But the core principles of the game haven't changed.

The point is that they have changed, and significantly, almost radically. Completely different logistics and micromanagement (no chains, red rush, growth, but with flag capturing, fog of war, etc.). Completely different methods and priorities for clearing the map (emphasis on development and leveling up over economy, unlike the third game's mechanics).

It seems that we have a situation where either you don't understand me, or I don't understand you. I'll try to clarify.

I meant the balance in using heroes and units from the Castle in the army. And not the balance between races (it's always been flawed).

Ah, that's what you mean. But this is exactly what plagues the second game. Because in the army, there can only be a few types of troops: archers, titans, minotaur kings, phoenixes, and skeletons. The rest are useless. And the hero is definitely a combatant.

But in the fourth game, the balance between heroes and the army is quite clearly defined. But! It should be noted that the fourth game is the most sensitive to the quality of the map. That is, the balance of the army depends entirely on the map. If it's a rush map, you'll have a combat hero and a scout; if it's too closed and rich, monsters will be used with a tactician. It's very difficult to find a middle ground, and that's why only maps created by online players or edited by them are played. And the lack of a random map generator also has an impact, of course (I'm not talking about Salnikov's generator, you can't play on it). The problem of maps is very acute. You start a fan-made map and start encountering imbalance after imbalance.

In my opinion, the fifth game has more tactical development options. The third game is more interesting in terms of the battlefield. IMHO.

I won't argue about how interesting it is. Tastes differ. But there's no point in arguing about invariance and depth. In the fourth game, on the battlefield, the heroes, that is, in one turn, a completely unimaginable number of actions can be performed. I don't consider myself a specialist in probability theory, but Horn, for example (I keep remembering him today), being a teacher of this discipline, didn't hesitate to calculate it. The battle in the fourth game is much deeper. Again, if you're interested, this can be easily illustrated with one game against a pro.

Added 3 minutes ago

All the previous parts, because the fifth game is largely an evolution of the fourth game, which, for commercial purposes, was disguised as "the third game with initiative."

The fifth game is too. Because the battle mechanics are from the third game. I'm not saying that's bad, but the third game's battle mechanics are objectively less invariant, what's there to argue about, even though it's complicated by creatures. The same applies to logistics and micromanagement. The fifth game greatly surpasses the previous ones in castle development, that's true, it's hard to argue with that. But the other aspects remain the same.