Two errors actually.
When you "Increase" something by 50%, you actually need to multiply it by 1.5. I think they said "increase" and implemented "percent of" (so just multiplied by .5 and left out the original 100%). Either way, to consistently change something in the same direction starting with the base value and taking a percent of it, all values have to be on the same side of 100% (higher or lower). So it should either have started at 150% and gone up, or at 66% (or something like that) and gone down. They could go either way depending on which value they applied it to. That would have resulted in a consistent improvement as the rally was upgraded.
On top of that, attack speed and time between attacks are inverses of each other. So they either applied the operation to the wrong one with the right operation, or applied it to the right one but did the wrong operation. The effect would be the same. My math used time between attacks because that is the easiest thing to measure directly.
When you "Increase" something by 50%, you actually need to multiply it by 1.5. I think they said "increase" and implemented "percent of" (so just multiplied by .5 and left out the original 100%). Either way, to consistently change something in the same direction starting with the base value and taking a percent of it, all values have to be on the same side of 100% (higher or lower). So it should either have started at 150% and gone up, or at 66% (or something like that) and gone down. They could go either way depending on which value they applied it to. That would have resulted in a consistent improvement as the rally was upgraded.
On top of that, attack speed and time between attacks are inverses of each other. So they either applied the operation to the wrong one with the right operation, or applied it to the right one but did the wrong operation. The effect would be the same. My math used time between attacks because that is the easiest thing to measure directly.