How do I take stat-differences and make them into one-stat difference?

Using hockey again as the best way to explain this, say I have projected a certain team's powerplay with go from being 19% efficient last year to 20% efficient this year, and I've also projected their penalty kill will go from 78% to 79%.

So I have my powerplay and penalty kill projections for next year, but how do I combine these two stats, or every stat, into one value that represents how much better they will be overall? Or how do I combine all the offensive ones into "offense" and the defensive ones into "defense?"

Another problem I'm having is this. 19% to 20%, that's 1% difference. 78% to 79%, that's 1% difference. The problem Im having is that when I do the divisions in my formula to calculate the differences, the size of the numbers is a problem. Like those two should have the same differences, right? Except when you divide 20 by 19, you get 1.05, but when you divide 79 by 78 you get 1.01 because they're bigger numbers. When really, the size of the numbers shouldnt matter because the range of powerplay and penalty kill are roughly the same. Most penalty kills range from 15% to 25%, and most powerplays range from 70% to 80%, for example.

So Ive confused myself a bit on the math, which is one problem, but the other im having is, with all these projections i have on how much better or worse "teams" will be in all these various statistical categories, I dont know how to distill that into less categories so it's easier to analyze. At the very least the offensive ones should be able to distill into one or a couple categories, and the defensive ones into one or a couple. Im not sure you can combine them but I cant even figure out how to do it without combining them.

So how do I do this? Thank you ppl!

I dont want to get 79% or any given number, I want the correct number in a format that can be compared to other stats. What I have been doing is diving this year's number last and then subtracting by 1. It works fine for each stat category just not when I need to combine different categories into a new stat.