(Reddit version of this post: Reddit - Dive into anything )
I’ve been seriously thinking about this for a bit now, with all the valid concerns popping up around mutaplasmids.
As they stand currently, mutaplasmids allow for “perfect rolls”, which really are the base most worrying thing about them, especially when fitted to the right ships. I see a lot of worrying, and rightfully so, but not many actual suggestions to solve the issues with them. Instead of allowing them to positively influence every stat, then, how about splitting it into a more evenly distributed, balanced system?
I’ll put together an example here - as it is now, this is the range for the different stats affected on an abyssal warp disruptor created using an “unstable” mutaplasmid (the potentially strongest-affecting one):
CPU usage: 0.7x - 1.5x
Optimal Range: 0.6x - 1.4x
Activation Cost: 1.0x - 3.0x
Currently, a “perfect roll” would look like this if you were using this mutaplasmid on, say, a Dread Guristas Warp Disruptor:
CPU usage: 0.7x23tf = 16.1tf
Optimal Range: 1.4x26,000m = 36,400m
Activation Cost: 1.0x20GJ = 20GJ
A perfect, all-around increase like this makes the module better than any other in any way - highly messing with the meta and affecting the game in very worrying, power-creeping ways. One possible solution to this problem would be to instead apply a “spread” to the potential increases and decreases, taking every affected stat into consideration.
Visualize the potential increase and decrease of each stat as a point-based system, where you have a balanced pool of potential change in both positive and negative directions - not talking about the solid numbers for the stat affectations themselves, but as a spread percentage increase or decrease for that stat - Actually, I’ll do it for you:
Activation cost is capped at a max positive increase of +0 points, as the range for potential change allows for no positive increase.
With a total of 100 possible positive points and 100 possible negative points distributed among the affected stats, the formula ends up being more of a thing where for every bit increased on one stat, another stat has an equal relative percentage decrease, balancing to a zero median. Here’s an example using the same Disruptor as above:
Example roll:
CPU usage: +25 points = (difference between max positive [.7x] and base [1.0]) x (25/100) = 0.075, sub from base = 0.925x
Optimal Range: +75 points = (difference between max positive [1.4x] and base [1.0]) x (75/100) = 0.3, add to base = 1.3x
Activation Cost: -100 points = (diff bt/w max negative [3.0x] and base [1.0]) x (100/100) = 2.0, add to base = 3.0x
Result:
CPU Usage: 23tf x 0.925 = 21.275tf
Optimal Range: 26,000m x 1.3 = 33,800m
Activation Cost: 20GJ x 3.0 = 60GJ
So while it’s still possible to get strong positive results, it comes with an equal negative - in this case, a 3x increase to activation cost for the module, making it cost 60GJ to activate. This range-balancing can apply across all mutaplasmid types, making for far less of a damaging impact on metas than if perfect, all-around better rolls were possible.
Hope this helps to serve as a possible inspiration or guide for balancing the new mechanics - they’ve got potential, but need polish.
o7
-Uriel