded there.

The split is by the aggressive side. Volume at the bid means a seller hit a resting buy order. Volume at the ask means a buyer lifted a resting sell order. Those two numbers side by side make a cell. The station calls this the print type Ask/Bid Split; the other three condense the row to a sum, a difference, or both.

What the cell does not hold: who traded, why, and who was on the passive side.

Reading one

Numbers invented, magnitudes realistic. A V500 bar in the NQ closes after 500 contracts and covered four levels.

Price level	Bid	Ask	Row total
20,148.75	18	104	122
20,148.50	30	118	148
20,148.25	40	55	95
20,148.00	96	39	135

Three things stand out. The strongest row is 20,148.50 with 148 contracts, framed yellow as the bar POC. The delta is 316 minus 184, so 132 in favour of the ask side. The distribution is lopsided: bid-heavy at the bottom, ask-heavy at the top.

That is a description, not a forecast. The same picture appears in an advance that carries on, and in an absorption where a larger seller takes the top. The footprint knows only the aggressive side. Which case it was shows in the next bar at the earliest.

Imbalance

The comparison is diagonal: the ask of one row against the bid of the row below, the bid of one row against the ask of the row above.

The reason sits in the order book. With a market one tick wide, the buy order rests at one price and the sell order one tick above. Buying at the ask takes the upper one, selling at the bid the lower one. Those two orders were in the book at the same moment. Bid and ask of the same row never were: one price, two different moments. A horizontal comparison pairs numbers that never stood against each other.

104 at the ask on 20,148.75 against 30 at the bid on 20,148.50 gives 3.47. The default is 3.0, written as 300 percent in the dialog, so the cell gets a mark. One row down, 118 against 40 is 2.95 and nothing is drawn. A minimum difference of 10 contracts applies too, so 3 against 1 raises nothing.

The imbalance frame says that and nothing more: aggression met little volume at the order directly opposite. It sits on the side where it happened, left cell edge bid, right ask.

Where the reading breaks down:

• The argument assumes a market one tick wide. With a wider spread the compared rows no longer sit directly opposite.
• Rows are grouped by default so they stay legible. The diagonal neighbour is then the next traded cell, not the next tick, and a different zoom gives the same bar different marks. Full tick resolution turns grouping off.
• Marks are not drawn in every print type: Ask/Bid Split has them, Delta and Delta + Volume do not.
• Historical prints carry no standing quote; the side is reconstructed through the tick rule, not measured.
• Several imbalances stacked mean the ratio repeated across several rows. No direction follows from that.

Three shapes that recur

P shape. Volume gathers in the upper part of the bar, the lower stays thin. The market ran through the lower half and traded at the top. What that does not say: whether positions were built there or handed over.

b shape. The mirror image, thick below and thin above. The same limitation, reversed.

Unfinished business at the edge. The outermost cell carries volume on both sides, so the auction was cut off there rather than completed. An extreme where it ran out looks different: at the high the bid side goes to zero, at the low the ask side, because nobody took the other side there. The station marks single prints; unfinished business has none and is read off the extreme cells.

All three are observations on a closed bar, not rules. P and b come from the profile of a whole session; on a single bar they rest on far less trading. With coarse grouping the zero at the extreme disappears, because one row holds several ticks. Bars filled from minute bars carry no footprint.

Location and occasion

The two tools have separate jobs.

Question	Volume Profile	Footprint
Span	session, several days, visible range	one bar
Supplies	the location: POC, VAH, VAL, volume nodes	the occasion: what happens there now
Changes	slowly	with every print

Alone, the footprint finds a conspicuous cell anywhere on the chart; alone, the profile draws lines where nothing happens. Together they say more precisely what happens at a price the market accepted or rejected over hours. The composite profile reaches back only as far as the tick history, three days by default.

That too is hindsight. What the cells hold has happened; whether it continues is not in them.

Volume across price

A chart plots volume under the candle. A volume profile turns the axis: not when but where — how many contracts changed hands at each price. One row per price, its length the volume.

In the Trading Station the tool is Volume Profile. Four types set what a row's length measures: Volume, Ask/Bid, Delta and Delta + Volume. Five periods set the span: Latest, Multiple per session, Composite across days, Visible for the viewport, previous day.

A profile needs order flow. Where minute bars filled a gap in the tick history, there are no rows.

Point of control

The longest row is the POC: the price with the most volume in the window. It counts contracts, not trades.

What it is not: an average, a fair price, a target. It describes the past and moves while the session runs.

Value area, VAH and VAL

From the POC a band grows until it encloses a set share of the total volume — 70 percent by default (value_area_pct). The upper edge is VAH, the lower VAL, the band between them the value area.

The band does not grow symmetrically. At each step it compares the two rows above the upper edge with the two below the lower edge and takes the heavier side. The value area is therefore rarely centred on the POC.

The 70 is a convention, not a law of nature; change it and both edges move. POC, VAH and VAL always come from volume, even while the rows display delta.

Term	What it measures	What it does not say
POC	the price with the most volume	that price returns to it
VAH / VAL	edges of that volume share	that price turns there
Value area	the range with that share	where price belongs

Thick and thin

The station calls the long and short rows high volume node and low volume node, off by default. Switched on, they colour single rows by fixed thresholds: an HVN is a local maximum across plus/minus two rows at 55 percent or more of the largest, an LVN a local minimum at 8 percent or less.

What that says: much traded here, little there. The common expectation that price crosses thin areas quickly and stalls in thick ones is already an interpretation. It fails as soon as the reason for trading changes: after a release, yesterday's profile describes another market.

Yesterday's profile

The previous day's profile can be laid over the running day, POC and both edges. The usual justification: many participants have the same edges in front of them. The station's data cannot show that; it measures prints, not intentions.

A profile is only as good as its history. In an earlier version, when the tick history did not reach the session open, the profile came from a fragment: POC, VAH and VAL sat wrong across the chart. The server now reports which span the data covers; Composite reaches back only as far as the backfill window, three days.

Zones that form during the day

Not every stretch worth marking fits a session. Dynamic Zones draw a profile over a range picked by hand — delta left, volume right, with a value area of its own. The zone describes that range and nothing beyond it.

VWAP

The VWAP is the volume weighted average price: price times quantity summed, divided by quantity. Not a mean of closing prices but a centre of gravity.

It is first an execution benchmark: anyone working a large order through the day sets fills against it. It grades; it does not forecast.

The station draws two lines. The first runs from the session open. The second re-anchors every vwap2_minutes inside the session, 30 by default; 0 turns it off. It is an anchored VWAP, not a moving one. A genuinely rolling VWAP, 1 to 240 minutes, sits in the Heatmap window.

Either line can carry a band of plus and minus one standard deviation, off by default. The dispersion is volume weighted over the same stretch. The band says how far price trades from the centre of gravity. Not that it comes back.

Such a number is sensitive to its denominator. In the Heatmap window every print once entered the numerator, while the denominator counted only volume with a known side. Ten percent side-less prints lifted it from 29,250 to 32,613.

What a level is

A level is a place where something can happen. Nothing more. POC, VAH, VAL, the previous day's edges and VWAP mark where a lot or a little traded. What happens there is decided by order flow, not the line.

The profile describes. It does not predict.

Time, Sessions and News

Futures run almost around the clock. That does not make the hours equal. The same footprint at 03:00 is not the same observation as at 15:35. Leave out the time of day and you compare unlike things.

Globex and RTH

The CME electronic trading day opens at 17:00 Chicago time, with a break before the next open. In Berlin that opening falls at midnight. The whole stretch is Globex, also called ETH. The Trading Station holds it as session_open, default 17:00, and cuts the day from one 17:00 to the next.

RTH is the shorter window inside it, the regular hours of the underlying market, fixed per instrument in the catalogue.

Instrument	RTH (Chicago)	Berlin
NQ, ES, YM, RTY, MNQ, MES, MYM, M2K	08:30 to 15:00	15:30 to 22:00
GC, MGC	07:20 to 12:30	14:20 to 19:30
SI, SIL	07:25 to 12:25	14:25 to 19:25
CL, MCL	08:00 to 13:30	15:00 to 20:30
6E, M6E	07:20 to 14:00	14:20 to 21:00
FDAX, FDXM, FDXS, FESX	02:00 to 10:30	09:00 to 17:30
Crypto	none	the column is a formality

Chicago and Berlin are seven hours apart, and six in the two short windows where the daylight-saving switches do not line up. The header carries both clocks, CT and local; the local one is the machine's, not the feed's.

A European day in three parts

Seen from Europe the day breaks into the Asian hours from midnight, the European open in the morning, and the US open at 15:30.

What measurably changes is not the price but turnover, spread and book depth. How much differs per instrument and per day. Three windows show it: Speed of Tape with vol/s and trades/s, the Volume Profile of the running session, and the Stats table with Vol(s) and Time(s) per bar. The Station sees prints and reported depth; who is behind them is not in the data.

Range bars and volume bars close less often at night. A forty-tick range bar can take minutes then and seconds in a fast market. Ten such bars are not ten equal stretches of time.

What hangs on the rollover

The session start carries the daily open, high and low, the previous day's O/H/L/C, and the session profile with POC, VAH and VAL. Delta Dly, the cumulative delta, restarts there, so it is a daily figure and not a running total. The Initial Balance is the high and low of the first thirty RTH minutes; a market without an open, meaning crypto, has none.

The reference time can be moved, to 08:30 for the cash open. Daily levels and profile change with it, and the chart label says so. The POC of the Globex session is not the POC of the cash session. That is not an error; it is a different question.

Scheduled numbers

The Economic Calendar pulls the public weekly calendar from ForexFactory and filters by minimum impact, medium by default. Warnings fire 15, 5 and 1 minute ahead and at the time itself; sounds are off on a fresh installation.

Two limits sit in the filter. A currency mapping exists only for NQ, ES, GC, CL and 6E; every other instrument falls back to a default list that contains ALL. And the calendar knows scheduled items only. An unscheduled statement is in no weekly plan.

What happens at a release is a measurement, not a rule. Whether the book thins and the spread widens shows in the DOM and in the spread, and it differs from release to release.

A stop order becomes a market order when it triggers and fills at the next reachable price. How far that is from the stop is decided by the book at that instant, not by the number typed in. What follows from that is not something the Station says, nor this text.

Where the reading fails: absorption against a thin book is not the same event as against a full one, and the DOM shows only the depth reported now.

Holidays and half days

Here the Station is weak: it keeps no holiday calendar. The RTH column is fixed and stands unchanged on a shortened day, and the Initial Balance computes as always. The calendar carries a holiday impact level, but it ranks below low and the default minimum of medium filters it out. A thin day looks like an ordinary one with little volume, and its levels are drawn across the next chart as though formed in the same traffic.

Why the clock is a feature

The range-bar forecast carries time as inputs of its own: time of day as sine and cosine over a period of 23 hours, weekday, an RTH flag, minutes since the RTH open and minutes to the RTH close.

Two further inputs exist but are not available everywhere: minutes to the next release, and a flag for a release within 15 minutes. There is no calendar source for the past, and the cache holds the current week only, so both are open to the recorded-data model alone.

One rule says the same from the other side. If a due prediction's outcome falls into a different session than the one it was made in, it is discarded and counted, not guessed. The same holds for a candle that never existed, over a weekend say. Otherwise a grade would stand over a move that never happened.

What closes a candle

A candle needs a rule for when it ends. That rule is not in the market. It is chosen, and it decides what falls into one candle.

Four rules are in use: a clock, a count of prints, a count of contracts, a span in ticks. The timeframe field accepts three.

Input	Closes on	Limits
<n>s <n>m <n>h	time	1–3600 s, 1–1440 min, 1–24 h
R<n>	span in ticks	1–2000
V<n>	contracts	10–200 000

Presets: 10s, 30s, 1m, 2m, 5m, 15m, 60m, R8, R12, R24, R40, V250, V500. The Station has no tick candles. WTS Analytics carries all four types on the same event stream.

Time candles

The clock is the only pace-setter independent of the market. Two candles are always the same length. That is the flaw.

A minute at midday may hold a dozen contracts, a minute after a number many times that. Both get the same width and the same footprint cells. On the thin minute, three contracts against one in the diagonally opposite cell already meet the 300 percent imbalance threshold — against that stands the minimum contract difference, ten by default. A fast minute merges several moves instead, and thrust and pullback cancel out. The candle never shows how much material is inside it.

Tick candles

A tick candle closes after a fixed number of prints. Each holds the same event count: long in quiet hours, instant in fast ones.

What it distorts: a print is not a size. A 300-contract trade counts as much as a one-lot. And the print count depends on how the feed aggregates — one market order may arrive as one print or as five.

Volume bars

A candle closes as soon as it holds n contracts. The print that fills it stays whole: the candle can hold more than n, never less.

What it makes equal: every candle carries roughly the same material, so cells stay comparable. A footprint needs that.

What it distorts: volume stops being readable. Heavy turnover is a tall bar on time candles, a dense run of candles on volume bars — activity moves from the height into the time axis. And n does not travel between markets: 250 NQ contracts are not 250 MNQ.

Range bars

A range bar closes when high minus low reaches exactly the set tick count. The breaking print opens the next candle; the Station follows NinjaTrader semantics.

Their role here is measurability. On a minute candle, “up or down” has three answers: up, down, and a flat band that blurs every evaluation. A range bar always leaves its span upwards or downwards, so “the next candle” is a two-way question that can be scored. That is why the range forecast runs per market on a range series, R40 (NQ) and R8 (ES).

What range bars distort: time leaves the axis. A bar may take an hour or a second — the forecast pointer uses five minutes there instead of three candles. Every span is identical by construction, so movement per candle is constant.

Range and volume bars also depend on where counting began. The chart series have no anchor; the series behind the range forecast cuts at 17:00 Chicago time, so two runs produce the same candles.

Same stream, different distribution

Over one session	across candle types
total volume	identical
total delta / CVD at the close	identical
high and low	identical
delta per candle	different
footprint cells, bar POC, imbalances	different
price-CVD divergence	different

All types read the same print stream, and every trade sits in exactly one candle. The candle type changes only the distribution across candles — and that distribution is what every order flow reading reads. Two traders on one market with different candles see different imbalances. Neither is wrong; they measure different things.

The upper half holds only while one source classifies direction. Live, the standing quote arrives with every print; in history it is missing and the aggressor side comes from the tick rule — equally deep in every candle type.

What follows

The candle type belongs to every observation. “Absorption at the high” without R40 or 5m is a memory, not an observation. Settings do not travel either: an imbalance threshold from 5m meets different content on V250. Even how often a tool fires depends on it.

One limit applies to every choice: where the tick history is too shallow, 1-minute bars fill the rest. Such bars carry no order flow — no footprint, no delta — and are marked “OHLCV only”.

Risk, position size, and what a result actually is

R as the unit

One R is the distance from entry to stop. A trade ending at the stop costs 1 R; one that runs twice the stop distance returns 2 R.

In money, two results say nothing about the risk behind them. $300 on a tight stop and $300 on a wide one are the same number and two different things. R removes tick value and quantity.

Where R fails: it says nothing about costs, nor whether the stop sat in a sensible place. The same path yields more R with a tighter stop. And 1 R is the planned loss, not the certain one — a gap through the stop costs more.

Position size

contracts = risk per trade / (stop distance in ticks × tick value)

Tick values from the station's wiki: NQ 0.25 points = $5, ES 0.25 points = $12.50. Assume $300 risk per trade:

Market	Stop	Contracts
NQ	20 ticks at $5	3
ES	8 ticks at $12.50	3

Both put $300 at risk; twenty ticks and eight ticks are the same thing here. Fix the quantity first instead, and it decides where the idea is declared wrong.

Where it fails: a contract cannot be split, and rounding moves the risk. The number is an intention — the fill comes at the next tradable price.

Stop and target belong to the idea

The stop is the price at which the idea is refuted. It comes from the chart, not from the account balance. If it does not fit the account, the size is wrong.

The station enters both as an entry bracket in ticks: slTicks, tpTicks, trailTicks, beTriggerTicks, beOffsetTicks; any 0 switches that part off. Trail and break-even take the stricter value, and a stop is never loosened. Between average entry and stop or target lies the pale risk and reward band.

Where it fails: the bracket is computed by the paper engine in the server. A stop in the server is not an order at the exchange.

MAE and MFE

The worst point during the holding period is the MAE, the best the MFE. The Trading Journal calls them max drawdown and max profit, in points, ticks and money with a timestamp.

Figure	describes	does not describe
max drawdown	this trade's adverse move	the next trade's
max profit	how far this trade ran	a reachable target

The trap is the MFE. It reads like a better target and is not one: it is known only once the trade is over. A target at past MFE levels stops being reached in exactly those trades that turned just short of it. And a touched price is not a fill.

The MAE holds the same trap in reverse: winners never more than eight ticks offside do not prove that a stop at eight ticks would have left them intact.

Where the figures fail: they exist only for trades recorded with journal mode on, and a tick buffer short of the entry reports nothing rather than a fragment.

Expectancy

E = p × average win in R − (1 − p) × average loss in R

Assume p = 40 %, win 1.8 R, loss 1.0 R: 0.12 R per trade. The hit rate alone says nothing. 70 % with a 0.3 R win and a 1 R loss gives −0.09 R.

Where the formula fails: it assumes a loser costs exactly 1 R on average. And as an average over many trades it says nothing about their order or the deepest drawdown along the way.

Costs

A calculation without costs is not a calculation. NQ, 20-tick stop, 1 R = $100 per contract, assuming $2.50 commission per contract per side plus one tick each of spread and slippage: three times 0.05 R, together 0.15 R. If the expectancy above was gross, what remains net is negative.

Measured in R the same costs weigh more on a tight stop: at a 10-tick stop they are 0.30 R, at 40 ticks 0.075 R. It does not follow that the wide stop is better.

The dashboard reports net, pnlNet = pnl − commission. Spread and slippage appear in no line item; they sit inside the fill price. And no report knows the orders that never came together.

Dispersion

Twenty trades prove nothing. At p = 40 %, eight winners are expected and the standard deviation is √(20 × 0.4 × 0.6) ≈ 2.2. At 1.8 R per winner and −1 R per loser, six winners give −3.2 R, eight give +2.4 R, ten give +8 R — all three from the same distribution.

Where this fails too: the calculation assumes equal-sized results and independent trades. A long series narrows the dispersion and still proves nothing: it cannot say whether conditions stayed the same. That is the question WTS Analytics answers with its count of tested hypotheses.

This chapter explains figures and their limits. It is not investment advice. Every number in the examples is set to show a calculation, not measured. Trading futures carries the risk of total loss.

Where all of this sits in the Trading Station

The earlier chapters described terms. This one says which window each lives in. Not a manual.

Term	Window	Form
Executions per price level	Chart	footprint cell
Book over time	Heatmap	one column per time step
Aggressor per print	Time & Sales	side per row

Chart

The footprint shows, per bar, the contracts executed at each price level. Bar POC framed yellow, value area shaded, diagonal imbalances marked on the cell edge: left bid, right ask.

Two places where the reading breaks down: marks appear only in the print types ask/bid split and volume combined; bars filled from 1-minute bars carry no order flow.

Beside it the Volume Profile: POC, VAH, VAL. Composite reaches back only as far as the tick history, three days by default. M-Effort places rectangles at swing points; the documentation states their predictive value is not established on three days of data.

Heatmap

Its columns carry the book, bid/ask, CVD and big trades. From those come a rolling VWAP and the Winkler AI Tracker, an open rebuild of Bookmap's Market Pulse algorithm: the last price's deviation from the price-weighted average, measured against the largest excursion of the same window, and normalised to it. A 100 on a quiet day is not the same movement as a 100 on a loud one.

Right-hand panel column

Speed of Tape measures the volume flow of the last seconds, ten by default. The DOM is a ladder of bid and ask sizes, the session profile behind each price. Time & Sales lists every print with its aggressor side. Big Trades marks large executions and lists them.

Two limits. Live the aggressor is exact, because the standing quote arrives with every print; in history it is reconstructed with the tick rule. And the DOM shows the reported book, not what lies behind.

Chart Trader

The entry bracket is set in ticks: stop, target, trail, break-even with trigger and offset, 0 meaning off. In paper trading the server's engine computes it. Stop and target keep their role colour: a sell stop below a long position is red because it limits loss, not because it sells.

Delta and CVD

Below the chart, the delta pane shows delta per bar with shadows for the intrabar high and low, optionally split into four order-size bands. The bands come from order size observed in the print stream; who stands behind an order is not in the data. The CVD pane carries cumulative delta as a line or candles. Both series break at the session change.

Two detectors sit on top, both off by default. CVD divergence compares pivots in price and CVD and requires a new extreme. Double tap takes the case beside it: two attempts at the same price, two ticks of tolerance by default, same session, no CVD confirmation. If CVD stands practically still, the station calls it absorption.

Both are observations: the second attempt may have found no aggression left, or two pivots simply landed close together.

Dynamic Zones

A tool under Tools, not an automatism. Dragged open it creates a profile zone with delta left, volume right and a value area, anchored to time and price, not pixels. The limit is the hand that draws it: another range yields another POC.

Forecast

A statistical prediction of the next candle, one minute by default. Nine named experts vote: momentum, bar delta, CVD slope, book imbalance and others; book imbalance counts only with real level 2. At the close it is graded and every weight adjusted. The panel shows weights, gradings and the hit rate; below ten graded predictions, none.

The documentation names 52 to 56 percent on one NQ R40 bar, and an after-cost break-even near 80 percent for ES R8. Readable as a direction, not as a trading signal. Not the three-candle forecast, which comes from a language model on demand.

Trading Journal

With journal mode on, every trade gets a file: an image at entry, one at exit, the view and market context. The server computes the roundtrip from the fills, plus highest profit, deepest drawdown, best reachable net result and efficiency. Without the trading window's ticks it stays empty rather than assert a number. Efficiency measures against a peak only hindsight knows.

Dashboard

It opens after login: connections, accounts, trades made. Throughput per bridge is measured, prints per second over the last minute. A bridge switched off turns nothing red; it is a decision, not a fault. A trade is the episode from flat to flat; a partial exit sits inside. Figures are net. Without a losing trade the profit factor shows a dash: undefined, not infinite.

What remains

The station shows and measures. It does not decide.

What order flow cannot do

The previous chapters show what the tools display. This one shows where they stop.

The number is softer than it looks

Delta is the ask volume of a bar minus its bid volume. Both sums rest on one decision per trade: who was the aggressor? Live, every print carries the standing quote, and the side is known.

Recorded data has no quote. The side is reconstructed: for each trade the last reported quote, then price against bid and ask. Three cases stay unclear. A trade exactly at the mid. A quote that is too old. The first trade of the day. WTS Analytics does not count them as balanced; it keeps them out of the delta and states their number above every analysis. A delta that rests on guessed sides for a third of its trades is not a measurement.

The Station's chart does not show that number. There the historical side comes from the tick rule: a good but not perfect reproduction, and nothing on screen says so.

Then the second gap. The tick history covers three days by default; beyond that, one-minute bars fill the series. Those bars carry no order flow, only price and volume, and are drawn dimmed. Read past them and a blank looks like a quiet market.

The threshold makes the pattern

A diagonal imbalance is not a property of the market. It is a comparison: the size in one cell against the size in the diagonally opposite one. The default marks from a ratio of 300 percent, with a minimum difference of ten contracts. Both numbers are settings. Lower them and the same data carries more marks.

Big Trades work the same way. In mode auto the threshold comes from a high percentile of the order sizes observed; on NQ it landed at four contracts. Big here means big relative to what was running.

Every pattern is obvious in a finished chart

Hindsight is hard to notice because it feels like insight. A closed bar shows a result. Live, the same bar is still open.

Observation	What it can mean	Where the reading breaks
Heavy sell delta, price does not fall	The bid is absorbing	The bar is open, the delta can turn. In history the side is reconstructed.
Diagonal imbalance in one cell	One side was executed more aggressively	The cell is final only at bar close, and the mark depends on a threshold you set.
Big trade at the high of the day	A large participant is stepping out	Order size says nothing about who is behind it. A block can be one leg of a larger trade.
CVD rises, price does not	Divergence	Delta and CVD break at the session boundary. A divergence is a state, not a moment.

None of the right-hand column goes away with better settings.

Search long enough and you find something

Test 500 hypotheses against a five percent threshold and twenty-five remarkable hits are to be expected when none is real. That is arithmetic.

An example from this project. The Effort Model engine required its absorption candle to show volume above twice the average and a range below 0.6 of the average. The two conditions fight each other: heavy trade means movement. Alone they matched 2.9 and 29 percent of candles, together 0.128 percent; across 20 sessions of 27,400 minute candles one signal was left. The thresholds were then loosened until it fired again. The choice was made on the number of signals, not on an outcome.

No better search fixes that, only bookkeeping: state how many hypotheses were tested, measure against the base rate instead of zero, leave the last stretch of data untouched by the search, write the decision rule down before the first run. That does not make a rule correct. It makes it checkable.

An observation is not a trading rule

A rule is checkable once another reader applies it to the same data and gets the same result. Most observations are missing four things: when it applies, how its start and its end are recognised, and what counts when both fall on the same print.

Leave one out and it is not a weak signal. It is none, because nothing can be recomputed.

Costs are not in the chart

Analyses like this run without spread, fees and slippage. An excess of 0.3 ticks is a loss when the spread is one tick. That is why the figures are given in ticks.

A cost calculation from the forecast work shows how hard that limit is. On ES R8 the break-even of a one-bar trade sits near 80 percent: 2.4 ticks of cost against a range of eight ticks. A directional estimate can be well calibrated and still be worthless as a signal. For NQ R40 over one bar it cuts the other way: anything above roughly 60 percent at full coverage is a leak until proven otherwise, meaning information from the future. The stated 52 to 56 percent is an expectation, not a measurement.

No guarantee

This chapter supplies no rule and no setup. Two things belong to every idea and no tool supplies them: an invalidation, the point at which the assumption is refuted, and a risk budget, the amount its failure may cost. Without both, an observation is not a basis for anything, only a guess with leverage.

Trading futures can lead to the total loss of the capital employed.

Glossary

The terms from the preceding chapters in alphabetical order, each explained the way the Winkler Trading Station uses it.

Term	Explanation
Absorption	One of the labels the station puts on a big trade: price moved by no more than two ticks and the delta of the running minute leans against the aggressor's side. Whether someone is defending that price or trading simply paused is not something the label decides.
Aggressor	The side that took an order resting in the book. Live it is exact, because every print arrives with the standing quote. In history the quote is missing, and the side is reconstructed with the tick rule.
Ask	The best price at which someone is willing to sell. Buying at the ask makes the buyer the aggressor.
Bid	The best price at which someone is willing to buy. Selling at the bid makes the seller the aggressor.
Big trade	An execution above the configured threshold, drawn as a marker in the chart and listed in the panel column. The threshold is either a fixed contract count or a percentile of the observed order sizes (99, 99.5 or 99.9), with a floor of 30 contracts by default in both cases. Large here means large for this instrument, not important.
Contract	The tradeable unit of a future, with a fixed tick and a fixed point value: NQ 0.25 points per tick, 5 US dollars per tick and 20 per point; ES 0.25 points per tick, 12.50 US dollars per tick and 50 per point. Every contract belongs to a contract month and expires.
CVD	Cumulative volume delta, the delta summed across bars, drawn as a line in its own pane. It restarts at the session change and is therefore a figure for the day, not a running total.
Delta	The difference between ask and bid volume in a bar: how much more was bought than sold by market order. It measures aggression, not direction — every buy has a sell on the other side. Bars without prints carry no delta.
DOM	Depth of market, the ladder of bid and ask sizes per price, fifteen levels per side by default. It is a snapshot of the book as the feed reports it and says nothing about what sits behind the reported depth.
Effort	Volume per traded tick: how much trading a price move cost. The figure is a ratio and only becomes readable next to the surrounding bars.
ETH	Extended trading hours, the time outside the core session (see RTH). In the station's crypto catalogue the same abbreviation is the Ether pair.
Exhaustion	The word for a push whose aggressors run out. The station no longer draws a box for it: exhaustion is only settled afterwards, which made the box a reading rather than an observation.
Footprint	The view of how many contracts were executed at each price level within a bar — as ask/bid split, volume, delta or delta + volume. Bars without print data carry no footprint.
Globex	The CME's electronic trading system. In the station the trading day starts by default with the Globex session at 17:00 Chicago time. That reference time can be changed, and changing it shifts every daily and prior-day level.
High volume node (HVN)	A profile row that is a local maximum across plus/minus two rows and reaches at least 55 percent of the largest row. That is a statement about a threshold, not about a direction.
Iceberg	A run of small executions on the same side at one price step, plus/minus one tick, inside a short window of 500 milliseconds by default: a lot of volume in small pieces. The station never sees the hidden order, only its trace.
Imbalance	An imbalance between two diagonally neighbouring footprint cells: the ask size of a row against the bid size of the row below it, the bid size of a row against the ask size of the row above it. By default it counts from a ratio of 3.0 and a minimum difference of ten contracts. Other thresholds produce other marks; the number is a setting, not a property of the market.
Initiative	The label for a big trade where price moved with the aggressor, three ticks by default, and the delta of the running minute points the same way. It describes what just happened and says nothing about the next bar.
Limit order	An order with a price limit. It will not fill worse than the stated price, and it may not fill at all. It places liquidity into the book instead of taking it.
Liquidity	The sizes resting in the book that a market order can take. Visible liquidity is not promised liquidity: it can disappear before anyone takes it.
Low volume node (LVN)	A profile row that is a local minimum and reaches at most eight percent of the largest row. In the Winkler preset it carries the same colour as a rising bar although it says nothing about direction.
MAE	Maximum adverse excursion, the worst point of a position against the entry. The station's journal keeps the same figure under max drawdown and computes it from the ticks inside the trade window.
Margin	The collateral a broker requires per contract: initial margin on entry, maintenance margin at all times. It is a deposit and not a purchase price — and it does not cap the loss that is possible.
Market order	An order without a price limit. It fills immediately, at whatever price stands in the book. It takes liquidity and makes the sender the aggressor.
MFE	Maximum favourable excursion, the best point of a position in favour of the entry. The journal calls it max profit and states next to it, as giveback, how much of that was returned before the exit.
OHLC bar	A bar that carries only open, high, low, close and volume, because no ticks exist for its window. It is drawn dimmed in the chart and has no footprint, no delta and no imbalances.
Paper trading	The path on which the station calculates against an account inside its own server, with no broker. Market orders fill against the standing quote and the entry bracket is built in the server. Queue position in the book, partial fills and a slide across several book levels do not occur in it.
POC	Point of control, the price row with the highest volume in a profile. It says where most trading happened, and nothing about what happens there next.
Point	The price unit of a contract, made up of several ticks: four ticks per point on NQ and ES. The cash value of a point is fixed per instrument.
R	The multiple of the risk taken: 2R is twice the amount that would have been lost at the stop. An R figure only exists once the risk per position has been fixed in advance; the station's numbers are in ticks, points and money.
Range bar	A bar that closes on range rather than on time: high minus low reaches exactly the configured number of ticks, and the breaking print opens the next bar (R40 = 40 ticks). Such a bar has no fixed duration — statements about time do not hold on it.
Response	The counterpart to initiative in market-profile language: trade that acts against the current move. The station carries no label of that name; what its tape flow can measure at that moment is absorption.
Roll	The switch from the expiring contract month to the next. The station measures it against the last trading day and not against a calendar mark: eight days before the third Friday for the quarterly contracts such as NQ and ES, 25 days for gold and silver, six for crude. Across the roll, two price series belong to two different contracts.
RTH	Regular trading hours, the core session of an instrument, 08:30 to 15:00 Chicago time for NQ and ES. It is narrower than the electronic trading day and is stored per instrument.
Session	The electronic trading day from the configured reference time onwards. The delta series and CVD break at its change, and daily and prior-day levels are computed from the print stream starting at that hour.
Single print	A price level that traded only once inside a bar. It marks a place price passed through quickly, not a place where something has to happen.
Slippage	The difference between the price expected and the price received. The station's paper engine does not model it, and the analyses in WTS Analytics do not deduct it. It belongs to the costs that decide whether a small measured edge is one at all.
Spread	The distance between bid and ask. It is paid on every entry and exit: a measured excess of 0.3 ticks is a loss when the spread is one tick.
Stop run	The label for a big trade at or near the session extreme, within five ticks by default, whose size is at least twice the average of the last twenty executions. Whether stops were resting there is not in the data — the label describes place and size, not intent.
Sweep	A series of executions on the same side across consecutive price steps within a few milliseconds, eight by default. The pattern is measured, not the party behind it.
Tick	The smallest price change of a contract, 0.25 points on NQ and ES. In everyday speech a single print is also called a tick; the station's tick history counts those prints.
Tick rule	The Lee-Ready method that reconstructs the aggressor side when the quote is missing: an uptick counts as ask, a downtick as bid, unchanged as before. It applies to history only and is a good but not a perfect reconstruction.
VAH	Value area high, the upper edge of the value area.
VAL	Value area low, the lower edge of the value area.
Value area	The price range in which a configured share of the volume traded, 70 percent by default. It grows outward from the POC, taking the larger of the two neighbouring rows each step until the share is reached. That share is a setting: another percentage moves VAH and VAL without anything having changed in the market.
Volume profile	The distribution of traded volume across the price levels of a chosen window, with POC, VAH and VAL. It needs order flow; bars without prints contribute no rows.
VWAP	Volume weighted average price, measured from an anchor, optionally with a band of plus/minus one standard deviation. The value hangs on the anchor: a different start time gives a different line.

Four words that mean two things

• ETH is the time outside the core session in the trading calendar and the Ether pair in the crypto catalogue. Which one is meant follows from the window it appears in.
• Tick is the smallest price change and also the name for a single print. A tick history of three days counts prints, not price steps.
• Absorption still exists in the station as a label on a big trade, not drawn by default. The absorption box of the structure detection was removed for the same reason as the exhaustion box: it was a reading, not an observation.
• MAE and MFE are called max drawdown and max profit in the journal. They are computed from the ticks during the trade, not from bar closes.

Being able to name a thing is not a trading decision. Every label in this list describes what stood in the data — not what happens next. Trading futures carries the risk of total loss.