Draw a map of points that need to be corrupted to compromise a network and/or cause harm and I assure you ETH has fewer points of corruption. Namely, I would corrupt the operators of the two largest pools and get a 51% attack.
ETH is delegated proof of work with 7 meaningfully elected positions with 2 controlling 51%
ETH recommends 12 confirmations, in a window of 12 blocks 7 will be produced by 2 people and the other 5 will be produced by 5 people for an average of 7 unique individuals confirming a transaction.
The economics of block production on ETH falls off dramatically once the number of blocks produced per day falls below the cost of operating a node. EOS and STEEM both have the same trail off on the long-tail which means that economically speaking, the total number of block producers will be the same in both systems. The difference is that DPOS makes sure the top 21 are all unique and equally weighted in each confirmation window (45 seconds) vs unequally weighted in each ETH confirmation window (3 minutes).
So do the math, their claims of decentralization and/or centralization are entirely bogus and unfounded.
RE: Response to Vitalik Buterin on EOS