Hello, I didnt get why you need only 33% to shutdown the blockchain.
Thanks for asking. I realize that math result is non-intuitive and difficult for a layman to believe.
See the “Math of liveness & safety” section of my blog. Study the explanation of the math that is quoted from my white paper.
It is because over the overlap of 2 elections for the relative pairing of two blocks, and the fact that there are N delegates who vote yet the quorum size must be T. We then solve for a mathematical relationship between T and N by equating the two expressions, because we have another requirement that the excess of N - T be equal to the aforementioned overlap 2T - N - 1.
The point is that it is the need for the ability to objectively prove the relationship between any pairings of block elections in the case that everything is relative because all actors are untrusted and the network propagation order is indeterminant (which relates back to the oft-cited FLP impossibility theorem which is mentioned in more detail in my yet unpublished white paper). This is why mathematically 33% can halt the blockchain, if we care about there existing an objective consensus (i.e. safety and consistency) and not just an unbounded number of subjective competing forks (i.e. double-spends and chaos). Of course a boastful, overeager FOMO-bag seller like Dan would claim this is incredulous, because Dan is apparently ignorant of the math and does not have his priorities on being expert on the research. I find it incredulous that Block.one raised ~$500 million thus far and apparently can’t even be bothered to hire some researchers to teach them about their technological weaknesses.
It is all in the math. I know it is non-intuitive but this result is from the 1980s (although the way I have derived it and explained the well known result may be my own unique insight). Sheesh and Dan is not even aware of it.
Realize the 33% liveness threshold (where 33% can shutdown the forward progress of the blockchain) only applies for an elected set of delegates and not to proof-of-work.