A trending market moves persistently in one direction, making higher highs and higher lows or the reverse, with pullbacks that do not reverse the direction. A ranging market moves back and forth between a ceiling and a floor without going anywhere. Every market does both, switching between them without announcement, and it spends more time ranging than trending.
Why it matters
Trend-following and breakout strategies make money when a move continues and lose when it does not. In a range they are stopped out repeatedly, buying the top and selling the bottom. Mean-reversion strategies make money when a move reverses and lose when it does not. In a trend they fade a move that keeps going, and the losses are large. Neither kind of strategy is wrong. Each is wrong half the time, and the half depends on the state of the market.
Measuring it
There is no clean measure, but there are usable ones. The average directional index, ADX, is a standard indicator that rises when price is moving consistently and falls when it is not; readings below about 20 are usually taken as ranging and above 25 as trending. The relationship between price and a long moving average, and the slope of that average, gives a similar reading more slowly. Simple inspection of a daily chart over a few months is often as good as either.
None of these identifies the state in advance. They describe what the market has been doing, and a strategy that uses them accepts that it will sometimes act on a state that has just ended.
Over a year, EUR/USD trends for about four months, rising 900 pips, and ranges for the other eight within a 300-pip band. A breakout strategy makes 500 pips in the trending months and gives back 400 in the ranging ones. A mean-reversion strategy loses 300 in the trending months and makes 450 in the ranging ones. Each is profitable for the year. Each has a period it would rather forget. Run together, the year is smoother than either alone.
The state of the market is the largest single reason an EA's results vary from year to year. An EA that only trades one way, always breakout or always mean reversion, will have long stretches that look like failure and are simply the wrong state. Its record should be judged across at least one full cycle of both.
Multi-strategy EAs exist largely to address this, by holding strategies that need opposite conditions. A well-built one also uses some measure of market state to decide which strategies are allowed to trade, and its documentation will say that it does, even if it does not say how.