It’s pretty common these days to hear the critique that Delegated Proof of Stake (DPoS) protocols are too centralized with respect to block creation. In today’s post I’d like to push back on that a bit — some DPoS protocols actually have extremely decentralized block creation.
As with many concepts in computer science, it’s all about tradeoffs, and this applies to different DPoS systems as well. Here’s a simple diagram to illustrate the biggest tradeoff for DPoS:
As you can see, the more delegates you have, the slower your network will be, and the fewer delegates you have, the faster your network will be.
I think the “all DPoS systems are centralized” critique mostly stems from thinking about EOS, which only has 21 delegates called Block Producers. EOS Block Producers will likely resemble corporate data centers with massive computing power — so the network will be able to process thousands of transactions per second — and there is an argument to be made that the network will be less censorship resistant because of this.
But one DPoS protocol that should be quite decentralized in terms of its block creation is Tezos. In Tezos, users with at least 10,000 XTZ can “bake” (that’s the Tezos term for stake) in the network. For users with fewer than 10,000 XTZ, they can delegate their coins to a baking pool which will allow them to create blocks too. With a total supply of ~760 million XTZ, the Tezos network will be able to support up to nearly 80,000 delegates at launch, and although the network surely won’t have that many bakers when it’s live, it will probably closely resemble the decentralization of the Ethereum Casper ecosystem, which has individual stakers plus staking pools like 1protocol. This bodes well for Tezos’ ability to be censorship resistant, however, it comes at a cost to throughput — the network will only be able to support ~40 transactions per second.