Staking

Geode implements NPoS (Nominated Proof-of-Stake), a relatively novel and sophisticated mechanism to select the validators who are allowed to participate in its consensus protocol. The NPoS model encourages the majority of GEODE holders to participate as nominators. Nominators may back up to16 validators as trusted validator candidates. Both validators and nominators lock (bond) their tokens as collateral and receive staking rewards when they produce blocks as a validator or when their nominations were effective.

The staking system pays out rewards essentially equally to all validators regardless of stake. Having more stake on a validator does not influence the amount of block rewards it receives. However, there is a probabilistic component to reward calculation (discussed below), so rewards may not be exactly equal for all validators in a given era.

Distribution of the rewards are pro-rata to all stakers after the validator’s commission is deducted. In this way, the network creates incentives for the nomination of lower-staked validators to create an equally-staked validator set.

How does staking work?

1. Identifying which role you would like

In staking, you can be either a nominator or a validator.  You can also be both, as long as you have a separate stash and controller account for each. As a nominator, you can nominate validator candidates that you trust to help you earn rewards in GEODE. The earned rewards can be bonded (locked) immediately for staking on your account, which would effectively compound the rewards you receive over time from your efforts. You could also choose to have them deposited to your account or a different account as free (transferable) balance. See the details on being a nominator or validator throughout the Geode documentation. 

2. Staking System Overview

Any potential validators can indicate their intention to be a validator candidate. Their candidacies are made public to all nominators, and a nominator in turn submits a list of up to 16 candidates that it supports. In the next era, a certain number of validators having the most GEODE backing get elected and become active. As a nominator, a minimum of 1000 GEODE is required to submit an intention to nominate. The nomination intents are placed in a semi-sorted list called the bags-list.

MINIMUM NOMINATION TO RECEIVE STAKING REWARDS – Although the minimum nomination intent is 1000 GEODE, it does not guarantee staking rewards. The nominated amount has to be greater than minimum active nomination on the system, which is a dynamic value that can be much higher than 1000 GEODE.

The bags list has two primary components, bags and nodes. The list is composed of bags that each describe a range of active bonded funds (e.g. the 1st bag will have nominators with 0 → 10 GEODE, 2nd bag 11 → 20 GEODE, etc). In each bag is a list of nodes that correspond to a nominator and their staked funds. The bags-list pallet is designed to be self-maintaining, with minimal effort from the blockchain, making it extremely scalable. 

Actions like bonding/unbonding tokens automatically rebags a nominator node, but events like staking rewards/slashing do not! The bags-list pallet comes with an important permissionless extrinsic: rebag. This allows anyone to specify another account that is in the wrong bag, and place it in the correct one. 

This sorting functionality is extremely important for the staking/election system. The bags-list is capable of including an unlimited number of nodes, subject to the chain’s runtime storage. In the current staking system configuration, the bags list keeps 50,000 nomination intents, of which, at most 22,500 come out as the electing nominators. 

MINIMUM ACTIVE NOMINATION THRESHOLD TO EARN REWARDS IS DYNAMIC – Once again, submitting a nomination intent does not guarantee staking rewards. The stake of the top 22,500 nominators is applied to the validators in the active set. To avail staking rewards, ensure that the number of tokens bonded is higher than the minimum active nomination. 

Once the nomination period ends, the NPoS election mechanism takes the nomination intents and their associated votes as input, and outputs a set of validators. The bags are iterated from the most staked to the least staked. This could leave the last touched bag to only be partially iterated. This means that in some edge cases, the order of members within a bag is also important. Recall that within each bag, the iteration order is simply the insertion order. If only 7 nodes have to be picked for the electing set, the nodes with 5 and 7 GEODE will be selected and the node with 8 GEODE will be left out. The bags-list pallet comes with an extrinsic: putInFrontOf which helps the node to move up in the bag. 

The “election solution” has to meet certain requirements, such as maximizing the amount of stake to nominate validators and distributing the stake backing validators as evenly as possible. The objectives of this election mechanism are to maximize the security of the network, and achieve fair representation of the nominators. 

We expect each nominator to carefully track the performance and reputation of the validators they back. Nominating is not a “set and forget” operation, but an active and ongoing effort by the Nominator.

3. Staking Rewards Distribution

To explain how rewards are paid to validators and nominators, we need to consider validator pools. A validator pool consists of the stake of an elected validator together with the nominators backing it. If a nominator n with stake s backs several elected validators, say k, the NPoS election mechanism will split its stakes into pieces s_1, s_2, …, s_k, so that it backs validator i with stake s_i. In that case, nominator n will be rewarded essentially the same as if there were k nominators in different pools, each backing a single validator i with stake s_i.

For each validator pool, we keep a list of nominators with the associated stakes. The general rule for rewards across validator pools is that two validator pools get paid essentially the same amount of tokens for equal work, i.e. they are NOT paid proportional to the stakes in each pool. There is a probabilistic component to staking rewards in the form of era-points and tips, but these should average out over time.

Within a validator pool, a (configurable) percentage of the reward goes to pay the validator’s commission fees and the remainder is paid pro-rata (i.e. proportional to stake) to the nominators and validator. Notice in particular that the validator is rewarded twice: once in commission fees for validating (if their commission rate is above 0%), and once for nominating itself with stake. If a validator’s commission is set to 100%, no tokens will be paid out to any nominations in the validator pool.

4. Rewards Mechanism

We highlight two features of this reward scheme. The first is that since validator pools are paid the same regardless of stake level, pools with less stake will generally pay more to nominators per-token than pools with more stake. We thus give nominators an economic incentive to gradually shift their preferences to lower staked validators that gain a sufficient amount of reputation. The reason for this is that we want the stake across validator pools to be as evenly distributed as possible, to avoid a concentration of power among a few validators. In the long term, we expect all validator pools to have similar levels of stake, with the stake being higher for higher reputation validators (meaning that a nominator that is willing to risk more by backing a validator with a low reputation will get paid more).

The following example should clarify the above. For simplicity, we have the following assumptions:

• These validators do not have a stake of their own.
• They each receive the same number of era points.
• There are no tips for any transactions processed.
• They do NOT charge any commission fees.
• Total reward amount is 100 GEODE tokens.
• The current minimum amount of GEODE to be a validator is 350 (note that this is not the actual value, which fluctuates, but merely an assumption for purposes of this example).

	A – Validator Pool
Nominator (4)	Stake (600)	Fraction of the Total Stake	Rewards
Jin	100	0.167	16.7
Sam	50	0.083	8.3
Anson	250	0.417	41.7
Bobby	200	0.333	33.3

	B – Validator Pool
Nominator (4)	Stake (400)	Fraction of the Total Stake	Rewards
Alice	100	0.25	25
Peter	100	0.25	25
John	150	0.375	37.5
Kitty	50	0.125	12.5

Both validator pools A & B have 4 nominators with the total stake 600 and 400 respectively.

Based on the above rewards distribution, nominators in validator pool B get more rewards per GEODE than those in pool A because pool A has more overall stake. Sam has staked 50 GEODE in pool A, but he only gets 8.3 in return, whereas Kitty gets 12.5 with the same amount of stake.

There is an additional factor to consider in terms of rewards. While there is no limit to the number of nominators a validator may have, a validator does have a limit to how many nominators to which it can pay rewards. In Geode, this limit is currently 256, although it can be modified via runtime upgrade. A validator with more than 256 nominators is oversubscribed. When payouts occur, only the top 256 nominators as measured by amount of stake allocated to that validator will receive rewards. All other nominators are essentially “wasting” their stake – they used their nomination to elect that validator to the active stake, but receive no rewards in exchange for doing so.

We also note that when the network slashes a validator slot for a misbehavior (e.g. validator offline, equivocation, etc.) the slashed amount is a fixed percentage (and NOT a fixed amount), which means that validator pools with more stake get slashed more GEODE. Again, this is done to provide nominators with an economic incentive to shift their preferences and back less popular validators whom they consider to be trustworthy.

The second point to note is that each validator candidate is free to name their desired commission fee (as a percentage of rewards) to cover operational costs. Since validator pools are paid the same, pools with lower commission fees pay more to nominators than pools with higher fees. Thus, each validator can choose between increasing their fees to earn more, or decreasing their fees to attract more nominators and increase their chances of being elected. In the long term, we expect that all validators will need to be cost efficient to remain competitive, and that validators with higher reputation will be able to charge slightly higher commission fees (which is fair).

Accounts

There are two different accounts for managing your funds: Stash and Controller.

• Stash: This account holds funds bonded for staking, but delegates some functions to a Controller. As a result, you may actively participate with a Stash key kept in a cold wallet, meaning it stays offline all the time. You can also designate a Proxy account to vote on governance proposals.
• Controller: This account acts on behalf of the Stash account, signaling decisions about nominating and validating. It sets preferences like payout account and commission. If you are a validator, it also sets your session keys. It only needs enough funds to pay transaction fees.

We designed this hierarchy of separate key types so that validator operators and nominators can protect themselves much better than in systems with only one key. As a rule, you lose security anytime you use one key for multiple roles, or even if you use keys related by derivation. You should never use any account key for a “hot” session key in particular.

Controller and Stash account keys can be either sr25519 or ed25519 but generally are sr25519 encryption.. 

Validators and Nominators

Since validator slots are somewhat limited, most of those who wish to stake their GEODE and contribute security to the network will be nominators. Validators do most of the heavy lifting: they produce new block candidates in BABE and vote and come to consensus in GRANDPA. Nominators, on the other hand, have far fewer responsibilities. Those include monitoring their validators’ performance (uptime), keeping an eye on changing commission rates (a validator can change commission at any time), and general health monitoring of their and their validators’ account. Thus, while not set-it-and-forget-it, a nominator’s experience is relatively lower pressure compared to a validator’s.

Slashing

Slashing will happen if a validator misbehaves (e.g. goes offline, attacks the network, or runs modified software) in the network. They and their nominators will get slashed by losing a percentage of their bonded/staked GEODE. Any slashed GEODE will be added to the Treasury. The rationale for this (rather than burning or distributing them as rewards) is that slashes may then be reverted by the Council by simply paying out from the Treasury. This would be useful in situations such as a faulty runtime causing slashing or forcing validators offline through no fault of their own. In the case of legitimate slashing, it moves tokens away from malicious validators to those building the ecosystem through the normal Treasury process.

Validator pools with larger total stake backing them will get slashed more harshly than less popular ones, so we encourage nominators to shift their nominations to less popular validators to reduce their possible losses.

It is important to realize that slashing only occurs for active validations for a given nominator, and slashes are not mitigated by having other inactive or waiting nominations. They are also not mitigated by the validator operator running separate validators; each validator is considered its own entity for purposes of slashing, just as they are for staking rewards.

As an example, assume BIG_COMPANY has 50 validators that all go offline at the same time, thus causing a 1% unresponsiveness slash to their nominators. In this example, the nominator has nominated five validators, two of which are with BIG_COMPANY (BC_1 and BC_2) and three are with other validators that do not belong to BIG_COMPANY (OV_1, OV_2, and OV_3). In this era, BC_1 is the active validator for this nominator, BC_2 and OV_1 are inactive, and OV_2 and OV_3 are waiting. The nominator will be slashed 1% of bonded stake, since BC_1 is the active validator. The inactive and waiting validators (BC_2 and OV_1 through 3) don’t have any effect on this, since they are not actively validating. Any nominator actively nominating BC_2 also receives a 1% slash, but any nominator actively nominating OV_1 is unaffected.

In rare instances, a nominator may be actively nominating several validators in a single era. In this case, the slash is proportionate to the amount staked to that specific validator. For instance, if another nominator had their stake split 50% to BC_1 and 50% to OV_1, they would receive a slash of 0.5% (50% of 1%). If a nominator were actively nominating BC_1 and BC_2, again with 50% of their stake allocated to each, they would still end up with a 1% slash, since a 1% slash is applied to both halves of their stake. Note that you cannot control the percentage of stake you have allocated to each validator or choose who your active validator will be (except in the trivial case of nominating a single validator). Staking allocations are controlled by the Phragmen algorithm. 

Once a validator gets slashed, it goes into the state as an “unapplied slash“. The Geode Portal shows it per validator and then all the affected nominators along with the amounts. While unapplied, a governance proposal can be made to reverse it during this period (usually 7-28 days). After the grace period, the slashes are applied.

The following levels of offence are defined. However, these particular levels are not implemented or referred to in the code or in the system; they are meant as guidelines for different levels of severity for offences. To understand how slash amounts are calculated, see the equations in the section below.

• Level 1: isolated unresponsiveness, i.e. being offline for an entire epoch. Generally no slashing, only chilling.
• Level 2: concurrent unresponsiveness or isolated equivocation. Slashes a very small amount of the stake and chills.
• Level 3: misconducts unlikely to be accidental, but which do not harm the network’s security to any large extent. Examples include concurrent equivocation or isolated cases of unjustified voting in GRANDPA. Slashes a moderately small amount of the stake and chills.
• Level 4: misconduct that poses a serious security or monetary risk to the system, or mass collusion. Slashes all or most of the stake behind the validator and chills.

Let’s look at these offences in a bit more detail.

Unresponsiveness

For every session, validators will send an “I’m online” heartbeat to indicate they are live. If a validator produces no blocks during an epoch and fails to send the heartbeat, it will be reported as unresponsive. Slashing may occur depending on the repeated offences and how many other validators were unresponsive or offline during the epoch.

Validators should have a well-architected network infrastructure to ensure the node runs to reduce the risk of slashing or chilling. A high availability setup is desirable, preferably with backup nodes that kick in only once the original node is verifiably offline (to avoid double-signing and being slashed for equivocation – see below). 

A comprehensive guide on validator setup is available in the documentation section for Validators here at geodechain.com

Here is the formula for calculating slashing due to unresponsiveness:

Let x = offenders, n = total no. validators in the active set

min((3 * (x – (n / 10 + 1))) / n, 1) * 0.07

The examples demonstrate how to calculate the slashing penalty for unresponsiveness.

In all of the examples, assume that there are 100 validators in the active set.

• No slashing would enact if < 10% of all validators are unresponsive.
• For example, if exactly 10 validators were unresponsive, the expression 3 (x – (n / 10 + 1))) / n would be 3 (10 – (100 / 10 + 1)) / 100 = 3 * (10 – (10 + 1)) / 100 = -0.03 which is rounded to 0.
• The minimum value between 0 and 1 is 0. 0 multipled by 0.07 is 0.
• If 14 validators are unresponsive, then slashing would occur, as > 10% of validators are unresponsive.
• The slashing penalty would be min((3 (14 – (100 / 10 + 1))) / 100, 1) 0.07 = min((3 (14 – 11))/100, 1) 0.07 = min(0.09, 1) * 0.07 = 0.6%
• Similarly, if one-third of the validator set (around 33/100) are unresponsive, the slashing penalty would be about 5%.
• The maximum slashing that can occur due to unresponsiveness is 7%. After around 45% of the validators go offline, the expression 3 (x – (n / 10 + 1))) / n will go beyond 1. Hence, min((3 (x – (n / 10 + 1))) / n, 1) *0.07 will be ceiled to 7%.

Equivocation

There are two types of equivocation…

• GRANDPA Equivocation: A validator signs two or more votes in the same round.
• BABE Equivocation: A validator produces two or more blocks in the same time slot.

Both GRANDPA and BABE equivocation use the same formula for calculating the slashing penalty:

Let x = offenders, n = total no. validators in active set

min( (3 * x / n )^2, 1)

As an example, assume that there are 100 validators in the active set, and one of them equivocates in a slot (for our purposes, it does not matter whether it was a BABE or GRANDPA equivocation). This is unlikely to be an attack on the network, but much more likely to be a misconfiguration of a validator. The penalty would be min(3 * 1 / 100)^2, 1) = 0.0009, or a 0.09% slash for that validator pool (i.e., all stake held by the validator and its nominators).

Now assume that there is a group running several validators, and all of them have an issue in the same slot. The penalty would be min((3 * 5 / 100)^2, 1) = 0.0225, or a 2.25% slash. If 20 validators equivocate, this is a much more serious offence and possibly indicates a coordinated attack on the network, and so the slash will be much greater – min((3 *20 / 100)^2, 1) = 0.36, or a 36% slash on all of these validators and their nominators. All slashed validators will also be chilled.

From the example above, the risk in nominating or running many validators in the active set are apparent. While rewards grow linearly (two validators will get you approximately twice as many staking rewards as one), slashing grows exponentially. A single validator equivocating causes a 0.09% slash, two validators equivocating does not cause a 0.09 * 2 = 0.18% slash, but rather a 0.36% slash (4x as much as the single validator).

Validators may run their nodes on multiple machines to make sure they can still perform validation work in case one of their nodes goes down, but validator operators should be extremely careful in setting these up. If they do not have good coordination to manage signing machines, equivocation is possible, and equivocation offences are slashed at much higher rates than offline offences.

If a validator is reported for any one of the offences they will be removed from the validator set (chilled) and they will not be paid while they are out. They will be considered inactive immediately and will lose their nominators. They need to re-issue intent to validate and again gather support from nominators.

Chilling

Chilling is the act of stepping back from any nominating or validating. It can be done by a validator or nominator at any time themselves, taking effect in the next era. It can also specifically mean removing a validator from the active validator set by another validator, disqualifying them from the set of electable candidates in the next NPoS cycle.

Chilling may be voluntary and validator-initiated, e.g. if there is a planned outage in the validator’s surroundings or hosting provider, and the validator wants to exit to protect themselves against slashing. When voluntary, chilling will keep the validator active in the current session, but will move them to the inactive set in the next. The validator will not lose their nominators.

When used as part of a punishment (initiated externally), being chilled carries an implied penalty of being un-nominated. It also disables the validator for the remainder of the current era and removes the offending validator from the next election.

Geode allows some validators to be disabled, but if the number of disabled validators gets too large, Geode will trigger a new validator election to get a full set. Disabled validators will need to resubmit their intention to validate and re-garner support from nominators.

Slashing Across Eras

There are 3 main difficulties to account for with slashing in NPoS:

• A nominator can nominate multiple validators and be slashed via any of them.
• Until slashed, stake is reused from era to era. Nominating with N coins for E eras in a row does not mean you have N*E coins to be slashed – you’ve only ever had N.
• Slashable offences can be found after the fact and out of order.

To balance this, we only slash for the maximum slash a participant can receive in some time period, rather than the sum. This ensures protection from overslashing. Likewise, the time span over which maximum slashes are computed are finite and the validator is chilled with nominations withdrawn after a slashing event, as stated in the previous section. This prevents rage-quit attacks in which, once caught misbehaving, a participant deliberately misbehaves more because their slashing amount is already maxed out.

Reward Distribution

Geode produces blocks every 6 seconds. Rewards are calculated per era (approximately twenty-four hours on Geode). These rewards are calculated based on era points, which have a probabilistic component. In other words, there may be slight differences in your rewards from era to era, and even amongst validators in the active set at the same time. These variations should cancel out over a long enough timeline. See the page on Validator Payout Guide for more information on how these are calculated.

In order to be paid your staking rewards, someone must claim them for each validator that you nominate. Staking rewards are kept available for 84 eras, which is approximately 84 days on Geode. If nobody claims your staking rewards by this time, then you will not be able to claim them and some of your staking rewards will be lost. Additionally, if the validator unbonds all their own stake, any pending payouts will be lost. Since unbonding takes 28 days on Polkadot, nominators should check if they have pending payouts at least this often.

Claiming Rewards

If you go to the Network > Staking > Payouts page on the Geode Portal, you will see a list of all validators that you have nominated in the past 84 eras and for which you have not yet received a payout. Each one has the option to trigger the payout for all unclaimed eras. Note that this will pay everyone who was nominating that validator during those eras, and anyone can call it. Therefore, you may not see anything in this tab, yet still have received a payout if somebody (generally, but not necessarily, another nominator or the validator operator) has triggered the payout for that validator for that era.

If you wish to check if you received a payout, you will have to check via a block explorer. 

Validators can create a cut of the reward (a commission) that is not shared with the nominators. This cut is a percentage of the block reward, not an absolute value. After the commission gets deducted, the remaining portion is based on their staked value and split between the validator and all of the nominators who have voted for this validator.

For example, assume the block reward for a validator is 10 GEODE. A validator may specify validator_commission = 50%, in which case the validator would receive 5 GEODE. The remaining 5 GEODE would then be split between the validator and their nominators based on the proportion of stake each nominator had. Note that validators can put up their own stake, and for this calculation, their stake acts just as if they were another nominator.

Rewards can be directed to the same account (controller), to the stash account (and either increasing the staked value or not increasing the staked value), or to a completely unrelated account. By using the Extrinsics tab (Developer -> Extrinsics -> Staking -> Bond) you can also send rewards to “None”, effectively burning them. It is also possible to top-up / withdraw some bonded GEODE without having to un-stake all staked GEODE.

Inflation

GEODE is inflationary; there is no maximum number of GEODE. Inflation is designed to be approximately 10% annually, with validator rewards being a function of the amount staked and the remainder going to treasury. 

There is an ideal staking rate that the network tries to maintain about 50-75%. The goal is to have the system staking rate meet the ideal staking rate. The system staking rate would be the total amount staked over the total token supply, where the total amount staked is the stake of all validators and nominators on the network. The ideal staking rate accounts for having sufficient backing of GEODE to prevent the possible compromise of security while keeping the native token liquid. GEODE is inflated according to the system staking rate of the entire network. According to the inflation model, this would suggest that if you do not use your GEODE for staking, your tokens dilute over time. 

If the amount of tokens staked goes below the ideal rate, then staking rewards for nominators go up incentivizing nominators to stake more tokens on the network. On the contrary, if it goes above the ideal rate, staking rewards drop. This is a result of the change in the percentage of staking rewards that go to the Treasury.

For instance, assuming that the ideal staking rate is 50%, all of the inflation would go to the validators/nominators if 50% of all GEODE are staked. Any deviation from the 50% – positive or negative – sends the proportional remainder to the treasury and effectively reduces staking rewards.

For those who are interested in knowing more about the design of inflation model for the network, please see here.

Why not stake?

With all this discussion about staking, you might wonder why someone wouldn’t stake all they can. There are plenty of good reasons to not stake everything, or for some users, to not stake at all…

• Tokens will be locked for about 28 days on Geode after unbonding. You may prefer to have your tokens more liquid than that.
• Punishments in case of validators found to be misbehaving (Slashing). You may not like the slashing risk.
• You want to use your tokens for other activities on the system. You might just have other plans for your tokens such as making claims and endorsements and using the various applications, voting, and a host of other things.