So the units have % values, like for initiative, or for the success of a thief.
But even with the random factor of % itself, and the additional random they implemented (for each round, afaik), it does not really explain how some things can happen.
Like, i take 5 thieves, each with 30% chance to poison the enemy (in) city. As they are already standing next to it, each thief has two tries.
So i try 10 times to poison the units there, and each fails. I loaded (saved right before, as 30% is pretty low anyways), try it again. Each fails. Again..each attemt failed.
So it was like 20-30 attempts, at 30% chance. Statistically, about 1/3 should work. 3/10, 6/20 or 9/30. While it is possible that i just had bad luck with the first attempts.. it should really work with 30 attempts.
I have to say, i was not surprised, even 80% feels more like 40% often..
Another thing: Initiative.
So i have one guy with initiative 50, and the enemy has one. In a battle of 8 or more rounds, he was first, each single turn.
Or i got a guy with 55 initiative (used spell/item to raise it), and the enemies (with 50) still gets first move, every single round.
And today i got one with 44 initiative, and although he was somehow faster then one of my "50" guys, each round, once again, he was never faster then my other "50" guy, and never faster then the enemies "50" guy.
So it looks pretty much like the game would give a **** about the value.
Why should i buff my army with + initiative when the enemy strikes always first?
And why do they even bother with displaying % chance when the actual chance is either way lower or outright zero?
Thats really a thing thats bugging me about this game.

For the thieves thing, it may just have been that you got bad luck. Also, the game might use a random number sheet like Fire Emblem, where it generates a ton of random numbers when the level starts, and draws on them in order whenever it needs one. That way, the outcome would be the same no matter how many times you reloaded your save.
Then again, it's possible that the % values are just lies. I have heard of one game that does this, but it was of a much lower standard than Disciples.

Well, the random number set would be a possibility. But as i tried at least 10 times, that would mean very bad numbers, all right after another, and then the % would be some kind of lie, too.
Random effects should determine which of two equals goes first, but not like it works in Disciples if there is such a thing.
If they do: Why % values when the random numbers make a greater impact? As they do, then. Otherwise it was really really bad luck, as without it matters if i load or not..
Also i wonder if the enemy can ever run out of ressources? Well, could be that he saved that much, as he has like 100% of all ressource locations right from start, but they cast that much, even stupid stuff, like casting Summon I/II, and then end the turn.
And he proceeded with it when i had everything except his capital.. makes me wonder if you can "siege" him until he runs out of ressources.

The AI definitively cheats. The manner and degree probably varies with difficulty, but castle upgrades, resources, unit upgrades, heroes, unit & hero levels, and researched spells are all things where the AI obviously cheats.
The percentage values though, I think outcomes are based on a unique, random number. At least, I think I've tried re-attempting just about anything with a percentile involved and gotten different outcomes.
I think it is more likely that you either had a run of really bad luck, or perhaps that a higher difficulty setting penalised your stats without telling you about it..

Well, then the values displayed are for the easiest difficulty.
At least i think i tried first at hard, then changed to normal after the first level, as it was really hard, like saving before and after each turn, otherwise the AI will first demolish my troops with some badass spell and then appear out of nowhere.
Wonder how it is on the hardest difficulty.

It is so that the thief's action's rate of success wil not change within the same turn, what you might try doing if they all fail is skipping one turn, or just face the facts - you had bad luck.
With initiative there is some variation, at times a unit with less initiative can act before a unit with slightly higher initiative (eg. a unit with 40 initiative can be faster than a unit with 45 initiative, but never faster than one with 50).
However this should be a rare thing to happen, and certainly not every round.
If both have the same initiative, I guess it's a random chance for either one to be first.
And saying the computer cheats is easy, it just feels that way but I have played so much Disciples that I don't think that's true.

Do you know that for a fact, or are you speculating? That's true for just about all variables in the games. It is indeed very easy. Because it demonstrably does cheat. A great deal. Whether it cheats on the random outcomes & creature/item stats, I don't know. It seems like creature stats are nerfed a bit on hard difficulty, but I suppose I could simply be biased or have had a couple of years worth of below average luck.
But it very obviously cheats with almost everything else. If you don't believe me, find a couple of cheats to dispel the fog of war and get infinite movement. Then make a thief and go check out the enemy/enemies.
That said, there's nothing wrong with cheating AIs. The best AI is a retard compared to a human player with a bit of practise, and even if that wasn't so, never being disadvantaged would get boring fast.

I wouldn't call that cheating, and my statement was within the scope of a battle, where the computer keeps finding those lvl2 stacks is out of my knowledge.
I once read an article by a big game designer and he put it this way: the player has the brain, the computer gets the cash.
So I'm guessing the PC has unlimited cash and can make wonderfull use of trainers in certain missions.
I however never cheat in my games - and did manage to bring very hard PC's on their knees, it's not so much about the huge quantities of armies the computer possesses rather than choosing the right battle at the right time. (Computer makes huge mistakes on that part.)

I always assumed that initiative was +\- 5. IE a 50 ini guy could in fact go after a 41 ini guy, if they got -5 and + 5 respectively.