Proposal #14 passed on January 1, 2022. That night we tweeted statistics on the jailed counts and cumulative share changes to which we got a few responses. Jacob Gadikian replied, "Wait tho Jailed is meaningless And I expected them to be jailed and come back Because they’re noobs so you’re hypothesis is based on the flawed assumption that uptime matters." To be clear, our hypothesis was that many validators "weren't ready to become active" and so we agree with Mr. Gadikian and expected them to be jailed as well. We also publicly stated, as acknowledged by Proposal #14 supporters, that jailing wouldn't produce any noticeable downside. However, 14 validators being jailed within the first 24 hours self-evidently supports the hypothesis that they were not fully prepared to activate.
Some specific claims about proposal #14 will elucidate why we're focusing on these metrics. We're focusing on downtime and delegations because the claim was made by Proposal #14 supporters that validators "are waiting for expansion to make their move for competition" and would increase delegations to activate. StakeLab, the proposal originator, specifically stated "they will redelegate their funds to their validator when in the active set." Since these are claims are verifiable, we think it's useful for the community to know whether the proposal had the intended effects.
The goal of the proposal was to increase decentralization of the network. We tracked the validator cumulative share for 8 days prior to the proposal and 8 days after the proposal to compare its effects. To us, meaningful decentralization would mean shifting the benchmark of the 2/3 cumulative share, which is what determines the prevote. Graph #1 displays the cumulative share controlled by the top 5, top 10, top 25, and top 50 validators. While less important than the prevote level, it is useful to more holistically observe decentralization, especially when evaluating lower timeframes such as our 16 day period. The thick black line in the center of the graph is representative of the proposal taking effect.
As is clear, the cumulative share did not change in any significant or noticeable way. However, there were a couple issues with our dataset that deserve recognition. Acheron, ranked 31, was jailed at the beginning of our study and then unjailed on January 5. We removed them from our dataset until January 8 to isolate the effects of the proposal on the dataset we had from before the proposal had passed. We reintroduced Acheron on the last day of data gathering to identify the true current distribution of validator cumulative share. It did not impact the dataset in any noticeable way. GStaking, ranked 41, was jailed on January 6. We continued to include them in the dataset until January 8 as they were active during the pre-proposal data collection period. We removed GStaking from the dataset on January 8 to identify the true current distribution of validator cumulative share. It did not impact the dataset in any noticeable way.
Ultimately, Proposal #14 has not proportionally increased network decentralization, as is evidenced by the ultra-mildly strengthening share of the top 10/25 and the stable share of the top 50. It is probable that in the long-run this will contribute to decentralization as opposed to an 85 validator set. However, the question over whether this was the right time to expand the set remains. The negatives to Proposal #14 are that the larger set is less efficient than the smaller set and increasing the set may fracture delegations to smaller validators. The positives are that many new validators entered the Akash ecosystem and can experiment with deploying infrastructure, increasing activity and participation within Akash.