I assume this means something is lost during compression and you lose more of it during “light” compression presumably because it’s faster it doesn’t process nearly as much, but I can’t quite figure out what would be lost. I mean, the whole point of baking is so you don’t have to calculate anything again, if I’m understanding things correctly, so does compression force you to have to re-calculate portions of the simulation when you reload the cache? If that were the case, why would “heavy” be more accurate if it also makes the file size much smaller than “light?”