But... When i try to reverse the last (let's say) month: I'd have to go back to (x-4464) => 6 blocks/hour * 24 hours/day * 31 days/month.
Your example is correct with the assumption that the attacker would start their 51% from block x-4464 when the main chain was already at block x. A 51% attacker could start their 51% attack from block x-4464 when the main chain was still at block x-4464, and simply not broadcast their blocks. When the main chain reaches block x, the attacker could broadcast blocks up to x+182 and rewrite an entire month in one go, without having to play "catch up" to the honest chain.You are correct, i didn't think about 51% attacks in a while, and i kind of forgot there are several ways of looking at this problem
I do still maintain that the numbers generated by crypto-5-1-.app should not be taken at face value without knowing the exact methodology... Adding numbers on a site and telling people you calculated them based on the price of nicehash's hashrate just doesn't cut it for me... The numbers might be ballpark correct, but without seeing the actual calculations, they're just numbers...