Wicc
Auto-translated
Sending a promising noob on week 5 with part of the army to a corner for leveling up isn't a negative move. Because the main hero doesn't lose momentum from this, and continues to conquer territory. The location for leveling up can be chosen randomly; experience is available in all unexplored areas.
I don't feel like searching for screenshots; just believe that this is done quite often.
However, sending the main hero on week 1 or 2 to various questionable locations can be a negative move. Especially if there might be an opponent there, especially if the phase of restarting after encountering the opponent hasn't passed yet, and especially if there's still something to do in the explored territory.
You still haven't tried to solve the problem, and that's a shame. Because it clearly demonstrates the principle of analyzing such situations.
Let's take an exaggerated example: consider the situation of capturing a "schoolboy" with armor. Suppose there are 100 players in total, and 10 of them are cheaters. That is, a random player, without any other information about him, is honest with a probability of 90%. Let's assume that the probability of an honest player capturing this "schoolboy" is 20% (there are fools). And for a cheater, it's 90% (not 100%, because there are smart ones). I'm taking these numbers out of thin air; they can be refined later, but the principle is what matters.
That is, if a player captures the "schoolboy," the probability that he is honest is (90*0.2) / (10*0.9 + 90*0.2) = ~67%.
That is, one episode is not enough to draw a sufficient conclusion, but based on a sufficient number of episodes, the estimated probability can be brought to extreme values.
At the same time, it is necessary to distinguish between suspicious and obvious moments. For example, if the probabilities are distributed as 90/10 + 99/1, then this factor practically does not affect the final assessment; it can't even be called suspicious, and it can be safely ignored.
Something like 50/50 + 95/5 in a single case gives too small an increase in probability to be taken seriously. In a superposition of such episodes, some noticeable probability may accumulate, but again, it will not be enough to reach extreme values. That is, in order for a set of such episodes to gain weight, there must be a very large number of them; one game is not enough.
However, the aforementioned 20/80 + 90/10 has a decent weight in itself and allows it to attract attention. And a combination of 3-4 such episodes gathers a critical mass, sufficient to make an accusation and even pass a verdict.
In essence, this is the core of my claims: the arguments you provide for the first game have too low a percentage to make an accusation, and they don't add up to the necessary weight. You need to see several such polished games that deviate from the theory of probability and the skill of the person in other components.
In the second game, however, everything is much more serious; the probabilities of an honest player and a cheater making such decisions are at different ends of the probability scale, so the critical mass is accumulated rapidly.
