Decentralization in Proof-of-Stake is a multifaceted topic. Many theories and designs have been suggested in the wider ecosystem on how a PoS system may decentralize power over multiple entities. The Cosmos Hub is one of few of these experiments that has been live for a couple of months. The following analysis covers stake distribution and diversification on the Cosmos Hub and aims to shine some light on how token holders have actually engaged with this early PoS network.

In a PoS network, the way token holders decide to participate in staking influences power distribution in the network. Do holders stake tokens themselves, do they pick one validator and stick with it, or do they diversify across validators? Diversifying stake should come with benefits: it lowers the impact a slashing would have on the token holder and helps to support different entities, thus strengthening decentralization and the ecosystem in general.

The Data

To find out how Atom token holders are participating in staking, I took a snapshot of data from the 5th September using the Stargazer API (thanks to Certus One for providing this API). The data takes into account all active delegations, as well as their size and the corresponding delegator and validator addresses. It should be noted that addresses don’t equal entities, there could be holders that hold their funds in different addresses, or holders that share addresses (exchanges/custodians), which will not be reflected in the following analysis.

Let’s start with a summary of the snapshot data. In total, at the time of the snapshot, 177,771,881 Atoms were being staked. The snapshot contains 11,387 active delegations from 6,459 addresses on the Cosmos Hub, which means the average address has 1.76 active delegations. From that we can already see that a large number of holders don’t seem to diversify.

Exploring Stake Distribution

Before we go more in-depth regarding stake diversification, let’s first consider the stake distribution. The following graph shows Atoms at stake with validators ordered by stake and colored according to the commission rate bracket that the validator belongs to (5% or lower, between 5-10%, and greater than 10%).

Stake distribution among the 100 active validators on the Cosmos Hub. Gini Coefficient = 0.7366. (Date: 05 Sep 19)

Looking at this we can see that the largest validators either charge a low commission rate below 5%, while others charge above 10%. These often signify funds that validate their own stake, or providers that focus on larger holders. Looking at the validator set ranked by stake visualizing the amount of addresses that delegate stake to them, we can see that low fee validators are popular especially among smaller holders. The top 3 validators by number of delegations charge no or a low fee:

Number of delegators among the top 100 active validators on the Cosmos Hub. (Date: 05 Sep 19)

This graph also contains some interesting outliers: e.g. Huobi Wallet on rank 43 has delegation from 636 addresses. Similarly, Coinone Node on rank 15 has 745 delegators. Both of these are exchange run validators. Another interesting mention is Cosmostation at rank 7, who managed to attract 678 delegators to their validator with a 12% fee, potentially a result of their successful wallet.

Exploring Stake Diversification

Now that we got an overview of stake distribution among validators, let’s take a closer look at delegations and the addresses they are coming from:

Summary statistics of the Cosmos Hub stake distribution data set. (Date: 05 Sep 19)

This data contains some interesting insights. For one, only about a quarter of addresses actually diversify their stake, but together these addresses account for about 64% of the total stake. What this means is that addresses that diversify on average hold about 5 times as many Atoms as those that don’t, providing us with insights that diversification is more popular with larger holders.

Taking a closer and counting the number of addresses given how many active delegations they have (n), we see a swiftly declining curve. While there are still 786 addresses with 2 delegations and 347 with 3; addresses with more than 5 delegations become very rare. There are a few interesting outliers that were cut off this graph. Specifically, an address maintained by B-Harvest sports 119 delegations with low Atom amounts that were carried for tracking purposes. 6 other addresses with seemingly similar patterns (<500 Atoms across 67 to 83 validators) exist.

Counting addresses based on the number of validators that they delegate to. X = Number of Delegations, Y = Amount of Addresses. (Date: 05 Sep 19)

Finally, if we visualize the Atom amount staked by addresses sorted by the number of validators they diversify amongst, we are presented with some more data points that stand out:

Atom holdings of stakers grouped by the number of validators that they are delegated to. X = Number of Delegations, Y = Amount in Atoms. (Date: 05 Sep 19)

The peak of holdings with 36 delegations is from the Tendermint team address, which stakes ~21 million Atoms. Another peak with 46 delegations and  ~7 million Atoms at stake is due to the ICF address. The most diversified address with significant holdings delegates to 51 validators. The two peaks with 16 and 18 delegations also signify well-diversified large holders.

Conclusion

In conclusion the data presented clearly indicates that Atom holders largely don’t diversify across multiple validators. In addition, we also saw a preference for low fees among smaller holders.

Our team at Chorus One believes that stake diversification is important to create a thriving validator set that engages with the network and contributes to ecosystem growth. It might be that currently token holders don’t care about diversification, are unaware of benefits, or that it is simply too hard, or not worth the effort to diversify (low slashing risks, diversifying requires increased research on validators, bad UX of having to carry out multiple transactions to delegate and to withdraw rewards,...).

We are researching ways to make participating in staking and diversifying easier. Join our Telegram to discuss and let me (@FelixLts) know what else you’d be interested in for future analysis. Interesting extensions could be to expand on differences between whales and small holders or to take a look at other Proof-of-Stake networks (e.g. Tezos). If you have a dataset for other PoS networks or want to help obtaining the relevant data please reach out.