Carl, please stop testing by just adding the shield value to the defense skill field. Every single unit in the game will perform better with this done, provided it has any shield points, so it proves absolutely nothing that a unit modified like this does better in combat.

I would also like to point out that my testing of armoured sergeants vs. mailed knights already helps neutralize one of the most random factors - the charge. I've noticed especially that spear units charge very oddly. Against cav, mostly they stop to group up before charging, but are overrun before getting up to speed and brandishing weapons. That's with a single-click charge. With a double-click charge they are considerably better about getting shield up and weapon out at the right time, but again still sometimes charge in without having weapons and shield set (i.e. fail to have a charge trigger), only to be completely slaughtered. The easiest fix was to not issue a single command to the unit during combat. I did absolutely nothing except set the unit to run speed while it was sitting still in order to chase the knights better when they try to recharge, which should make the situation vary not at all from run to run as I issue no commands. The end result is that my runs for each stat set looked practically identical, down to when the knights would pull out, where they would go to, how the spearmen chased after them, how many of each unit were lost on initial contact... pretty much every way imaginable. The 1 exception from that first post is the first entry under 5/9/1, the 26/32 entry. This one I recall in particular, because it went a little differently. The knight general was killed on charge impact, causing his unit to rout early. Even so, the results appear to be on their way to the typical kill ratio for that stat level.

That leads to another factor that can be removed, though: the morale factor. IIRC, the EDU lists a possible morale factor that locks morale in place so the unit is unable to rout. This seems very promising as it would force combat to the last man instead of allowing the early endings we sometimes get currently.

As an additional note, switching sides in a battle is not a good way to go about things. The idea in testing is to isolate the effect of stat changes to only one of the units in the combat. This means you must do everything identical in your runs, only changing unit stats for ONE of the units for some of the runs in order to gauge what effect that change has on the combat. The idea is that the AI should be forced as much as possible to do the same things in combat, to make the stat change the only possible reason for any difference in combat results.

I think later on (when I'm at home) I'll run some tests using 2 identical units like the DFK and DNK that Revenant mentions in post #18. I'll be doing the following:

1. Units with vanilla stats except: charge set to 0, morale locked if possible to implement. With charge 0, charging should have no impact, which means I can safely leave the unit in my control sit without adversely affecting it... which is exactly what I intend to do.

2. One of the units will have shield value added to skill (+6 skill pts I'm guessing), the other will use 0 shield (probably this is -6 shield pts), so I'm testing to see if skill points equal to shield has the same effect on a unit that removing its shield does. A result of dead even battles will illustrate that positive skill points directly equal negative shield points, i.e. that shield points operate exactly in equal magnitude but the opposite direction than they should. I hypothesize this will be the case, with let's say less than 10 of either group ever surviving the combat.

3. I'll then run a control test to make sure the units are in fact even in vanilla. This should be more important than just noting how even their combat is with #2 stats, as the vanilla stats have to provide results similar to those with #2 stats applied to show the correct relationship between positive skill points and negative shield points.

4. All tests will be run with default custom battle settings. I select grassy plain and hit go, select the 2 factions, pick each unit, and begin battle. I have been doing this already whenever I test, but wanted to be clear what my practices are.