Since somone recently mentioned elsewhere that the 90% rule isn't really described in the wiki, here is my own writeup of all the knowledge I could gather about it in the forums. I think it - or a section like it, should go into the wiki:
The 90%-rule
Warbarons restricts the impact of luck through its "90%-rule".
Each battle is simulated a thousand times, and the number of occurrences of each outcome is noted.
The battle is then performed one more time, to determine the outcome that counts.
However, if that rolled outcome was within the left or right 10%-fringe of all outcomes, then
the entire battle is rerolled until the outcome is within the middle 80%.
This eliminates extreme results, such as winning without losses, or losing against very weak opponents.
The 90%-rule does not apply to ruin exploration.
The consequence of the 90%-rule is, that if the battle dialog of the game displays a win chance of over 90%,
then the attacker will definitely win with at least one unit surviving.
Similarly, the attacker will definitely lose all units if the game comes up with a win percentage below 10%.
Warbarons never displays chances that account for the 90%-rule (it is nowhere mentioned within the game).
Since the application of the 90%-rule depends on random-based simulation, any outcome remains possible.
A single weak Scout may still kill 8 strong Dragons,
but the 90%-rule makes the chance for this to happen very slim indeed:
the Scout must win at least 100 of the 1000 battle simulations,
and then he must also win the last battle that counts.
This is extremely unlikely, but not impossible.
The 90%-rule may sometimes have weird consequences,
such as smaller stacks having better individual survival chances:
e.g. a single spider has a better individual survival chance
attacking an elf in a city with no wall,
than a spider and a crow attacking together in this order.
The outcome percentage value that is also displayed in the battle dialog of the game
may already give away the result of the battle:
it is the probability of the survivor pattern that this particular battle will have.
Note that this outcome percentage may well be below 10%, despite the 90%-rule,
since it might be a rare outcome in the middle of all possible outcomes
(e.g. a stack having a weak unit wedged in between strong units, with the weak unit reaching the front line
and surviving).
Note that the winning chance shown by the game likely differs
if you repeat a battle with identical units,
due to the random chances within the simulations that were used to determine the win chance within the game.
In fact, the game may produce any winning percentage, as the scout-vs-dragon example above shows.
I have more text like this, e.g. about bonus mechanics or naval battles - things that I believe are missing in the wiki. I had already contacted piranha about it through the forum PM, but I guess I should have used the Warbarons PM instead, since I got no reply. So let me know if you can use my text and if you want more.