Atlas Sessions (GC) guide

What Atlas Sessions (GC) is

Atlas Sessions (GC) divides the gold futures day into four sessions. At each session open it draws the previous session's high and low and tells you how often, historically, a session like this one went on to take the high, the low, both, or neither.

The four figures add up to 100%. They are measured frequencies from eleven years of COMEX gold history (2015 to 2026), not forecasts. If the table reads Take high only 37%, it means that in the measured history, comparable situations took the previous high and left the low alone 37% of the time.

Yes, one of those boundaries moves during the year. That is deliberate, and the section on the two clocks below explains why.

This is classification, not prediction, and it is not a signal. It answers "what usually happens from here", so your own analysis starts from the base rate rather than from a guess.

Why eleven years

The gold data goes back a good deal further than the figures do. Gold's intraday behaviour changed materially in 2015, and averaging across that change would blend two different markets into one set of figures. The figures are measured only on the market you are trading now.

Distance comes first

How far the previous session's high and low sit from where the new session opens is the main thing that decides whether they get taken. That much is intuitive.

The less intuitive part is that "near" and "far" are not fixed numbers of points. Gold's four sessions behave very differently, and a gap that is a short walk in the US morning can be an unrealistic ask overnight. A trader who reads the same number of points the same way in every session will be wrong in opposite directions at different times of day.

That gives you one habit worth building:

Every figure rests on a deep sample, and the Sample size row in the table tells you how much history stands behind the reading in front of you.

Four sessions, two clocks

Gold does not run on one clock, and the session map reflects that.

Most of the boundaries are New York times: the 18:00 open, 08:00, 11:30 and the 17:00 close. The overnight boundary is not. It is fixed in Shanghai, at 13:30 local time, because that is where gold's overnight activity changes gear, a conclusion the research reached from the data rather than from the calendar.

The result is visible on your chart. The right-hand edge of the Asia (SGE) box sits at 01:30 New York time in summer and 00:30 in winter, because China does not change its clocks and the United States does. That is the indicator working as designed, not drifting. A fixed New York boundary would put half the year's overnight bars in the wrong session and quietly distort every figure that depends on where the session started. The boundary was checked by replaying the chart on both sides of a daylight-saving change.

Because one edge of the overnight block follows New York and the other follows Shanghai, Asia (SGE) runs an hour longer during US summer time and Eurasia an hour longer during US winter time.

Why there is no London session

London matters a great deal to gold, but its influence arrives inside the trading blocks rather than as a step at the London open, and drawing a boundary there would assert something the data does not support. The block that covers the end of Asian trading and the European morning is called Eurasia for exactly that reason. Naming it London would claim a boundary that is not there.

Asia (SGE) is named after the Shanghai Gold Exchange, whose clock sets its far edge.

Setup

Getting it

Atlas Sessions (GC) is free on any TradingView plan, including the free one. Add it from its TradingView page ↗. This guide describes v1.7, the published build. On your chart it appears as AtlSess GC.

Symbol

GC or MGC futures only. The figures are measured on COMEX gold; MGC is the same market at one-tenth the size. They do not apply to spot gold (XAUUSD), which trades a different day on different venues. On any other symbol the table adds a warning.

Timeframe

Use a timeframe that divides 30 minutes: 1, 2, 3, 5, 6, 10, 15 or 30 minutes. Any other intraday timeframe, including one hour, does not line up with the session boundaries, and the table says so. The figures are the same on every supported timeframe because the engine works in clock time; only the precision of a printed time, such as the time on a TAKEN label, follows your bar size.

1 to 5 minutes is recommended. Above 5 minutes the table adds a note that the view is coarse, because a wide bar can straddle the moment a level is taken. A 15-minute chart is still useful for one thing: it loads far more history than a 1-minute chart, so the Hit rate row rests on a larger sample.

Chart history

Contract rolls

Gold futures roll every two months, and on a continuous chart the switch to the next contract can move price by the spread between the two. On the session where that happens, the indicator shifts the previous session's levels into the new contract's prices so they line up with the bars in front of you. That session is greyed and left out of the running tally, and the warning row shows the size of the shift (on a chart already showing another warning the line just adds "contract roll"): Contract roll: prior levels shifted, the amount in points, then greyed. If your chart was open across the roll, reload it once so the shift is picked up. On a back-adjusted chart there is no gap to correct, and the session is shown normally.

Sunday's opening session

Sunday's Asia (SGE) session opens against Friday's US PM levels across the weekend. That is not the same event as an intraday handoff, so the session is greyed, as the limitations section explains. The same applies after a holiday break. Unlike a contract-roll session, a greyed weekend session still counts in the Hit rate row.

Settings

The settings dialog has seven groups. Everything optional ships off, so a first run shows the levels and the shares on their own.

Reading the chart

Session boxes

Each session is drawn as a shaded box in its own colour, by default light blue for Asia (SGE), amber for Eurasia, green for US AM and purple for US PM, and the box grows with the session's range as it trades. The most recent sessions carry their names above their boxes; the session whose levels are in play is named on its level labels instead.

The level lines

The previous session's high and low, drawn from the start of the session that made them and carried across the session now open. By default they take that previous session's colour. Each carries a label with the session name, the price and a share for the session now open: Eurasia High, the price, a "|" separator, then the figure. The figure on the high's label is the Take high only share and the figure on the low's label is the Take low only share, the same numbers you see in the table.

A level counts as taken when price trades through it; a touch at exactly the level does not count. When that happens the label switches to TAKEN with the New York time it happened, and stays that way for the rest of the session.

If a session opens beyond one of the previous levels, that level is taken on the first bar. The four-way shares then read "-" for that session, because they describe sessions that open inside the previous range, and the session is left out of the Hit rate tally.

The four-way shares

At each open you get four percentages that add up to 100%: high only, low only, both, and neither (the session stays inside). These are the at-open figures. They are fixed when the session opens and do not change while it runs.

Live mode (optional, off by default)

Live-update % (per-side marginal, advanced) in the Display group replaces the at-open figures on the level labels with a live P(take) that updates every bar: given that this level has not been taken yet, and given where price is now and how much of the session has passed, how often was it still taken by the close. It is a separate measurement rather than a rescaling of the at-open shares, and it is where the 15:00 rule shows up. The live figures are per side: the high and the low are scored independently, so they will not add up to 100% the way the at-open shares do. Live mode also adds a Live P(take) H/L row to the table.

On a level label, a "<" in front of a live figure means the level is further away than the measured history covers well, so read it as "at most".

The EQ line

A line at the midpoint of the previous session's range, solid by default and restyleable in the Level lines group under S2 EQ line (midpoint of last completed session). Whether the new session opened above or below it is one of the context rows in the table.

Median penetration guides (optional, off by default)

Dotted lines labelled Med pen high and Med pen low, showing how far beyond each level price typically continued during the session once that level was taken. Tick Median penetration lines in the Display group to draw them. Hover either label to see the typical push 15 and 30 minutes after a take. As with all such figures, this is measured drift, not momentum: use it for target geometry, not as evidence that a break will run.

The push levels (optional, off by default)

15m/30m push levels, also in the Display group, draws dotted levels labelled 15m and 30m above the previous high and below the previous low: the typical further push beyond the level within 15 and 30 minutes of it being taken. They are known in advance, since a take only starts the clock, and they move to the new levels at each new session. The tooltip calls them a pace guide, not an edge, and that is the right way to read them.

Both guides ship off so that a first run shows the levels and the shares on their own, which is the product; the guides are geometry laid over the top. Together, the level, the penetration guide and the push levels give you destination candidates grounded in where price has historically tended to reach from a situation like this one, rather than in round numbers or hope.

Quarter levels (optional, off by default)

Project 25% / 75% levels, in the Level lines group, adds the quarter marks of the previous session's range alongside the midpoint, drawn across the current session. They are dashed and grey by default so they read as secondary to the high and the low, which remain the levels the figures are about.

They are context, not measured figures: no share in the indicator is attached to a quarter level. What they give you is a quick sense of where price sits inside the previous range. Sitting just under the previous high and sitting in the middle of the range are very different starting points, and the quarters make that difference visible at a glance.

Reading the status table

The table reads top to bottom in the order you should read it: the score, then the context block under Previous Session Summary, then the figures under Probability Metrics. Rows that start with a session name, such as "Eurasia range", always refer to the previous session, the one whose high and low are drawn.

The Hit rate row

The Hit rate row sits second from the top on purpose: it is the honesty check, and it is worth knowing what "good" looks like. The figures imply roughly 59% accuracy at picking the top outcome of the four. The naive benchmark, always picking the most common outcome overall, gets about 36%, and a uniform guess gets 25%. So the useful comparison is not 59% against 100%; it is 59% against 36%. The row puts your chart's realised rate next to the rate the figures expected for the same sessions.

Roll sessions, and sessions that open beyond one of the previous levels, are not scored. We also published, week by week from August to September 2026, what the gold build showed at the US AM open and how each session closed: see the gold Sessions record.

A "~" beside a share means that particular situation was less stable across the measured years than the rest. It is marked rather than hidden, and the marker can be switched off in the Display group.

The warning row

The row marked "!" appears whenever something would weaken the read. When more than one applies, they share the line.

The check marks

As the session runs, a check mark appears beside the share that has actually happened. Take the high and nothing else, and the tick sits on Take high only; if the low goes later, the tick moves down to Take both sides, because that is now the outcome in force. On the First sweep row the tick marks the side that went first, which is the row's whole subject.

Inside session never ticks while the session is live, and the reason is worth stating: it is the only one of the four that is an absence. The others are events: a level trades, and the tick lands. Inside is settled only by the session ending without either level trading, at which point the table has already rolled forward to the next session. A tick there would have to be a prediction that nothing further will happen, which is not what these figures are. On gold that matters more than it looks, because inside is the most common single outcome in US PM, so the row you would most like to see ticked is precisely the one that cannot be.

Taken together, the ticks turn the four shares into a running scorecard: you can see at a glance which of the four the session chose, and how that compares with the share the table gave it.

Compact mode, for phones and small screens

Compact table (mobile) in the Table group strips the table back to the rows you need at a glance: the header, Hit rate, the four-way shares and First sweep. The Previous Session Summary block, Med pen and Sample size are hidden.

It is off by default, and it changes nothing about the figures, only which rows are drawn. Live P(take), Recal n and any warning row still appear when they apply.

One thing to be aware of: Sample size is one of the rows compact mode hides, and that is the row telling you how well supported the current reading is. If a number looks surprising on a phone, switch the full table back on before acting on it.

The 15:00 rule

Gold's afternoon has a hard edge in it. Take activity falls away from about 15:00 New York time, an hour before the day's range goes quiet and well before the 17:00 close. A previous-session level still untouched at 15:00 is, historically, a level that mostly is not going to be touched: the remaining probability drops towards single digits and then to almost nothing.

This is not a rule added to the display. It came out of the measurement, it holds across the whole measured period, and it is built into the live figures rather than drawn over them, which is why you can watch it happen with live mode on.

It also explains the most distinctive figure in the gold tables: Inside session is the most common single outcome in US PM, more common than either side being taken. On an index, an afternoon that goes nowhere is a quiet day; on gold it is the base case.

The practical use is timing discipline. A plan built on a level being taken has its historical tailwind early in the session, and live mode lets you watch the decay as it happens rather than feeling it out. When the figure is in single digits and the clock is past three, waiting is not patience. It is a position with the base rate against it.

The scenarios table

Optional and off by default. Show scenarios table, in the Table group, adds a second small table (bottom right by default, moved with Scenarios position) that states the figures conditionally, one row per side:

• ▲ Session high taken → before a take, "low also" with how often the low also went before the session ended once the high had, and "run" with the median push beyond the high in points.
• ▼ Session low taken → the same for the low side, "high also" and "run".

The "also" figures are worked out from the four-way shares at the open: both sides, divided by that side only plus both sides. Once a side trades, its row switches to TAKEN with the New York time and keeps the figure for the other side; when both have gone it reads "both sides done". With live mode on, the other side's figure switches to its live value and is tagged "(live)", so you always know which one you are reading.

The day-of-week adjustment

One contextual adjustment ships with the indicator, and it is off by default: Day-of-week tilt (US AM and US PM) in the Contextual adjustments group.

It is not what you might expect, and the difference matters. It is not directional. What held up in the research is a pattern in decidedness: how often a session resolves at all, taking one side or the other rather than staying inside the previous range, varies by weekday in two of the four sessions, US AM and US PM. With it on, the weekday shifts the balance between Inside session and the three outcomes that take a level, and leaves the balance between high and low untouched. It tells you something about whether the session commits, never about which way.

It changes the at-open shares, and with them the Hit rate pick, and nothing else. The live figures, and the Asia (SGE) and Eurasia sessions, are not adjusted. It ships off because the base figures are the product and you should read them unadjusted first. Turn it on if you want the weekday context folded into the shares.

Other candidates for this layer were examined and are not in the indicator, because they did not hold up.

Where the numbers come from

The figures come from COMEX gold futures minute data spanning nearly two decades, of which the published figures use the period from 2015 onward. Prices are the contract's true traded prices, checked against known historical highs and lows to the tick. Contract-roll sessions and handoffs that span a weekend were left out, so that neither can pass for a move.

Everything the indicator displays is a measured frequency, frozen when the build was made. The shares are not estimated on your chart: it looks up what happened, in comparable situations, in the measured history.

Getting there took a long and deliberately sceptical piece of proprietary research. Many candidate influences on how a session resolves were considered, from where price opened to what the previous session had done, the shape of the handoff, the calendar and a good deal more, and each was held to the same strict standard. Most did not meet it. What you are using is what was left standing, which is why the indicator has a short list of controls rather than a long one: the short list is the finding.

The engine was checked for three things when it was built: the output must be the same on every supported timeframe, the chart must match the research data session by session, and the overnight boundary must land on the right bar either side of a daylight-saving change.

Keeping the figures current

Gold's centre of gravity has been drifting towards Asian hours for years: Asia's share of gold's daily activity has roughly tripled since 2007 and is still climbing. A change like that could slowly wear on fixed figures, so it has been checked against the published ones, and so far they have held up.

Recalibration (optional, off by default)

Live recalibration blend nudges the live figures towards what the sessions loaded on your own chart show, weighted towards the published figures by Recal strength (default 150; higher means trust the published figures more). It moves the live per-side figures only; the four-way shares at the open are unchanged. It needs a deep chart to matter, which is why the Recal n row suggests a 15-minute view while its count is thin. The published figures change only with a new version of the script.

Limitations

• Frequencies, not predictions.
• Session boundaries are fixed. They are part of the measurement, not a display option.
• Asia (SGE) changes length twice a year. Its far edge follows Shanghai and its near edge New York, so it is an hour longer in one half of the year than the other. That is correct behaviour, and it means the Asia figures cover both halves of the year rather than being split by season.
• Some situations rest on less history than others. The Sample size row tells you when you are in one, and a "~" marks the readings that were least stable across the measured years.
• Sunday's Asia (SGE) session is greyed. It opens against Friday's US PM levels across a two-day gap, which is not the same event as an intraday handoff and was not part of the measurement. Its levels and figures are shown in grey with the warning "Weekend wrap: greyed (unvalidated)", and when the Sunday open sits outside Friday's range the shares read "-". Treat any figure it shows as unmeasured. A session that opens after a holiday break is treated the same way.
• Contract rolls. Gold rolls every two months. Roll sessions were removed before measuring. On your chart the roll session's levels are shifted into the new contract, the session is greyed and left out of the tally, and the chart needs one reload if it was open across the roll.
• Short history shows nothing. With fewer than five completed trading days loaded, the shares read "-" until enough history arrives. On a free plan, use 3m or 5m.
• The quarter levels carry no probability. They are geometry drawn from the previous session's range, offered as reference. Nothing in the figures scores them.
• Measured on history. The figures describe the period they were measured on, and their stability across that period is the main evidence that they carry forward. Your Hit rate row and the published record are where that gets tested.
• GC and MGC only. Not spot gold, not silver, not anything else.
• Take probability is context, not a trade in itself. A level being likely to be taken does not, on its own, make it a profitable place to act. The next section explains how to use the figures.

Using it in practice

Use it for sequencing and context, not as an entry trigger

This is the core guidance, and it follows directly from what the indicator measures.

The take probability tells you what a session is likely to do to a level: which side tends to go first, whether the session is likely to stay inside, how far price typically travels once a level goes. That is real, measured information about session behaviour.

What it does not tell you is that the level is a good place to trade. The value is in knowing what the session usually does, and staging your own plan around that: the same entries you would have taken anyway, now judged against the base rate. Use it to decide sequencing, whether to expect range or expansion, and where your own setup is working against the usual behaviour. The entry logic stays yours; this supplies the base rate it should be judged against.

Read that way, the indicator does what it was built to do. Read as "the level will hold" or "the level will break, so trade it", it will disappoint, and rightly so.

• Start with distance. Before reading any percentage, look at how far away the levels are, bearing in mind that near in Asia is not near in the US morning. Close levels tend to get taken; far ones tend not to. The percentages are a more precise version of that.
• Use the 15:00 rule. It is a clear timing asymmetry, and gold's afternoon is where many traders wait for something that has already stopped happening.
• Use Inside session as a range warning. A high inside share is the history telling you this session has often gone nowhere. In US PM that is the base case, not an unusual reading.
• Use First sweep for sequencing, not direction. It tells you which side has tended to go first, which matters for how you stage entries, not whether the session closes up or down.
• Watch the Hit rate row. It is your own audit of the figures on your own chart. The gold Sessions record shows the same kind of check published week by week for the US AM session.
• Let it set the day's temperature. When the shares say range, you can stop paying for breakouts that have rarely carried from here; when a level is within reach, you have a destination backed by measurement rather than a hunch. A plan set at the open, with the odds known, is easier to follow than one improvised mid-session.

Using it alongside Atlas IB (GC)

The two indicators answer different questions about the same day, and the difference is in what they measure from.

Atlas Sessions measures against the previous session, using reference points that exist before the new session opens and never move. Does this session take the previous session's high, its low, both, or neither?

Atlas IB measures against the current window. Price builds its own range in a chosen 30- or 60-minute window, and the targets come from that range, so nothing is known until the window completes.

Sessions therefore gives you the session frame, whether this is likely to be a session that expands and takes levels or one that sits inside, while IB gives you the execution frame within it: specific targets, distances in points, and how often they were reached.

The natural pairing on gold is the US morning, where an 08:00 or 09:00 60-minute IB window forms inside the US AM session that Sessions is describing.

Disagreement is the more informative case. A high Inside session share alongside a wide IB extension target suggests the target is probably too far for the day you are in. A strong lean towards one side in Sessions alongside an IB abstain means you have an expectation for the session with no window structure to express it through. Either is worth knowing before you commit.

The practical order: read Sessions first for the session's character, then IB for the geometry inside it. Sessions rarely changes what you do at the window level; it changes how much you expect the day to travel at all.

Tracking it yourself

The Hit rate row scores the at-open shares on the history loaded on your chart. We also keep a public record of what Atlas Sessions (GC) showed at the US AM open and how each session closed: week by week from August to September 2026, and month by month from October 2026. See the gold Sessions record.

FAQ

Is it free?

Yes, on any TradingView plan, including the free one. On a free plan, use a 3-minute or 5-minute chart so there is enough history.

Does it repaint?

No. The at-open figures are fixed at the open. Live mode is explicitly a live feature and is labelled as such; it does not rewrite history. The one thing that changes on a reload is a contract roll that happened while your chart was open: the shift is applied when you reload.

Why the same numbers on 1m and 15m?

The engine works in clock time, so the figures are identical on every supported timeframe. Only the printed time of a take follows your bar size. Use 1 to 5 minutes for the sharpest live reads, and 15 minutes if you want the biggest sample on the Hit rate row.

Why do the shares read "-"?

Usually one of three things: the chart does not yet hold five completed trading days (the warning row says so), the session opened beyond one of the previous levels, or Sunday's session opened outside Friday's range.

Why does the Asia session end at a different time in winter?

Because its far edge is fixed in Shanghai, and China does not observe daylight saving. On a New York chart that lands at 01:30 in summer and 00:30 in winter.

Why is there no London session?

Because London's influence on gold arrives inside the trading blocks rather than as a step at the London open. Drawing a boundary there would assert something the data does not support.

Can I change the session times?

No. They are part of the measurement, not a display option.

Why is my hit rate "only" 59%?

Because the honest benchmark is 36%, not 100%. Picking one of four outcomes, with 25% for a uniform guess, is a hard problem, and beating the naive favourite by more than twenty points is the result.

Can I use it on NQ, spot gold or silver?

No. It is measured on COMEX gold. There is a separate build for NQ, linked from the Atlas Sessions page.

Can I see the tables?

No. The source is protected because the measured figures are the work product.

Support

• support@algotradernqgc.com
• @AlgoTraderNQGC

Questions about interpreting a specific session are welcome; send a screenshot.