Putting It Together on Gold

≈32 min read · Real gold futures examples with XAUUSD equivalents · Free

In this chapter: You will turn the separate tools of the previous chapters (VWAP, volume profile, auction theory, delta, footprint, the order book and the tape) into one repeatable decision process: Location first, then Confirmation from flow, then Risk. You will learn how to build a short map of gold levels before the session, how to write a confirmation-or-veto checklist, how to write a one-page pre-market plan, how to size a position per tick on GC and MGC and measure results in R, why daily loss limits matter and why trailing drawdowns quietly shrink your room, how to keep a journal that can answer real questions, how a full gold day looks when you follow the process, and how gold behaves around CPI, jobs reports and Fed decisions. Everything here is education about process. Nothing in this chapter is a trade signal.

Location First: The Framework

By now you have a large toolbox: delta and cumulative delta, profiles, VWAP, footprint cells, the DOM, the heatmap, the tape, sweeps and icebergs. The danger of a large toolbox is that you start using every tool all the time.

Gold futures produce dozens of "events" every minute: tape bursts, stacked imbalances, large prints, walls of offers that appear and vanish. If you react to all of them, you are reacting to noise, and every reaction costs commission and slippage. A trader who takes forty small decisions a day because "the flow looked strong" has multiplied their costs by forty and cannot tell which decisions were good.

The framework in this book has three steps, always in the same order.

Step 1 — Location: where?

Before the session opens, you choose a small number of price levels that have an auction logic behind them. These are places where many participants are watching and where the previous auction showed balance or imbalance:

• Yesterday's value area high and low (VAH and VAL) and yesterday's point of control (POC).
• A naked POC: an earlier day's POC that price has not traded back to yet.
• VWAP for the session, and sometimes a weekly VWAP or a VWAP anchored to an event.
• Session highs and lows: Asia, London, the prior day and the overnight range.

At these levels the market is more likely to make a decision: accept the price (trade there with time and volume) or reject it (move away quickly). Between them, most of what happens is rotation and noise.

Step 2 — Confirmation: is the flow agreeing?

Only when price reaches one of your pre-chosen levels do you look closely at the order flow. Are the aggressive buyers or sellers making progress, or are they being absorbed? Is liquidity stacking in front of price or being pulled away? The flow has three possible jobs at this point:

1. Confirm: the evidence is consistent with your hypothesis about the level.
2. Time: it helps you decide when, not just where.
3. Veto: the evidence contradicts your hypothesis, so you do not act.

Of these three, the veto is probably the most valuable. A veto removes a decision, and a removed decision costs nothing.

Step 3 — Risk: where am I wrong, and how big?

Before any position exists, you know where your idea is proven wrong (usually beyond the level), how large the position must be so that being wrong costs a fixed amount, and where the next reference level is.

Why this order?

Why not let the flow find opportunities for you? Because, in published research and in our own testing, order-flow signals on their own did not reliably predict direction for a non-high-frequency trader once realistic costs were included. Many ideas that looked like "something" before costs became nothing after them (Chapter 12).

What remains useful is order flow as a filter in context. At levels chosen in advance, it reduces the number of decisions you take and improves the quality of how you evaluate them. That is a process, not a guarantee of profit.

Key idea: Order flow is not where you look. It is what you check once you are already at the right place.

Common mistake: Adding a new level during the day because "price reacted there". A level drawn after the reaction is hindsight, not location. If it was not on your map before the open, it is not a level for today.

Common mistake: Fifteen lines on the chart. When everything is a level, nothing is a level. A map with too many lines guarantees that price is always "at a level", and the location step stops filtering anything.

"Order flow alone tells you the direction" is a popular claim, not proven. The honest version: order flow tells you more about who is active, where and how hard, and that information is most useful at a place you chose before the session.

Try it: Take a gold five-minute chart from a recent day. Before looking at that day's action, mark only five levels from the prior day (VAH, VAL, POC, the London high and the session VWAP). Shade everything between them grey and label it "no-trade zone". Then scroll through the day. Notice how much of the activity happened inside the grey areas.

Building Your Gold Level Map

A level map is a short list of prices where you will "listen" today. It is not a prediction of where price will turn; it is a list of places where a market decision is more likely, written down in advance so you cannot rewrite it later.

The standard sources

All of these were defined in earlier chapters; here is how each one earns a place on the map.

Yesterday's value area and POC. The value area is the price range where roughly 70% of yesterday's volume traded, with VAH and VAL as its edges, and the POC is the single price with the most volume (Chapter 5). In auction market theory (Chapter 7), opening outside yesterday's value and then returning inside, or being rejected from one of the edges, are among the most important decisions of the day.

Naked POC. A POC from an earlier session that price has not traded at since (Chapter 6). It marks a price where a previous auction found strong agreement and that the market has not revisited.

VWAP. The volume-weighted average price of today's session (Chapter 4): the average price at which today's participants have actually traded. A weekly VWAP or an anchored VWAP from an event (for example, the moment of a big data release) can be added when relevant.

Session highs and lows. Asia, London, the prior day and the overnight session. In gold, the London high and low deserve special attention, because a large part of price discovery happens during London hours, before New York is fully active.

Large round numbers. Every $50 or $100 in gold, for example. The usual explanation is human behaviour and clustered orders. This is a popular claim, not proven; treat round numbers as secondary levels at most.

Ranking: confluence first

Not all levels are equal. Give the highest priority to levels where two or more independent sources line up. If yesterday's VAL sits within a few ticks of today's London low, that zone is more interesting than either alone, because two different groups of participants have a reason to care about it.

Merge levels that are within two or three ticks of each other. Four lines in a $0.30 band are one zone, not four levels.

Write it down, with "what I want to see"

For each level, record four things: the price, the source, the confluence, and what you would want to see from the flow if price arrives. This last field is what turns a line on a chart into a testable hypothesis. It also blocks hindsight bias: after the session you cannot say "I had that level too" if it was not on the list.

Here is an example map in table form. The numbers are hypothetical, chosen to illustrate the format. XAUUSD equivalents use an illustrative gap of about $28 (GC minus ~$28), and the real gap must be measured on the day, as explained in the appendix.

Example — not a signal. Prices are illustrative; the XAUUSD gap is approximate and varies by broker and by day.

Five to seven levels is enough. If you find yourself with twelve, rank them and cut the bottom half.

What a level is, and what it is not

No level "works" in a certain sense. A level is a place where the probability of a decision is higher. A claim like "level X holds every time" needs to be tested on your own data with the methods of Chapter 12.

On strong trend days, price can move through several levels without pausing. That is not the map failing. That is information: the market is accepting new prices, and the other-timeframe participants are in control.

Two pieces of folklore deserve labels. "A naked POC always gets filled, like a magnet" is a popular claim, not proven. The so-called "80% rule" (if price spends two 30-minute periods inside yesterday's value, there is an 80% chance it travels to the other side of value) is also a popular claim, not proven. Both may be useful as ideas to test; neither is a law.

Gold note: "Yesterday" must match your session setting (Chapters 4 and 6). With full Globex sessions (18:00–17:00 ET), yesterday's value comes from the whole session; with the day session only (about 08:20–13:30 ET), the overnight hours form their own profile. Do not mix the two.

Try it: Build tomorrow's map tonight. Limit yourself to six rows. For each row, fill in all four columns, including "what I want to see". Tomorrow, do not edit it after the open; only add notes in a separate column.

Confirmation From Flow: Yes, No, or Veto

When price reaches a level on your map, the question is not "is the flow bullish or bearish?" The question is narrower: is the flow evidence consistent with my hypothesis about this level, or against it?

Let us work through one hypothesis: "Price is rejected at yesterday's VAH and rotates back down into value." (Example only.) What would count as evidence for, and against?

Evidence consistent with the hypothesis (confirm)

• Footprint and delta: absorption. Aggressive buying at the level is heavy, but price does not make progress. Strong positive delta with no movement is the classic sign of effort without result (Chapter 9): someone passive is taking the other side of all that buying.
• Footprint and delta: exhaustion. Positive delta that fades quickly as price pushes into the level; buyers run out of urgency at the extreme.
• DOM and heatmap. Sell-side liquidity above the level stays in place or builds, and it is confirmed by trades: orders actually get filled there rather than only being displayed.
• Tape. A burst of buying at the level, followed immediately by quiet, and price drifting back.
• Sweep. The high of the level is swept (several prices taken in quick succession) and price closes back inside.

Evidence against the hypothesis (veto)

• Aggressive buying with progress; price trades above the level with time and volume. That is acceptance above the level.
• Offers above the level are pulled before price reaches them.
• Cumulative delta moves in the direction of the breakout with no divergence.

A traffic-light way to think about it

Amber is a legitimate state. Many level tests end in amber, and the correct response is to keep waiting or to let the opportunity pass.

Write the checklist before the day

The most important rule of this lesson: the checklist is written before the session, in objective terms, not improvised in the moment. "Flow looks strong" is not a checklist item. "At least one one-minute bar at the level with strongly positive delta that does not close above the level" is a checklist item, because two people looking at the same bar would agree on whether it happened.

Keep it short. Three to five items is plenty. More conditions mean fewer opportunities, which is not necessarily bad, but there is a trap here: mental overfitting. After every loss, it is tempting to add a new condition that would have filtered out that specific trade. Ten losses later, your checklist describes your last ten losses perfectly and the next ten market situations not at all. Chapter 12 explains the statistical version of this trap.

If you use an order-book indicator for the DOM and heatmap side of this checklist (ours is called AK FlowBook, but any reliable heatmap and DOM tool will do), the same principle applies: decide in advance which of its readings count as confirm, wait or veto.

Honest limits

No combination of evidence is certain. Absorption can fail later: a resting iceberg that was absorbing buyers can be cancelled in a millisecond, and then price moves through the level. All the checklist does is improve your odds and define your risk. Two popular claims need their labels: "absorption always leads to a reversal" is a popular claim, not proven; "three stacked imbalances make a certain level" is a popular claim, not proven.

Your checklist is a hypothesis. It needs to be recorded, counted and evaluated like any other hypothesis.

Key idea: The most useful thing order flow may tell you on a given day is: do not take this trade.

Common mistake: Editing the checklist during the session to justify an entry you already want. If the checklist changes while price is moving, it is not a checklist; it is a story.

Try it: Write a three-item confirmation list and a three-item veto list for one level type (for example, "yesterday's VAL"). Then go through ten historical tests of that level type and, without looking at what happened next, record only what the checklist said: confirm, wait or veto.

The Pre-Market Plan for Gold

A pre-market plan is a short document written before the session, so that important decisions are made calmly instead of under the pressure of moving prices. In a "cold" state you weigh evidence; in a "hot" state, with a position open and price moving, you react. The plan moves decisions from hot to cold.

It fits on one page and has six parts.

1. Context: observe, do not predict

What kind of day was yesterday: trending, balanced, rotational? Where is gold opening relative to yesterday's value: inside or outside? What did the overnight sessions (Asia and London) do? Did anything notable happen in related markets, such as the US dollar index or Treasury yields? Write observations, not forecasts. "London made a new high above yesterday's VAH and then returned inside value" is an observation. "Gold will rally today" is a forecast and does not belong in the plan.

2. News calendar

Which data are released today: CPI, the jobs report (NFP), an FOMC decision, a scheduled speech? Write the exact time in New York time (ET). Then write your rule for before and after the release (see the last section of this chapter).

3. Level map

Your five to seven levels, ranked, each with "what I want to see".

4. Scenarios (if–then), each with an invalidation

Two or three scenarios are enough. For example:

• If price reaches yesterday's VAL and I see evidence of sellers being absorbed, then my hypothesis is a rotation back toward yesterday's POC.
• If price is accepted below VAL (time and volume trading below it), then that hypothesis is invalid; I do not act on it and I wait for the next level.

Every scenario needs its invalidation written next to it. A scenario without an invalidation is a hope.

5. Today's risk rules

Risk per trade (in dollars and in R), the maximum number of trades, the daily loss limit, and a stop condition such as "after two losses in a row, I stop for the day". These are the subject of the next two sections.

6. Personal state

Sleep, focus, stress. This is not a soft extra. If your state is poor, the plan should say "reduced size" or "no trading today". Most professionals treat their own condition as an input, the same way they treat the news calendar.

A filled example

Example — not a signal. All numbers are hypothetical.

XAUUSD values use an illustrative gap of about $28; measure the real gap on the day.

Why it works, and its limit

As a process, the plan works because it moves decisions into a calm state and creates something to compare against in your journal. Did you follow it? Did the day look like any of your scenarios?

Its limit: the plan is a discipline tool. Its quality depends on the quality of the hypotheses in it. A beautifully formatted plan built on levels without logic only looks professional.

Common mistake: A plan built around a fixed bias ("today gold goes up"). Bias makes you see confirmation everywhere and vetoes nowhere.

Common mistake: Writing the plan and then ignoring it. Your journal should measure "plan adherence" so you can see how often this happens.

Key idea: Almost every decision made after the open is worse than the version of it you could have made before the open.

Position Sizing Per Tick: GC vs MGC and R-Multiples

Position sizing is where all the abstract talk about risk becomes arithmetic. On gold futures, the arithmetic is pleasantly simple because tick values are fixed.

Tick values, again

Both GC and MGC move in ticks of $0.10 per troy ounce.

• GC is 100 troy ounces, so one tick is worth $10 per contract.
• MGC is 10 troy ounces, so one tick is worth $1 per contract.

A $1.00 move in gold is 10 ticks: $100 on one GC and $10 on one MGC. On XAUUSD, profit and loss per $1 depends on how many ounces your lot represents, which varies by broker; check your broker's specification.

The sizing formula

Number of contracts = Allowed risk in dollars ÷ (Stop distance in ticks × Tick value + Estimated costs per contract)

"Estimated costs" means commission and fees for the round trip plus an allowance for slippage. Always round down.

Worked example

Hypothetical, for learning the arithmetic. Not a trade idea.

Suppose your rules allow you to risk $150 on one trade, and the logical point where your idea is proven wrong is $3.00 away from where you would enter. $3.00 is 30 ticks.

On GC: 30 ticks × $10 = $300 per contract, before costs. Even one contract is twice your allowed risk. GC does not fit this trade.

On MGC: 30 ticks × $1 = $30 per contract. Add an estimate of a few dollars for commission, fees and one tick of slippage, say about $3. That gives about $33 per contract. $150 ÷ $33 ≈ 4.5. Round down: 4 MGC contracts, risking about $132 including costs.

Costs are illustrative; use your own broker's commission and fees.

Notice the last row: at $500 of allowed risk, one GC (about $300 plus costs) and 15 MGC (about $495) are both possible, and the MGC size uses the allowed risk more precisely. This is exactly why the micro contract exists.

Stop first, size second

The single most important sentence in this section: the market decides where your stop belongs, not your wallet. First find the logical place where your hypothesis is invalid (on the other side of the level, beyond where acceptance would prove you wrong). Then compute the size that makes that distance affordable.

The reverse, shrinking the stop until "one GC fits", produces a stop with no logic. It sits at a price that means nothing to the market, so it gets hit by ordinary noise. MGC gives you the flexibility to keep a logical stop at a small size.

Measuring in R

R is the amount you lose if your stop is hit: your initial risk on that trade. Express every result as a multiple of R:

• A full loss at the stop = −1R.
• A gain of twice the initial risk = +2R.
• A loss with extra slippage might be −1.2R.

R makes trades of different sizes comparable and summable. Two +$200 wins are very different achievements if the risks were $100 and $400: in R, they are +2R and +0.5R. Once results are in R, you can compute expectancy: the average R per trade over many trades. That number, not the win rate, tells you whether a process has any edge at all.

The R-multiple approach was popularised by Van Tharp. The rule "risk 1% of the account per trade" is a common conservative convention, not a scientific law. "Risking 1% guarantees survival" is a popular claim, not proven: survival also depends on your real expectancy and on whether losses come in clusters.

Slippage breaks the neat −1R

A stop-market order becomes a market order when it is triggered. In a fast market, especially around news, it can fill several ticks beyond your stop price. Your real loss can then be more than 1R. A 30-tick stop that fills 9 ticks worse on MGC is a 39-tick loss: about −1.3R. Always include a slippage allowance in the sizing calculation, and record actual slippage in your journal.

Key idea: One tick of gold is $10 on GC and $1 on MGC. That difference decides whether your logical stop fits your risk.

Try it: Risk $200, stop distance 20 ticks, ignore costs. How many MGC? (20 ticks × $1 = $20 per contract; $200 ÷ $20 = 10 contracts.) Now add $3 of costs per contract and recompute. (Answer: $23 per contract; 8.7, rounded down to 8.)

Common mistake: Reporting results only in dollars. Dollar results from different sizes cannot be compared or added meaningfully. Record R.

Daily Loss Limits & Why Trailing Drawdowns Are Dangerous

The daily loss limit

A daily loss limit is an amount, expressed in R (for example 2R or 3R), that ends your trading for the day once you have lost it. You decide it in the pre-market plan, not after the second loss.

Losses tend to come in clusters: some days the market does not behave the way your hypotheses assume, and some days your own state is poor. Decisions right after a loss are, on average, worse; the urge to "win it back" brings larger size, looser criteria and trades outside the plan. This is revenge trading. A daily limit stops it and puts a ceiling on your worst day.

Drawdown and the three types of floor

Drawdown is the fall from the highest point of your account. Prop firms (companies that give traders a "funded" account after an evaluation) usually set a maximum drawdown: a floor below which the account is closed. The rules differ by firm; what follows is a general description of the mechanics, not an endorsement or review of any firm. Read each firm's rules at the source.

There are three common types:

Why trailing floors are dangerous: a worked example

Hypothetical numbers to illustrate the mechanics.

You have $2,000 of drawdown room on an account with an intraday trailing floor. You open a trade on gold. It moves in your favour until the open profit is $600. At that moment, with intraday trailing, your floor has just risen by $600: the room you have is still $2,000, but it is now measured from the higher peak.

Then gold turns. The trade goes back through your entry and you close it at a loss of $400.

• Your net result for the trade: −$400.
• Your remaining drawdown room: $2,000 − $600 (open profit given back) − $400 (loss) = $1,000.

You lost $400 of money but $1,000 of room. On an EOD trailing account, the same trade would have cost only $400 of room, because the open profit never became an end-of-day peak. On a static account, also $400.

So an account can be closed without any single large loss: ordinary swings of open profit, given back, consume the room step by step. In gold, which can move several dollars in seconds around news, the effect is stronger.

Practical consequences

• Set your daily loss limit smaller than your remaining drawdown room, with a margin to spare.
• Compute risk per trade as a fraction of the distance to the floor, not of the account balance. If the floor is $1,000 away, risking $250 per trade means four full losses end the account.
• Track the floor's position in your journal every day, so you always know your real room.

Three myths: "if I don't take a loss, my drawdown isn't used" (on intraday trailing, giving back open profit uses it); "one good day will make it all back" (usually followed by bigger size after losses); "a prop account means zero risk" (evaluation fees are real money).

Key idea: You can close the day down only $400 and still lose $1,000 of room on an intraday trailing account. Know which kind of floor you are trading against.

Common mistake: Sizing from the account balance instead of the distance to the floor. A $50,000 account with the floor $1,500 away is, for risk purposes, a $1,500 account.

The Trading Journal That Actually Teaches You

A journal is not a diary of feelings. It is a database you will ask questions of later: "Do my trades at yesterday's VAL do better than at VWAP?" "Is my slippage worse just after 08:30 ET?" "Do I break my plan more often after a loss?" It can only answer questions it was built for.

The fields

Every trade, and every trade you vetoed, should have these fields:

MAE (maximum adverse excursion) is the furthest the trade went against you; MFE (maximum favourable excursion) is the furthest it went in your favour, both measured in R. Over many trades, they show whether your stops are too tight (many trades with MAE just beyond −1R that later went your way) or whether you leave a lot on the table (large MFE, small results).

Why log the trades you did not take?

Because without them you cannot know whether your filter removes bad trades or good ones. If your veto rule removed twenty situations, and those twenty would mostly have worked, your filter is costing you. If they would mostly have failed, it is helping. This is the logic of a control group in research: you need to see what happens to the cases you did not treat.

Example rows

Educational, hypothetical. Not a record of real trades.

Note the first row: a loss, recorded honestly, with the plan followed. In process terms, that is a good day.

Analysing the journal

After at least several dozen trades, look at expectancy (average R) broken down by level type, scenario and time of day. But be careful. With 40 trades split into 10 categories, each category has about 4 trades. Any difference between them is almost entirely chance. This is the multiple-testing problem of Chapter 12 on a small scale: the more ways you slice a small sample, the more "patterns" you find that are not there.

A related myth: "My journal says setup X wins 80% of the time" based on ten trades. That is a popular claim, not proven, and with ten trades it cannot be anything else.

Fill it in before you know the outcome

At least the entry part of the journal (level, scenario, evidence, stop, size) should be filled in before you know the result. Otherwise memory quietly rewrites your reasons to match the outcome. Winners get "strong evidence" and losers get "I wasn't sure about that one". That is hindsight bias, and it destroys the value of the database.

Key idea: A good journal can answer a question you have not asked yet.

Common mistake: Logging only the "interesting" trades, or changing the whole system after five bad trades. Both make the journal useless as evidence.

Try it: Build the table above in a spreadsheet. Add two calculated columns: cumulative R and a running count of trades per level type. Keep it for twenty sessions before drawing any conclusion.

A Complete Gold Trading Day, Walked Through

This section follows one full day with the framework, phase by phase, in New York time (ET). It is a walkthrough of process, not a record of profit. The most instructive day is usually an ordinary one, including the long stretches where the correct action was to do nothing.

Pre-market (about 07:00–08:15 ET)

Review the overnight sessions. Asia trades from the 18:00 ET reopen the previous evening; London becomes active from about 03:00 ET. Where is gold relative to yesterday's value? Where are the London high and low? Check the economic calendar. Write the one-page plan: context, news rule, level map, two or three scenarios with invalidations, risk rules, personal state.

The main gold session opens (about 08:20 ET) and 08:30 data

The traditional COMEX day session begins around 08:20 ET. Many US data releases arrive at 08:30 ET. On a news day, follow your pre-written rule: for example, no new positions from a few minutes before the release until a set time after it. The New York stock market opens at 09:30 ET, which usually adds another wave of activity.

The first test of a level

Suppose price reaches the London low, which on today's map overlaps with yesterday's VAL. Now, and only now, you read the flow in detail:

• Is aggressive selling hitting the level without progress (absorption)?
• Are bids stacking and being confirmed by trades?
• Or are bids being pulled, and is price being accepted below the level (veto)?

Decision and execution

If the checklist says confirm: the position is opened with the stop already defined (beyond the level or beyond the sweep), the size taken from the formula, and the next reference level (for example yesterday's POC or VWAP) noted as the logical objective. If the checklist says veto: nothing happens, and the veto is logged in the journal with the same care as a trade.

Management

Management follows rules written in advance, for example "the stop is moved only after +1R" or "the stop is never moved". Emotional management, such as widening the stop "to give it room" or closing at the first discomfort, is usually the largest single source of errors.

Midday and settlement (about 13:30 ET)

Activity usually slows through the middle of the New York day. Around the GC settlement window (13:29–13:30 ET), behaviour can differ, because settlement prices are computed from trading in that window. Many day traders finish before or around then.

Review (after the session)

Complete the journal, compare the plan with what happened, save the screenshots and write one short lesson. One sentence is enough.

The day as a timeline

Two pieces of folklore to label here: "The first 15 minutes always tell you the direction of the day" is a popular claim, not proven. And showing only great days, the ones that make the process look brilliant, is cherry-picking. A process is judged over many days, never on one.

Common mistake: Trading in the middle of the day out of boredom. A quiet market between levels is not an opportunity; it is the grey "no-trade zone" on your map.

Key idea: The process is the same on every day. The outcome of a single day tells you very little about whether the process is good.

Try it: Pick a past gold session and replay it bar by bar from 07:00 ET. Write the plan using only information available at 08:15. Then step forward and record every decision point: level reached, checklist result, action. Do not skip the quiet hours.

Gold on News Days: CPI, NFP, FOMC

Gold is sensitive to major US data because it changes expectations for interest rates and the dollar. Three releases matter most.

Around FOMC, the reaction can change direction more than once during the press conference.

What the order book and tape usually look like

What follows is a typical pattern as an observation, not a guaranteed rule.

Before the release. Market makers reduce their risk ahead of an unknown number. The book thins out: there is less resting size at each price. On a liquidity heatmap the area around price gets visibly "darker" (emptier). The spread widens. The tape slows down.

At the release. The tape bursts: a huge number of trades in a very short time. Price can "jump" through several prices that have almost no resting liquidity, which looks like a small gap within the day. Stop orders are triggered and multi-price sweeps occur. CME has protective mechanisms such as Velocity Logic, which can pause matching for a few seconds if price moves too far in too short a time.

After the release. Liquidity gradually returns. Sometimes the first move continues; sometimes it reverses completely. Both are common.

Practical consequences

1. Slippage. A stop-market order triggered at the moment of the release may fill many ticks beyond its price. Your real loss can be larger than 1R. A stop placed tightly just before the release, on the assumption that it will fill exactly at its price, is built on a false assumption.
2. Reading the flow in the first seconds is almost meaningless. Everything is a burst. Absorption, imbalances and delta signatures in that window mostly describe the chaos of the release itself.
3. Trailing drawdowns do the most damage here, because open profit can appear and vanish within seconds.
4. Many traders have a rule such as "no new positions from X minutes before a major release until Y minutes after". Some prop firms have their own news rules. Whatever your rule is, it belongs in the pre-market plan, written before the day.

The number versus the expectation

Gold does not react to the number itself; it reacts to the difference between the number and what the market expected, and to the broader situation at the time. "Higher-than-expected CPI always means gold goes down" is a popular claim, not proven. "The first move after NFP is always fake" is also a popular claim, not proven. Neither should be part of a plan.

Gold note: Release times are in New York time. If you trade XAUUSD on a broker platform, the platform clock is often set to the broker's server time, not New York time. Convert the release times before the session, and remember that the US and Europe change daylight saving time on different dates, which shifts the gap between London and New York by an hour for a few weeks each year.

Common mistake: Placing a tight stop right before the release and assuming it will fill exactly at the stop price. When liquidity has been pulled, price skips through empty levels.

Try it: For the next CPI or jobs-report day, watch (without trading) the heatmap and tape from 08:25 to 08:40 ET. Write down three observations: when the book started thinning, what the spread did, and how long it took for resting liquidity to come back.

Chapter summary

• The framework is always Location → Confirmation → Risk: levels before the session, flow only at those levels, risk before any position.
• Order flow has three jobs at a level: confirm, time, veto. The veto (not trading) is a valid and often valuable outcome.
• A level map has five to seven ranked levels (prior value, naked POCs, VWAP, session highs/lows), each with "what I want to see"; confluence raises priority.
• The confirmation checklist is written before the day, in objective terms, and treated as a hypothesis to be tested.
• The one-page pre-market plan covers context, news, levels, if–then scenarios with invalidation, risk rules and personal state.
• Position size = allowed risk ÷ (stop ticks × tick value + costs). GC is $10 per tick, MGC $1 per tick. Stop first, size second. Results are recorded in R.
• Daily loss limits cap the worst day and block revenge trading. Intraday trailing drawdowns count open profit, so giving back open profit consumes room.
• A journal is a database: structured fields, results in R, actual slippage, MAE/MFE, plan adherence, and vetoed trades as a control group. Small samples prove little.
• On CPI, NFP and FOMC days the book thins, the spread widens, the tape bursts and slippage grows. Gold reacts to the surprise versus expectation, not to the raw number.

Checklist

• My level map has no more than seven levels, each with a source and "what I want to see".
• Levels with confluence are ranked highest; levels within two or three ticks are merged.
• XAUUSD equivalents are written next to GC levels using today's measured gap.
• My confirmation and veto checklist is written before the session in objective terms.
• My pre-market plan contains context (observations only), news times in ET, scenarios with invalidations, risk rules and my personal state.
• I find the logical stop first, then compute the size, and I round down.
• I know my daily loss limit in R and my remaining distance to any drawdown floor.
• I know whether my account's floor is static, end-of-day trailing or intraday trailing.
• I log every trade and every veto, with the entry part filled in before the outcome is known.
• I have a written rule for major news releases.

Quiz

1. In the framework of this chapter, when should you start reading the footprint and DOM closely?
2. Which level deserves higher priority: a round number on its own, or yesterday's VAL that lines up with today's London low? Why?
3. Your allowed risk is $200 and your logical stop is 20 ticks away. Ignoring costs, how many MGC contracts can you trade, and why is GC unsuitable here?
4. On an intraday trailing drawdown account, what raises the floor?
5. Why does a stop-market order often fill worse at a CPI release?

Quiz answers

1. Only when price reaches a level you chose before the session. Between levels, the flow is mostly noise for this framework.
2. Yesterday's VAL that lines up with the London low. Two independent sources (the prior auction's value and the overnight session's extreme) give more participants a reason to care about the zone. A round number alone is a secondary level at best.
3. Ten MGC contracts: 20 ticks × $1 = $20 per contract, and $200 ÷ $20 = 10. One GC would risk 20 × $10 = $200 before costs, so with any costs at all it exceeds the allowed risk, and it leaves no flexibility.
4. Any new peak in account value, including open (unrealised) profit. That is why giving back open profit consumes drawdown room.
5. Liquidity is pulled before the release, so the book is thin and the spread is wide. When the stop triggers and becomes a market order, price skips through nearly empty levels, and the fill can be many ticks away from the stop price.

Education only. Not financial advice. Trading involves substantial risk of loss.