Battle engine: Peasants vs Armoured sergeants

[R'as al Ghul]

Hello,

I've made some tests on Peasants. The idea is to understand the combat model/battle engine better. We're all aware of the fact that Peasants can be a dangerous unit. They're able to beat Armored sergeants in melee or at least inflict enormous casualties. Peasants have an attack of 4 (charge 0) while Armoured Sergeants have an armour value of 5 (plus a shield of which we don't know if it's used in melee). Because of the armour value of the AS, the Peasant shouldn't be able to inflict much casualties. yet it does. Why is that?
My test setup:
Grassy plains, fine weather, medium difficulty, unit size=75.
Human controls 5 units of AS, AI controls 5 units of Peasants.
Human issues no orders but clicks "Start battle".
All tests play out like this:

Human and AI are default deployed in a line of 4 units, Gen behind the lines not engaging. The Peasants march up to the AS and charge them. AS just receive the charge and defend. After a while the attacking Peasants will route. When a (Peasant-) unit displays Routing/broken the game is paused and the remaining number of soldiers is noted. I unpause to wait for the next unit to rout. When all 4 line units are routing the battle is aborted and the number of casualties for each side is noted. The general units do not engage in fighing for the whole time.

Results of 3 tests:


This is just to display a trend. I've more results but am reluctant to copy them all manually. Note that of the 300 men engaged in the 4 units the AS lose between a third and half of their soldiers.

Now, I've edited the Peasants to have 0 combat values (0 charge, 0 attack, 0 defend skill). AS remain as they are, same setup as above. Only noteable difference: The Peasants stop after marching at a distance and wait for a while, then they walk up to my AS and start fighting.
3 exemplary results:


So, the Pesants kill a lot of spearman although they don't have any combat points. With 0 attack they shouldn't be able to kill at all inflict that much casualties. Of 300 AS at least a third are killed.
If somebody has an explanation, I'd be happy. Right now, I'm as clueless as this guy:



I agree that the number of samples is low but I've done more tests than I've documented here and the results displayed can be taken as exemplary. BTW, the tests are easy to setup and repeat for yourselves.
The results leve me with the question how the combat really works/ is calculated. Obviously the combat values aren't of high importance!? Shouldn't a difference of 4 attack points less result in less inflicted casualties for the AS?