Thread:FixSRwiki/@comment-46261745-20200903015444/@comment-26544499-20200911200209

Wow, I'm running late. Busy day...

I think that we agree on what the unit categories are: pre-existing in army but not scripted to appear, spawned for this battle, and pre-existing but scripted to appear. That's what my memory says anyway. Your guess about military power is probably right.

I think that a lazier and more consistent solution for BRD would be to make an array of values for what effect it has, but make BRD2 start earlier in the array. For example, if we had the array a=[9,7,5,3,1,0] and a shared counter for how many times either move has been used then BRD1 could start at 5 (a[counter+2]) whereas BRD2 could start at 9 (a[counter]). See what I'm aiming for? It avoids any strange arithmetic and makes things really easy to change if the results aren't liked. Of course, I've got no idea what the actual numbers should be... darn.

You probably want the description to make it clear that the diminishing returns is shared between versions.

I agree on those two Charge moves, but don't forget that there's two versions of Commander Charge - One for Warriors and one for Cavalry.

I'm unsure about not nerfing BRU. Don't forget that the player can capture generic Cavalry and upgrade them until they have it. The move cost is a bit of a pain, particularly on generics (what's their maximum move count, four?), but I could definitely break the game with them. It probably only needs some diminishing returns. Just make the increase look at an array (like I've suggested above) rather than have it be a constant.

Unless the enemy is winning by a truly massive amount or it's very near the end of the battle, then perhaps with some RNG, I think the priority should be something like: truly awesome buffs > full power BRD2 > important buff removals > full power BRD or mid power BRD2 > mid power BRD or low power BRD2 > not great buffs or nerfs > certain kills > terrible BRD/BRD2 > normal attacks. However, I can recognise that what I've just listed probably isn't very helpful or insightful, so maybe this just needs testing.