According to CivIII Combat Calculator these are the percentages.
Beserker Army defence = 2 + (3*2/6 = 1) + terrain bonus 10% = 3.3
Knight Attack = 5
the Attacker wins 9.36% of the time
So yes the
Knight was lucky, but not that lucky.
Combat in Civ can be wildly unpredictable which is why some players prefer to be builders because there are few random elements involved.
Lucky = Modern Armor (24) vs Spearman (2) + terrain bonus = 2.2
Defending Spear wins 1.359% of the time.
Note: I don't use the Combat Calculator as a rule. I downloaded it just to see what the percentages were. At first glance it did seem unlikely a Knight could defeat an Army.
In addition there was a 36.27% chance that you would take more than 2 points of damage in the other example you quoted (Army attacking Knight)
The problem with
beserks is the low defense, Carthage has a similar problem in Rise of Rome. Heavy Cavalry also has a low defence value so they are similarly dis-advantaged.
As a side note Blaze has lost two Heavy Cavalry armies in a Rise of Rome scenario we are playing, and did not react in the way you did.
The lesson is check your facts before making a statement such as
I will consider this another thing almost imposible.
almost 1 time in 10 this can happen. Of course against the AI you would never have got attacked as the AI likes a 50% or higher chance of success.
In another PBEM I had two 4 unit armies (Riders) attack MI in consecutive attacks and though victorious they suffered 4 (8.87%*) and 5 (3.37%*) points of damage respectively. In the same game I also got 3 great military leaders in 4-5 turns

but can't tell you what the chances of that were. Did my opponent do anything but comment that's a lot of leaders. Especially when I got 2 in the course of 3 combats.
* % chance of suffering this much damage or higher
As a further sidenote there used to be a board game journal called 'The General' which would feature a replay of one of their board games, part of this replay would show how great a degree luck had played in the outcome. If you are prone to be suspicious of other players good luck perhaps you would like to collect the results of combats in a spreadsheet.
This is why Battle Reports should be accurately posted, so that players can be less suspicious of the results of battles. PBEM requires the players trust one another to some degree and can take some comfort in analysis of the numbers should they become suspicious.
I hope this helps clarify things.
Wow this post became much longer than I anticipated
I truly am a gregarious GreyGamer
