That "study" is funny. First of all, the difference from average is unimportant in what is trying to be measured. What is important is the correlation between wins and rank in those categories. More important, the assumption in the analysis is that those teams ignored offense in their attempt to field well.
This isn't the case. It is about run differential. I think the Sox knew they would have a less productive offense, and the attempt was to mitigate that by an equal level of run prevention. If you can keep the run differential static, your expected won/loss record improves as the overall scoring decreases. For example, with a run differential of 162, you are 162-0 when your run scoring for the season is 162-0 (1-0 win every game), and you are 82-80 when your run scoring is 2162-2000. Yes, these are unlikely extreme examples, but they are used to quickly demonstrate the concept.
These teams didn't just try and win the UZR/150 measure. They tried to offset any offensive loss with defensive gain. Whether or not that will happen over a full season remains to be seen. My guess is that forecasting run prevention is much more difficult than forecasting run scoring, so there is more risk attempting this strategy.