Nice. I think that an ideal to work for is "Tier represents how many T1 recruits you can take on average." For example, 1 T6 banner knight can kill 6 T1 recruits, but no more than that. To me that seems nice and easy to remember, and reasonably realistic.
Since I have nothing better to do, let's analyze this statement and why it's not a good idea.
To simplify things, let's assume the TN troop is standing and surrounded by N T1 troops hitting him, while TN is hitting one T1 at a time until the T1 killed.
Let's say S is the number of seconds TN needs to kill a T1, and D is the damage per second a T1 inflicts on TN.
So for one on one combat, TN will kill T1 in S seconds, and receive S*D damage in return.
You with me still? Then let's see how much damage TN receives as the number of T1s grows.
1 T1: S*D damage
2 T1: S*D (from the first T1 to be killed) + 2*S*D (from the second)
3 T1: S*D + 2*S*D + 3*S*D = (1+2+3)*S*D
N T1: (1+2+...+N)*S*D = N*(N+1)*S*D/2
So you can see that the combat skill (weapons, armor, skills) of TN as N grows needs to grow at a square rate. That means T3 (N^2=9) needs to be at least twice as good at combat than T2 (N^2=4). T6 is nine times better at combat than T2, etc. That''s a pretty steep rate.
A more conventional tiering system would be a linear one, where T6 is only three times better at combat than T2.
The conclusion is that the number of T1 killed is not a useful standard for determining what a tiered troop should do. Stick with linear increases in weapon/armor stats and skills or something.
Or you could turn this around and say let's adopt a linear system, but still use how many T1 a troop can kill - use square roots.
For example, a T2 should be able to kill 1.4 T1 troops, a T4 should be able to kill 2 T1 troops, etc. or in general a TN should be able to kill sqrt(N) T1 troops.
Hope that helps!