Polkadot is a layer-0 protocol that allows entire blockchains to be built on top of it, unlike layer-1 blockchains built for protocol deployment on connected layers (see here for more). It features shards, or parachains (i.e., chains running in parallel), as well as the Relay Chain, which connects the parachains.
Polkadot runs Nominated Proof-of-Stake (NPoS) consensus, consisting of two key players:
Nominators: DOT holders earn rewards by nominating validators, which uses their DOT to elect these validators toward the active set, and entitles them to a share of the rewards received by, as well as the penalties levied against, these validators.
Here is the current breakdown of validators to nominators; you can see that nominators are clearly in the majority in terms of stake:
Every era (about 24 hours for Polkadot), a new batch of 297 validators is chosen to be in the active set. These validators are selected based on the total stake they have, which includes DOT that the validators have staked plus any DOT they have received through nomination. The top 297 validators by total elected stake are chosen to be listed as ‘active’ and ready to be selected for participation in consensus.
Rewards are paid out every era and each validator receives the same amount of DOT (ignoring slashing for the time being). After the validator takes their commission, the remaining DOT is split between the validator and the nominators in proportion to their stake.
This structure incentivizes validators to stake the minimal amount of DOT to be included in the active set; by staking the minimal amount of DOT, validators maximize profitability. This metric is known as the Minimum Stake Required (MSR hereforth). Taken to an extreme, this tends toward validators holding an equal amount of stake; i.e., all validators seek to stake the minimum amount that is just enough to be included in the active set. You can see this lack of dispersion in the top fifteen validators:
Penalties are applied for various behaviors, ranging from less than 1% for more benign infractions, such as being offline, to a validator’s entire stake (i.e., 100%) for actions judged to most likely be malicious, i.e., an attack on the network. For a more complete discussion of slashing see here.
In addition to having some stake taken away, all infractions come with chilling. In the event that this is part of a punishment, the validator is un-nominated, excluded from the remainder of the current era and the validator is removed from the next election. It should be noted that validators can voluntarily engage in chilling, if they wish to remove themselves from the active set for some period of time.
Nominators can select up to sixteen validators for a particular era. There are some aspects of choosing validators, which are fairly obvious, such as choosing validators that perform well and are least likely to be slashed. Another somewhat intuitive idea is that nominators should generally choose validators with less stake; although slashing would be the same regardless of the validator chosen, rewards will be higher when a nominator chooses a validator with less stake (because all validators receive the same rewards, slashing aside, and the rewards are split proportionally between the validator and nominators, after commission).
However, there are some nuances that may not be immediately apparent. For example, staking rewards are only distributed to the top 256 nominators of a particular validator (ranked by stake size).If someone is unlucky enough to be number 257, they don’t receive any rewards from that particular validator during that particular era. Unfortunately, slashing is not the same - all nominators will be slashed, including unlucky number 257. In other words, it is very important not to nominate an oversubscribed validator.
In addition, when deciding which sixteen validators to nominate DOT holders would be well-advised to follow a portfolio building approach. Specifically, DOT holders should recognize the benefits of diversification. A nominator who only elects a single validator will only earn rewards when that validator is picked for consensus. Confusion over this fact often leads to nominators not understanding why they are earning intermittent rewards - a diversification of elected stake across up to 16 validators maximizes rewards for nominators so that they can earn (on average) as close to the inflation rate as possible.
Validators holding a large amount of DOT are incentivized to run multiple validators (as discussed above) and as can be seen here:
Suppose as a nominator you believe a provider to be the best operator (i.e., validator), with the least risk of downtime/slashing. You might be tempted to nominate your stake to as many of a single provider’s validators as possible, but this could be a mistake. Should a provider run all of their validators on the same infrastructure and it were to go down, you lose across all of your nominations. If however, you only make a few nominations to a single provider, and make others based on their ability to diversify (i.e., operators running in different locations and on different setups/infrastructure) you have created a superior portfolio of nominations on a risk-adjusted basis.
This approach also potentially reduces the risk of nominating an oversubscribed validator. If everyone flocks to the best operators, there will be many nominators that lose out on rewards altogether because they are not in the top 256. Should a nominator know that they have the largest stake (compared to other nominators) then this is not a concern - the top 256 nominators who earn rewardsare determined by stake size.
It’s important to understand how large DOT flows can affect the dynamic mentioned above - validators maximize returns by accruing the minimum amount of DOT but enough to still be included in the active set; implying a large DOT holder would do best to spread their stake across multiple validators in this active set.
Outflows are fairly straightforward - because the unbonding period takes 28 days, users often have advanced notice about how much DOT is set to unbond at a given point:
If the amount of DOT bonded is set to decrease, this will, typically, lower the minimum amount of DOT needed to be included in the active set. Although not always possible, profit maximizing DOT holders could seek to spread their stake over a greater number of validators with this information.
On the other hand, the same is not true for DOT staking inflows - whichare often harder to predict. Staking inflows are bonded and begin earning rewards in the next epoch (~1 day), meaning that a nominator can bond and nominate their stake on one day, then earn rewards from the active validators the next.
Crowdloans are an interesting source of potential inflows. New projects wishing to launch on Polkdadot need a slot on the relay chain. To obtain this slot, they need to bid for it in an auction in competition with other projects bidding for the same slot. Projects can use DOT they have acquired to bid, or they can seek crowdloans from other DOT holders. These DOT holders lock up their DOT for the length of the slot lease (see here for more). The important point is that it is possible to predict when DOT will be released from these crowdloans. The problem is that one then has to infer what the DOT holder will do next - stake? Find a new crowdloan? Sell their DOT? Etc.
When the two of these mechanics (inflows/outflows) interact, they work together across the nominations of validators to determine the Minimum Stake Required for a validator to get into the active set- making it so that this dynamic ‘floor’ to what it takes to be a validator can fluctuate. This means that for ideal reward extraction, nominators with a significant amount of capital should distribute their stake across the validators where their stake has a maximal impact on election, therefore maximizing their portion of rewards.
Maximizing profit is therefore a dynamic and active process for both validators and nominators. Validators must consider how thinly they can potentially spread their DOT across validators while being assured they will be in the active set. Nominators need to figure out which validators to bond their DOT to, while ensuring that they are in the top 256 positions by stake size and that they are well-diversified.