Moving average convergence divergence, developed by Gerald Appel, is arithmetic on top of two exponential moving averages. The MACD line is the twelve-period EMA minus the twenty-six-period EMA. The signal line is a nine-period EMA of the MACD line. The histogram is the difference between the two. Those three numbers are the whole indicator, and the default settings come from Appel's original work rather than from anything about crypto.
What the MACD line measures is the gap between a fast and a slow average — a proxy for the rate at which the trend is changing. Above zero, the short average is above the long one and the trend is up on that timeframe. Rising above zero, the trend is not just up but accelerating. The signal line crossing is a smoothed way of marking when that acceleration turns.
Because it is built from moving averages, it inherits their lag exactly. A MACD crossover cannot occur until enough closes have accumulated to move two averages past each other, so it confirms rather than anticipates. In a sustained trend that confirmation keeps a position on the right side. In a sideways market the two lines cross repeatedly around zero, producing a stream of signals that cost fees and go nowhere. This is not a defect in the settings; it is what happens when a trend-following construct is applied to a market with no trend.
The histogram is the more informative part and is often ignored. It measures the gap between the MACD line and its signal, so it shrinks before they cross. A histogram contracting while price continues to make new extremes is momentum fading ahead of the crossover — the same observation RSI divergence makes, arrived at differently.
Divergence applies here too. Price making a higher high while the MACD line makes a lower one says the second advance was weaker in momentum terms. As with RSI, this describes character rather than timing, and it can persist far longer than a position can be held against it.
Two cautions specific to this market. Crypto's volatility means the default twelve, twenty-six and nine were calibrated on a different asset class and a different session structure, and there is nothing sacred about them. And on very short intervals the indicator is mostly measuring noise: it is a construct of averages, and averaging a minute of a thin book produces an average of nothing much.