What a divergence claim actually compares
Divergence compares two pairs of pivots. You take two peaks or two troughs in price, then the corresponding peaks or troughs on an oscillator, and you ask whether the second one moved the same direction in both. That is all the comparison contains. Which oscillator you use changes the arithmetic but not the logic.
It is worth knowing what the common oscillators compute. RSI compares the average size of up closes with the average size of down closes over a lookback, expressed on a bounded scale. MACD is the difference between a fast and a slow moving average, with a signal line that smooths that difference. Both are functions of closing prices over a window, which means both are describing the size and pace of recent legs. The crucial implication is that an oscillator peak is lower than the previous one mainly when the latest leg was smaller or slower than the earlier one. Keep that sentence in mind, because it explains most of what follows and most of the disappointment people experience with divergence.
Regular divergence, stated as a rule
Regular divergence, also called classic divergence, is the version most people mean. Bearish regular divergence is price printing a higher high while the oscillator prints a lower high. Bullish regular divergence is price printing a lower low while the oscillator prints a higher low. The reading usually attached is that the extreme was reached with less force than the previous one.
Written that way it sounds like independent evidence, but the oscillator peak fell because the leg into the new high was smaller or slower. Regular divergence is therefore a formalised way of saying the latest push was weaker than the last one, which is something you can read directly from candle size and the time taken. That does not make it useless. Having it stated mechanically stops you from arguing with yourself about whether a push looked weak. It does mean you should not treat it as a second opinion. It is the same opinion, computed. Nothing in the definition says the trend ends, only that this push cost more and delivered less.
Hidden divergence points with the trend
Hidden divergence, sometimes called concealed or reverse divergence, swaps the roles. Bearish hidden divergence is price printing a lower high while the oscillator prints a higher high. Bullish hidden divergence is price printing a higher low while the oscillator prints a lower low. Because the price extreme is the less extreme of the two, this is normally read as a continuation condition rather than a reversal one.
The mechanism is easier to see with an example in a downtrend. Price makes a lower high, so structurally nothing has changed. The oscillator makes a higher high, which means the rally into that lower high was relatively energetic for its size. One reading is that the bounce used up more effort than it gained in ground, which is a weakness argument for the counter trend move. Note the direction carefully, because this is where people make mistakes. Hidden divergence at a lower high in a downtrend is a bearish observation, and the same pattern inverted is a bullish one. It is the only one of the three families that argues for continuation.
Exaggerated divergence: the flat leg case
Exaggerated divergence is the double top or double bottom variant. Price makes two highs at roughly the same level while the oscillator makes a lower high, or two lows at roughly the same level while the oscillator makes a higher low. Structurally neither price pivot exceeds the other, so there is no new extreme to interpret, and all the difference sits in the oscillator.
This is the most fragile of the three, for a simple reason. Two highs at the same level with a lower oscillator reading tells you the second approach was slower, which is often just a consequence of the approach taking more candles. Spend longer grinding toward the same price and almost any momentum measure will read lower, without anything meaningful having changed. Treat it as a note about pace rather than evidence of a turn. It is most useful as a prompt to look at the actual behaviour at the level, which you can do with ordinary tools: did the second test produce a rejection, or did it simply arrive slowly?
Settings and pivot choice change the answer
Two decisions determine whether a divergence exists, and neither belongs to the market. The first is the oscillator lookback. Shorten an RSI period and its peaks move; divergences appear and vanish accordingly. The same is true of the two averages behind MACD, where changing either one changes the shape of the whole line.
The second is which pivots you connect. On any gold chart with a few swings in it there are several candidate peak pairs, and some produce divergence while others do not. Because nothing in the method tells you which pair is correct, the trader chooses, and the choice is made with the current position and the current hope already in mind. The remedy is procedural rather than analytical. Fix the oscillator settings and never adjust them to produce a signal. Define a minimum swing size that qualifies a pivot. Mark the pivots before you ask the question. If a divergence requires you to skip over an inconvenient intermediate peak, it is not a divergence, it is a preference.
Where it misleads in a gold trend
The dominant failure mode is the divergence chain. In a sustained gold trend, each push can be smaller or slower than the one before while the trend continues, so the oscillator prints a lower high on every leg. You get a sequence of bearish readings, all technically valid, through a move that keeps going. Anyone who traded the first one was early, the second one was also early, and only the last one looks correct in retrospect.
This is why divergence is better used as a filter than as a trigger. It can tell you to tighten management, to stop adding, or to require more before taking a continuation entry. It is a poor reason to take a position against an established direction on its own. The same caution applies to divergence read against volume on gold, because what is plotted on spot charts is a count of price updates rather than traded contracts, which is covered properly in the volume divergence discussion. If you want to study the chain effect, scroll back through a trending stretch on the live chart and count how many valid divergences appeared before the move actually ended.
A workable way to keep it
If divergence stays on your chart, give it a defined job. Use regular divergence as a reason to stop adding to a position and to tighten where you would exit, not as a reason to reverse. Use hidden divergence as a supporting note for a continuation idea you already had for structural reasons. Treat exaggerated divergence as a prompt to go and look at behaviour at the level rather than as a signal in itself.
Above all, write down the settings and the pivot rule, then leave them alone. Most of the inconsistency traders experience with this tool is not caused by the tool. It is caused by the two free choices inside it being made differently each time, usually in whichever direction supports the position already open. Fixing those choices turns an unreliable idea into a consistent one, even if what it describes turns out to be less than the vocabulary implies. Consistency first, then you can judge whether it earns its place.
FAQ
Which divergence type signals a reversal?
Regular divergence is the one usually read that way, but note what it actually states: the latest extreme was reached with a smaller or slower leg than the previous one. That is a comment about the last push, not a claim about the trend. Hidden divergence argues for continuation, and exaggerated divergence is mostly a note about pace.
Why does divergence keep failing in a gold trend?
Because a trend can make each push slightly smaller or slower than the last while continuing for a long time. Each of those legs produces a valid divergence, so you get a chain of readings through a move that does not turn. The signal is not broken. It is being asked to do something it was never able to do.
Does changing the RSI period change whether divergence exists?
Yes, which is one of the most important things to understand about it. The oscillator peaks move when the lookback moves, so a divergence can appear or disappear purely from a settings change. Fix the setting before you look and never adjust it to produce the reading you want.
Can I use divergence on spot gold volume?
With care, and with a caveat. The volume plotted on spot gold charts counts price updates rather than contracts traded, so it describes activity and volatility rather than buying against selling. A divergence between price and that series is an activity observation. It is not evidence about which side was being aggressive.
What is the single best discipline for using divergence?
Mark the two pivots before you ask the question. Most unreliable divergence is produced by choosing the convenient pair of peaks after a position is already open, and skipping an intermediate peak that would break the comparison. A fixed pivot rule plus fixed oscillator settings removes the majority of the problem.
ⓘ See these ideas on real price: open the free XAUUSD live chart.