Parameters and decentralization: the way ahead

Shelley continues to evolve and, as it does, IOHK is committed to supporting the healthy development of the Cardano ecosystem. There are many factors at play in this regard: everything from the wallet experience for ada delegators to network tools for stake pool operators; from governance and funding frameworks like Project Catalyst and Voltaire to community-led initiatives. Central to the health of Cardano as a vibrant proof-of-stake network is the manner in which delegators and stake operators are rewarded for running it, and how those rewards are shared to maximize decentralization and secure the network.

The system is still transitioning towards full decentralization and community ownership. Very shortly, the majority of blocks will be minted by community pools as the influence of the genesis nodes supporting the network is gradually reduced. The steady reduction of d (the parameter governing this process) will see it reach zero by the end of March 2021. This will mark the full decentralization of all block production on Cardano.

Some four months on from the deployment of Shelley, it is time to adjust the parameters of the rewards mechanism so that the next phase of the network’s growth is fueled. Since launch, we have been monitoring network behavior, while seeing how the community and ecosystem are developing. We have also been carefully modeling various scenarios based on data as it emerges. We have now finished a full review and have aligned – from a philosophical, scientific, technical and practical perspective – the next steps. Before we make further adjustments to the parameters of the rewards scheme, we wanted to share with the community what we hope to achieve with the underlying mechanism and associated parameter adjustments.

Encouraging sustainability through pool diversity

First and foremost, the rewards-sharing mechanism of Cardano aims to reward people fairly for supporting the platform in a sustainable way, not as a short-term windfall. One of the goals of staking has always been to encourage long-term holders of ada: the people who are committed to the success of the ecosystem.

In the longer term, the system can be successful only if it is widely decentralized. Philosophically, the design of the rewards scheme aims to encourage a wide and diverse set of stake pool operators. This secures the platform against attacks, spreads any available rewards evenly across the community, and makes the system more resilient to change.

Many economic systems have a tendency towards consolidation and a smaller number of strong players. On the other hand, blockchains can only be successful when control is decentralized. The rewards-sharing scheme ensures that smaller and medium pools can contribute meaningfully to the ecosystem without becoming subsumed into larger operators and consortia, as has happened with other blockchain systems, particularly Bitcoin.

One way to discourage a trend towards a few large pools that collectively control the system is to introduce a counter-incentive to a pool's growth. The rewards-sharing scheme we designed is an example of this novel (in the cryptocurrency space) concept. As soon as stake delegation to a single pool increases above a threshold, the rewards automatically diminish, encouraging ada holders delegating to that pool to find a new home to improve their rewards. This mechanism limits the delegation that can be rationally made to any pool and spreads delegated stake more evenly across a larger number of pools.

All about k

The parameter of the rewards scheme that sets this 'soft-cap' on the pool size is called k. The mechanism is designed so that, at equilibrium, assuming rational participants and no external factors, the stakeholders' best response behavior converges to k pools of equal size delivering the same level of rewards per unit of stake to their delegates.

For the deployment of Shelley on mainnet, we started with k=150, which limits pool size to 210 million ada. This was a modest increase from the parameter choice used in the incentivized testnet (ITN), which had a value of k=100. At the time, this was considered to be a relatively conservative choice, made to ensure the ITN environment would smoothly move to the mainnet. The launch of Shelley has seen enormous interest from the community and a large number of pools. Over the past few months, we have observed the way the staking pools are operating, and recognize that k needs to be adjusted upwards.

It is worth noting that the k-parameter is not amenable to small, gradual increases (unlike, for example, the d-parameter, which lends itself to a gradual reduction). Each increase in k requires pools and delegators to take action. For pool operators, this means a careful adjustment of their parameters and in particular their margin; for delegators, it means choosing new pools to delegate their ada to, especially if their current choice becomes oversaturated.

Therefore, the best strategy for an upwards adjustment of k is to move in larger, less frequent increments – and to move it as far and as fast as practical network dynamics and economics will allow. The ‘how much’ and ‘how soon’ has been a topic of intense debate and discussion within the team, made more complex by a number of technical factors. The best solution is one that minimizes disruption for successful pools and their delegators, while maximizing the opportunity for medium and smaller pools to mint blocks and attract more stake. Equally, it is crucial to be always focused on the longer-term strategic goal as a community to broaden decentralization as widely as possible.

Moving to k=500 in December

We are committed to a measured and deliberate set of changes and will be using the data we gather to inform subsequent decisions. Therefore, we plan to implement the k change in a staged manner. First, we shall move to k=500 at epoch 234 (21:44 UTC Sunday, December 6, 2020). The move to k=500 will give small- to medium-sized pools that are struggling an improved opportunity to attract delegation. It will also have the effect of limiting pool size to 64 million ada, which means more than 100 of the largest pools will become saturated.

Ada holders can redelegate at any point between now and the change. If you are currently delegating to one of the largest pools, and wish to continue receiving optimal staking rewards, you may need to move your ada before or during epoch 233, before the new epoch boundary at 21:44 UTC on December 6. We certainly encourage delegators to keep an eye on their favorite pools in the saturation meter in the Daedalus wallet a day or two before December 6. If it shows significantly more than 64 million ada at this point, you should certainly consider redelegating. It is important to note that rewards will still be payable from slightly saturated pools, but these steadily diminish the greater the saturation over 64 million ada in the pool. To be clear, though, no one delegating to an over-saturated pool will lose any of their stake. It is simply that the return on their stake will be reduced if they stay delegated to a saturated pool. We always encourage the community to keep an eye on their delegation choices, and this will be especially important around this time.

Modeling the long-term viability of stake pools, we found that k values of 1,000 were stable in the long term. As a result, our aim is to move to k=1,000 during March 2021. We recognize the importance of economic factors that also strongly influence pool profitability and will continue to consult widely with the community on the plan; the social dynamics of the network should also not be underestimated. A number of opportunities will be presented (including community panels and seminars) for further discussion, while helping the community understand more about the changes and contribute their perspectives.

From a strategic and philosophical perspective, we believe this is the correct approach for Cardano. We wish to encourage decentralization across a large number of stake pools that are independently operated, while recognizing the positive contribution of the pools currently leading in the rankings. We shall also pay heed to the practical dynamics of the evolving ecosystem. By the end of March, d, the decentralization parameter, will also be set to zero, meaning that Cardano block production will be fully decentralized, and responsibility shared across nearly all the pools currently running the network (and hopefully, some new entrants too).

No silver bullet

We believe that changing k to 500 will benefit the ecosystem, despite a period of disruption and change. However, it is not the whole solution. We are continuing to develop our thinking in other areas that will tangibly contribute to Cardano’s decentralization. Hardware wallet delegation support (coming very soon) should help open up ada supply, to the benefit of all. We shall soon add the ability to delegate to several pools from a hardware wallet, which will help Trezor (initially) and subsequently Ledger owners spread their stake across a range of pools. Improvements to stake pool servers will in time allow community members to start curating their own pool lists to help shape and steer delegation choices. We shall also be retiring all but one of our public IOG pools and encouraging delegators to switch their ada to community pools, while developing our own delegation strategy. On the parameter side, we are now finalizing some modeling around pledging by pool operators, another factor that will help shift network dynamics in favor of a broader spread of ada. Expect more news on all these topics soon.

We recognize that in the short term, the move to k=500 will mean significant change for some. If delegators to larger pools do not react, some pools will become oversaturated and rewards will be unclaimed (note that no rewards are lost; everything goes back to the system’s reserves for the community to draw upon in the future). As a result, pool operators will need to adjust their margin and cost in the short term to stay profitable and incentivize delegators to take action. While this will require some effort from the community, it is an essential step for the Cardano ecosystem to maximize its decentralization. As smart contracts and multi-currency support will be coming soon to Cardano, a high level of decentralization will be the jewel of the ecosystem, and a strong competitive advantage over other blockchains.

The increase in k will be a significant step forward in delivering Cardano’s mission. Because this change will bring some disruption, we want to give the community plenty of time to absorb the changes and adjust their strategy. Moreover, we would like to help the community make the right decisions for the long-term sustainability of Cardano. We will be publishing more content over the next few weeks and months to support this approach (including a guide to making good delegation choices) as we continue to improve the experience.

Achieving the highest level of decentralization is the ultimate objective of any blockchain system. Decentralization is the solid foundation over which the Cardano ecosystem will thrive. True decentralization cannot be conjured purely out of mathematical theorems, and the tweaking of parameters and formulas will never be enough by itself – even though with Cardano we have taken huge strides compared to any other blockchain. Nevertheless, in the end, true decentralization will emanate from the collective will and actions of the Cardano community. We will continue in this path together, investigating additional approaches to allow stake pools to build their brand, communicate with their delegators, and highlight their own contributions and mission more directly.

Thanks to Aggelos Kiayias, Colin Edwards, Olga Hryniuk and Francisco Landino for their input to this piece. Thanks also to researchers Aikaterini-Panagiota Stouka and Elias Koutsoupias.

We’ll be posting throughout the next month to support delegators and stake pool operators as we approach the December 6 deadline. Please be sure to follow us on Twitter and subscribe on YouTube for all the latest updates.