Game of Stakes 5: Moving Beyond Centralization

TL;DR: Having participated since block 0 and accrued 0.86% of network voting power and 1,307,274 total bonded tokens, Stake Capital will not be joining Game of Stakes 4, instead opting to participate in Game of Stakes 5.

Game of Stakes has been plagued by a centralized cartel that has now grown to the point of holding 53% of the network voting power. One single entity controlling a majority of the voting power poses a serious threat to the integrity of the network, thus it is in the best interest of all validators to fork out the cartel members.

The Evidence

A number of validators in the Cosmos community have conducted extensive analysis into the cartel. Clear patterns arose from cartel validators that repeatedly suggested centralized control. These included:

  • All cartel validators launched at the same time for all three network starts.
  • All cartel genesis transactions came from brand new GitHub accounts.
  • All cartel validators halted during the transaction spam attacks — early on in Game of Stakes 3.
  • All cartel validators exhibited the same patterns of withdrawal/delegation (auto-delegation script updates occurred at the same time across all of their validators).

To go deeper into the evidence, see the work from our friends Certus One and Joe Pindar here:

https://medium.com/certus-one/uncovering-a-game-of-stakes-cartel-f895d9591da1
https://medium.com/@joepindar_22415/identify-potential-cartel-relay-network-in-cosmos-gos3-6c995d69ce06

Credit: Joe Pindar — See his findings here
Credit: Leopold Schabel — See his findings here

Game of Stakes 5

As was evidenced by the network halt (when the cartel validators couldn’t handle the spam attacks), a single entity controlling more than 33% of the voting power compromises the integrity of our network. While it has been a fantastic experiment and enabled us to test the strength of our architecture (e.g. against the transaction spam attacks) and see how a cartel behaves and affects the stability of the network, it is time to move forward with our experimentation. To create a stronger BFT network and thoroughly test Cosmos out in the wild, we must fork out the cartel and move on to Game of Stakes 5: a fully decentralized network.

Certus One have created the Game of Stakes 5 genesis file here and Cosmos validators are actively showing support for the fork here. The launch of Game of Stakes 5 has been set to Saturday, February 2 at 15:00 UTC, giving validators some time to learn about the cartel and examine the evidence.

Why Should Validators Join?

By not validating on GoS 4 and preventing the network from starting (< 66% voting power), we provide time for Zaki to review the pull request and community members to examine the evidence of the cartel.

If, on the other hand, validators join GoS 4 and the network does start (>= 66% voting power), it may be very difficult to get validators to switch over to GoS 5. And, since the cartel cannot be forked out using on-chain governance proposals (as only 33% of the voting power is required to veto proposals), the cartel will remain in GoS 4 and the network will remain vulnerable.

The validators in red represent cartel members. As you can see on the right, most non-cartel members will benefit from a higher ranking in Game of Stakes 5 (once the cartel is removed).

As long as at least 33% of validators refused to participate in Game of Stakes 4 on Friday, it will not begin and Game of Stakes 5 will start on Saturday.

Zaki has stated his support for Game of Stakes 5:

Here is my take right now.
I am super pleased with how this is going down.
It’s been a long day and I don’t have the energy to fully review the created genesis file so I don’t think I should merge it until I’ve reviewed it.
I would be very impressed if GoS 5 goes live.
I think the argument laid out by Certus One in Favor of the GoS 5 token distribution is very persuasive and voting with your validator to start GoS 5 if the token distribution agrees with you, I support starting on GoS 5

Big thank you to all validators who contributed to the research, in particular Leopold Schabel and Hendrik Hofstadt.