But they become less useful over time, as anything you got would be obsoleted (it's surprising how different those percentages look when compared to "reasonable" games from 1 year ago), any effort put into these will have become wasted time in the future.
To the point where you get more accuracy by removing old games from the equation, specially if modern engines are able to find stronger moves that you didn't have, and then old weaker moves are just adding noise. This is a reason people still lose games, they follow old lines produced by old engines that made them look good (and by "old", I mean July 2019 lines, 6 months are enough for things to become obsolete. The percentages you produce today will become obsolete by September 2020, unless engine improvement stops to a halt, then they'd last longer.)
You can come up with the strongest 1.g4 book and be very happy about it, but chances are by 2021 you'll take a look and see how poor it'll have aged.