I believe spells like arrows, log, freeze, rocket, zap, snowball, and clone should receive more attention.
Arrows and log are overpowered, with usage of 36% and 32% respectively in Grand Challenges (see image below).
Freeze is statistically balanced around 7% usage, but it may be frustrating to play against since it can freeze your defense for quite long. Thus, reduce duration and elixir cost to compensate.
Rocket, snowball, zap, and clone have usage between 0% and 4%, so they need buff.
Mirror is underpowered, but since it can only be changed in coarse intervals (level boost and elixir cost), it is very difficult to balance. However, small adjustments may make the card somewhat more viable in casual play, without making it overpowered (an example of a possible change is provided under the image).
7-day Grand Challenge usage rates on Dec 30:
For your consideration:
Arrows (nerf), radius 4 → 3.5
The log (nerf), knockback distance 0.7 → 0.5
Freeze (rework), elixir cost 4 → 3, duration 4s → 2.75s
Rocket (buff), crown tower damage 25% → 30% (or a radius increase)
Snowball (buff), damage 179 → 192 (same as zap), slow duration 3s → 3.5s
Zap (buff), stun duration 0.5s → 0.7s
Clone (buff), radius 3 → 4.5
Mirror (buff), no mirror animation, it is possible to mirror a card after playing a champion, mirror to increase the level of spells by 2 instead of 1 (except goblin barrel, graveyard, clone, barbarian barrel, and royal delivery)
(rationale: Troops/buildings get both damage and health increased by 10%, but for spells it only increases damage by 10%, thus it is more worth it to mirror troops. This bias is removed if spell level is increased by 2, and should only be applied to spells that don’t spawn troops.)