When there's an imbalance of supply one way or the other, that is what causes extended moves. It's the move with high velocity over a set amount of time that creates usable and trackable levels based on certain perimeters. However, means reversion strategy alone is very difficult, mainly because it's hard to effectively define your risk. I find means reversion most useful, if lets say you have an open level to the upside or downside that hasn't been filled yet and you get a corresponding buy or sell signal, than you take the trade and look for the means reversion. This increases the odds of it being filled while also giving you a better risk management of it (as you're taking a buy or sell signal you know and have already determined a stop / risk profile).
A reasonable model of price behavior is pink noise. Therefore, a trend is just waves with a longer periods (lower frequency) as the dominant market frequency, and mean reversion is catching waves with smaller periods (higher frequency) as the dominant market frequency. When prices transition from shorter frequencies to longer frequencies, and back again, is randomly distributed in time, which is why almost all indicators don't work, or lead to positive expectancy methods. Trying to use adaptive indicators that adjust according to the dominant cycle present in the data don't seem to work well in my opinion due to lag (group delay) or insufficient attention of the higher frequency data in the stop band.