Direct answer
Divergence in MACD means that the movement in MACD does not match the movement in the underlying price at key moments. In plain terms: price may be making a new high (or low), while MACD is not making a new high (or low) to the same extent. This difference is often interpreted as a possible momentum weakening or strengthening, not as a standalone prediction.
What MACD divergence is (definition and mechanics)
To understand divergence, start with how MACD is constructed.
MACD is typically based on two exponential moving averages (EMAs) of price (often of the closing price), plus an additional line or histogram derived from the relationship between those EMAs and a signal line. A common way to describe it is:
- Compute a “fast” EMA and a “slow” EMA.
- MACD is the difference between them.
- A signal line is an EMA of the MACD value.
- The histogram (if used) represents the difference between MACD and the signal line.
MACD divergence is then defined relative to price behavior:
- Bearish divergence: price makes a higher high, but MACD (or its histogram) makes a lower high.
- Bullish divergence: price makes a lower low, but MACD makes a higher low.
A simple model for the idea is comparison, not proof: you are checking whether the momentum measure (MACD) confirms the newest price extreme. If it does not, that suggests the momentum behind the extreme may be weakening.
How it works in practice (with an example model)
Because MACD uses EMAs, it smooths price. That matters.
Assumptions for the example model (no live data):
- Consider a period where price forms a second peak slightly higher than the first.
- MACD is computed using fixed EMA parameters and updates each time a new bar closes.
- You compare the relative height of the MACD peak near the times of the two price peaks.
If the second price peak is higher, but the MACD peak is lower, the divergence indicates that the MACD-based momentum signal did not expand as much as price did. A frequent interpretation is “momentum lag” or “momentum weakening.”
However, note what the definition does not guarantee: a divergence is a description of mismatch at selected points. It does not automatically specify when (or whether) price will reverse.
Material limitations and common failure modes
-
Parameter sensitivity MACD divergence depends on MACD settings (EMA lengths, signal line length) and whether you measure the MACD line or the histogram. Changing those settings can change where peaks and troughs appear, which can create or remove “divergence.”
-
Confirmation limits A mismatch might happen briefly inside a larger swing. Without an explicit, consistent rule for “what counts” as a peak/trough and when divergence is considered confirmed, two analysts can legitimately mark different divergences.
-
Hindsight bias Many divergence explanations are written after the fact, when the later outcome is known. That leads to selecting only the divergence cases that preceded a good move and ignoring others. Even if divergence sometimes coincides with later turns, hindsight bias can make the relationship look stronger than it really is.
-
Market and condition variability Historical chart patterns do not establish a stable, transferable relationship. Costs, execution timing, and shifting volatility regimes can change how momentum signals behave.
What you can verify independently
You can verify the concept without assuming it is predictive:
- Replicate the MACD calculation with the same parameters on your chosen price series.
- Mark the exact two moments used for the divergence decision (which price swing points, and which MACD or histogram extrema).
- Check whether the divergence appears consistently under your definition, rather than being created by “moving the goalposts.”
If you want to reduce confusion, ask next:
- Which component are you measuring for divergence: MACD line, signal line, or histogram?
- What rule determines the swing points to compare?
- Are you testing the idea with a consistent method rather than telling a post-hoc story?
Divergence in MACD is best understood as a structured observation about momentum mismatch, with limited confirmation and a high risk of interpretation bias.