Now for a "fun" experiment.
It took me some time to figure out that the (non-boss) enemy difficulty grows with number of rounds and damage dealt to castle. But if you can avoid dealing too much damage to the castle, a lot is possible. Like having to sell lvl 3 Things to make space. Or stacking several rows of coins each round.
To reduce dependence on luck you basically have to spy every round (3 x free, then 2 coins per round) for quite some time.
2 lvl 3 Bat Curse books can take out 6 units which is really helpful. But then there are rounds with 7 (or 8?) archers, so a third book was nice to have. Even if they don't die right away, replacing their bows is great for planning.
Also keeping different books inactive behind the front line can be beneficial for adapting to current enemies every round.
It feels a bit weird to rearrange the units every time to maximize your own casualties – e.g. supporter in front, but it can be done … even if it costs some of your own health during the first few rounds.
Using multiplying summons makes the planning rather difficult – as do random ranged attacks –, so most of them loose their use soon.
For this delaying strategy it is very useful that lvl 3 units only deal the same 2 damage to the castle as lvl 2 units, so you don't have to delay the upgrade too much.
I could have probably squeezed out a few more rounds if I had sold the succubi … but they are simply too cute <3.
EDIT: … And a whole lot more when focusing on only 3 units … as several older comments on previous pages point out. Why didn't I check there before? Now I feel a bit dumb. But this was still quite fun for a first try of a delay approach.
 |  |
 |  |