Where the reading comes from
The calculation starts with two simple quantities per bar. The up move is how much the current high exceeded the previous high. The down move is how much the current low fell below the previous low. Only the larger of the two counts for that bar, and if neither extended, the bar contributes nothing. Those raw values are smoothed and then divided by a measure of the bar range, producing two lines known as +DI and -DI.
The next step is where the sign disappears. The index takes the absolute difference between +DI and -DI and divides it by their sum. A market where almost all of the extension is upward gives a large difference and a high result. A market where up and down extension are roughly balanced gives a small difference and a low result. That ratio is smoothed once more, and the output is ADX. Two smoothings sit between the raw bars and the line you read, which matters later.
Why the output carries no direction
Taking an absolute value throws away which side won. A steady decline where every bar extends the low produces the same large difference as a steady advance where every bar extends the high, so the index rises in both. This is not a flaw, it is the design: the question being asked is how one sided the extension has been, not which side it favoured.
In practice that means a rising index is a statement about the market being organised. It could be organised downwards. Traders who glance at a climbing ADX and read it as bullish are importing a direction that is not in the number. Direction has to come from somewhere else: structure, the slope of a reference average, a higher timeframe read. The index then tells you how much respect that direction currently deserves.
Slope matters more than the level
The habit of quoting threshold levels, often a figure in the low to middle twenties for a trend and something lower for no trend, comes from the original description of the tool. Those numbers are conventions, not constants, and nothing in the mathematics makes them special for gold or any other market.
What carries more information is the direction the line is moving. A reading that is low but rising says the extension is becoming one sided, and that happens at the start of moves. A reading that is high but falling says the market was organised and is becoming less so, which often arrives as a trend starts to broaden into a range. The level alone cannot distinguish those two states, and they call for opposite handling. Traders who find the index useful usually watch its slope and treat any threshold as a rough fence rather than a trigger. The same point applies to the slope of a reference average in moving averages on gold.
What the two DI lines add
The +DI and -DI lines are where the direction lives, and they are the weaker half of the package. A crossover, where upward extension overtakes downward extension, looks like an entry signal and behaves like a late one. Both lines are smoothed, so the cross is confirming a shift that has already been underway in the bars.
Used as a state label the lines are more honest. Which one is on top tells you which side has been doing the extending recently, and how far apart they sit gives a rough sense of how lopsided that has been. Convergence of the two lines is often the earliest visible sign that an organised move is losing its one sidedness, which tends to show up before the strength reading rolls over. None of this is a trigger. It is a description of the current state, and the entry still has to come from a level or a structural event.
The lag you are paying for
Double smoothing is the reason the line is readable, and also the reason it is late. A breakout that a trader can see on the chart in real time will not show up as a clearly rising strength reading for several bars, because the raw ratio has to feed through two averages before the output moves far.
That lag has a practical consequence. By the time the index clearly confirms a trend, a good part of the initial expansion has happened, and entries taken on that confirmation are entries into an extended move with a wider stop. This is the standard cost of any confirming indicator, and the way traders usually handle it is to stop using the index for entry at all. It becomes a filter that is checked before the session, or a context reading that decides which playbook is live, while the entry comes from price itself.
Using it to choose a model, not a trade
The most defensible use of the index is as a switch between playbooks. When the reading is low and flat, the market is extending in both directions and mean reverting approaches at range boundaries are at least coherent. When the reading is rising, pullback and continuation approaches are coherent and fading the extremes is not.
That switch also governs stop placement. In an organised move, a stop placed at the distance that worked in a quiet range will be taken out by ordinary noise, which is why volatility based distances exist. The method is described in using ATR for stop placement on gold. The Market Regime label on the live chart does the same job as a plain statement rather than a number, and it is worth saying that its real contribution is telling you when to stand aside, not when to act.
Where it misleads on gold specifically
Gold has a particular habit that the index handles badly. A scheduled release or a headline can produce one enormous one sided bar followed by a return to drift. Because that bar extends in a single direction by a large amount, the raw ratio spikes and the smoothed line begins to climb. For several bars afterwards the index reads as though a trend is building when all that happened was a single repricing.
The honest limit is that the index cannot distinguish a repeatable one sided process from one violent bar. Checking what caused the expansion before accepting the reading is the only real defence, which is part of the routine in trading gold around news. A strength reading built mostly from a news bar deserves to be ignored until ordinary bars confirm it, and sometimes they never do.
FAQ
What ADX level means gold is trending?
Threshold numbers in the low to middle twenties are a convention from the tool as it was first described, not a property of gold. The slope carries more information than the level: low but rising describes a move becoming one sided, while high but falling describes an organised move breaking down into rotation.
Can ADX tell me whether to buy or sell gold?
No. The calculation takes an absolute difference, which removes the sign, so a strong decline and a strong advance produce the same rising reading. Direction has to come from structure, a reference average or a higher timeframe view. The index only says how one sided recent extension has been.
Which timeframe suits ADX on gold?
Higher timeframes give a steadier reading because fewer bars are dominated by spread and single ticks. Many traders check it on the four hour or daily to label conditions and then execute on a lower timeframe. On very short charts the double smoothing makes it late enough to be of limited use.
What is the difference between ADX and ATR?
ATR measures the size of recent bar ranges, so it describes how far price moves in a typical bar. ADX measures how one sided that movement has been. A market can have large ranges with no direction, which shows as high ATR and low ADX, and the two readings answer separate questions.
Does ADX work during gold news spikes?
It reacts, but the reaction can be misleading. One very large one sided bar pushes the smoothed line up for several bars afterwards, which reads as a trend forming even if price went straight back to drifting. Confirming the move with ordinary bars before trusting the reading is the usual safeguard.
ⓘ See these ideas on real price: open the free XAUUSD live chart.