They need to create a universal death algorithm that, has realistic randoms, inclusive of the player battle and all battle, the type of battle, the age, the medicine factors, outnumbered battle, yes or no, how many lives of the cat? For each lord and noble. Were they wounded recently? Did they come into battle wounded? Many things.
This is simply one long calculation, based on +1, -1, +.5, to then impact the random calculator.
This then applies to all real-time and simulations. Then they need to run a total game simulation, and ensure over the total, the sandbox remains alive, and clans still functioning, and only in rare events, do clans get wiped out, but it should happen on occasion.
--> This requires about twenty or so factors, to contribute to a Noble receiving a greater chance of death, or less.
The current state of the death in player only battle, I would classify as very bad design idea, or a bug. That makes the player a death bringer, and of course, the lop sided clans.
Example: When a player is battling 500* vs 50, death should be .15 percent chance, yet still a chance. This chance should then increase to .45, if a character comes into battle 90% wounded.
In the reverse situation, 50* vs 500, the death chance, is .25 percent chance. This chance should then increase or decrease, per co-factor.
(my calculations might be out of line here in these percentages, though it is not the point of accuracy, only a basework programming of how the function should work -- the chance of each battle, should fit within the greater wheel of the death target in battle for the entire map)
This is all fitting within a reverse engineer goal macro-total goal, of the total average death rate of all Kingdoms. First create the target rate on average for the total worldwide map of <death in battle>. Then reverse to find the micro of those instances with the inducing factors, so that it simulates the average, of long-time sandbox play (death in battle per hours of real clocked gameplay time), the extremes coming into play, but not affecting ten individual game distinct average.
The death calculation on the whole, needs to be redone, and reworked. It ought to be not too realistic, but include the fun factor of play. If there is to be a custom death slider, that can bring total realism, this could cause most clans to be wiped out, as this is a game of war, and not total reality, where war and battle is always happening.
So in this case of the game, fun factor, I would create a mean, or a average death rate goal, for all clans, to die in battle. This is the goal.
This would be, for example, 1 Noble die in battle per 3 month [3 months is only my rough balance assumption] per kingdom, on average.
This would allow the sandbox to continue. This average can be taken to some extremes, based on unlikely coinciding factors.
These factors might superficially be, above age 45 and in battle, bad medicine skill in the party, then you might at the extreme have two Nobles die per 3 months in the Kingdom via battle, and no more than 3 at most, from differing clans.
This algorithm must work through total simulation of all activity in the map, so that about 1 Noble is dying per time, as per a realism in battle, yet not randomly chosen, it must be as per the many factors that allow it. It could also be the inverse, that no nobles die per a length of time because of those factors. Though, these extreme factors, must be come within a ten game average, for ten distinct game simulated results.
Then, the same code, should apply to the real-time battle, and the real-time battle should have nothing to do with the death rate.
So the beginning, is to create target, of How many nobles do we want to die in battle on average when we totally simulate the game? How does this work with the longevity of the sandbox, and this should also include the possibility of weaker clans completely getting wiped out.
It is supposed to inflect a realism of warfare, including death in battle, yet also perseverance of the sandbox.