You wanna know why quantity > quality ? Here's the answer !

I was lloking into Floris' Module Systeme. I guess it works for Native too.
the "strength" of a unit for AI battles is calculated by the formula :  2*(LVL+4)² 
The wage of a non-mounted, non-mercenary is (LVL+3)²/25
In this graphics, you can see quality in blue (divided by 100 for scaling reasons; we don't care, since strength is an abstract calculation unit) and base wage (so it's reduced with leadership or garrisonning)
in the table, quality is f(x) and wage is g(x)
So, fore example, 5 troops at LVL 6 will have the same strenght in autocalc as a lone lvl 18 guy, but the 5*LVL will cost 15 denars, while the LVL 18 costs 18 denars.
If you want the AI not to attack your castles, just stack peasants if you are short on money !
(and remember that quality is usually better in manually fought battles)

I think the domestic policy "Quality VS Quantity" changes these values


Graphics powered by www.mathe-fa.de

I see that Math Detective bisthebis25 is on the case!

Seriously, though, thanks for sharing this. I always thought that the best option was to get a bunch of mid-level troops for a combination of strength and cheapness. Now I'll just place 500 Recruits in Wercheg and hope that I don't get trapped inside 

Actually, my previous graph was a bit of an exageration. To be honest, medium units are good too, and are far better in the case of a player-fought battle.
This graph is more accurate at describing cost by strenght. The green curve is the ratio of power/cost, and I've set blue and red line to same scale.
Low levels units are really the best (especially lvl 0... sadly it doesn't exist  ) but the difference between lvl 10 and lvl 30 is a bit under 10% of power/cost.
This makes the game far more balanced that what I initially thought, since quality army is easier to move... that's a true dilemma 
The simplest rule to understand is that quality woulld equal cost if the wage paid by a unit was thewage actually paid by a unit one level under. (What I mean is that if a unit of level N received the wage of a N-1 unit, quality and cost would be directly proportionnal : we always pay "one level too much"

So would a bunch of tier 2-3 guys be the best combination, or would tier 3-4 be optimal?

In terms of pure "efficiency in autocalcbattle"/upkeep, the lower tier the better.
But higher tier units are easier to move from castle to castle and are better in player-fought battle, and depending on your situation, having some high tier might be easier than getting thousand of cheap guys.
But if you want to be sure the AI won't attack your castle/city, just put 5000 peasants.

This is an old Warband trick but it's nice to see some graphs of it lol.