The general concept of reversion implies that a trend comes to an end and reverses to some degree. This kind of price action can sometimes occur after recent strong performance - as there are always investors who want to lock-in some profits and so sell their investments.
If this type of profit-taking price action starts to accumulate and other investors follow suit, this can lead to reversion. If for example there is no new fundamental news out to help support the stock, then this profit-taking can cause the stock to fall quite quickly as others panic and sell, as they fear they have missed their profit-taking opportunity.
Reversion is a behavioural factor and therefore is not based or backed up by balance sheet data like for example value or quality. Reversion is a price-based signal and construction methods for it differ from user to user.
The Fama and French market neutral short-term reversion factor is gross of transaction costs and therefore presents a very optimistic set of results. Reversion in practice is also notorious for being difficult to capture the theoretical results - this is due to missed fills, slippage and trading impact playing a large role. These can be difficult to capture fully in a backtest.
If we look at the years in isolation the picture becomes even more rosy and over-optimistic for Fama-French's reversion factor. Apart from the massive drop in performance of the reversion factor after the Fed stimulus package took effect and the bull market returned, the short-term reversion factor has apparently delivered flat to 15% annual returns every year since 2010. Too good to be true it must be said.