Imagine an RTS game where so-called basic units play the main role throughout a whole game session.
In my game basic units are actually programmable by a player (like in Colobot/Ceebot), that's why it is so important to keep them significant from the first to the last second of each session.
Problem
I found out that end-game of a typical session is quite boring, no impression of... say, development of a game since middle-game. Don't know how to describe this feeling correctly.
I'm looking for reference to other games where such problem of end-game was resolved.
Thanks in advance.
UPDATE
Some more background:
The game is about splitting tasks which are best solved by chip and by brain. Player does not have to micromanage, he or she writes algorithms (in advance) instead. Players are to make strategic decisions and give high-level orders. I.e. if you have better combat algorithms for a group of bots == you're better at micromanagement. The same goes for engineering units in home front.
The typical border between middle-game and end-game is when all "free" resource sources are conquered by one of players, resources are being drilled, roads a constructed and factories produce bots at full speed (everything is automated by players in a way they want). Frontier line is more or less clear and stable.
From this point there is no choice of rush/produce/defend, you are to collect forces, invent tricky maneuvers and attack.
At this moment there is only one work for players brains: tricky tactics. It is something human brain excels at (in comparison to algorithms).
Balance is so that frontal attacks are very costly (good defense can destroy a great amount of units if an attack is frontal). Therefore attacking from multiple directions, "red herring" и fast relocation along the front line become very important (exactly like at wars in the middle of XXth century, e.g. WWII).
Again, this is where the human brain is much better than any program, so it makes no sense [for players] to write algorithms for such tasks.
My concerns is that in end-game it becomes just like a hexagon game with a very limited amount of units since all underlying tasks are "encapsulated" by algorithms.
Supreme Commander resolves such problem by introducing experimentals, Populous does it with spells (I remind, Populous: The Beginning has only five basic units).
Again, thanks in avance :-)