Conviction and Locked Stake
Conviction and locked stake describe a Bittensor staking feature where subnet alpha stake can be locked so that it builds a time-based commitment signal (Conviction and locked stake, Staking and delegation overview).
The feature is commitment vocabulary. It does not create a separate token, a separate subnet role, or a general staking recommendation.
Locked Alpha Stake
Locked stake is subnet alpha stake constrained by a lock. The conviction staking documentation explains that the locked amount limits how far the staked alpha balance on that subnet can fall while the lock is active (Conviction and locked stake, Glossary: Alpha Tokens).
That makes locked stake narrower than ordinary stake. The stake still belongs to a subnet alpha context, but the locked portion carries an additional commitment constraint.
The subnet context should stay visible whenever the term is used. Conviction is tied to locked alpha in a subnet position, not to every staking balance in Bittensor.
Conviction Score
Conviction is the score that grows from locked stake over time. The locked amount and the passage of time both matter, so conviction should be read as a maturing commitment signal rather than a simple balance number (Conviction score, Conviction and locked stake).
This keeps conviction separate from stake itself. Stake names the amount associated with a staking position; conviction names the time-based signal built from locked stake.
Lock Modes
The conviction model includes decaying and perpetual modes. Decaying mode reduces the locked amount over time, while perpetual mode keeps the locked amount from decaying until the mode changes (Decaying and perpetual modes, Conviction and locked stake).
The mode affects how the commitment signal evolves. It should not be treated as a separate asset type or as a promise about validator rewards.
Subnet and Stake Context
Conviction and locked stake apply to subnet alpha already staked inside a specific subnet. Alpha tokens are subnet-specific, and staking into subnets places stake in subnet-local context (Glossary: Alpha Tokens, Staking into Subnets).
The concept is therefore separate from Root Subnet TAO staking and from ordinary unlocked stake. An alpha position can be staked without being locked; conviction begins when part of that staked alpha is locked for commitment signaling.
Development Stage Context
Bittensor separates localnet, testnet, and mainnet environments. Conviction examples from one environment should not be treated as evidence for another because lock state, alpha balances, and subnet conditions are environment-specific (Bittensor Networks, Introduction to Bittensor: Subnet development).
Localnet examples are isolated development examples. Testnet examples are shared non-production examples. Mainnet conviction and locked stake concern production subnet alpha and production subnet state.
Relationship to Yuma Consensus
Conviction and Locked Stake and Yuma Consensus describe related parts of Bittensor’s incentive system. Yuma Consensus is the on-chain process that aggregates validator weight signals within a subnet into miner incentives and validator dividends, applying consensus clipping, bonding, and emission calculation (Yuma Consensus).
For readers, conviction and locked stake names a specific part of that incentive picture, while Yuma Consensus names the consensus process that turns validator weights into the resulting incentives and dividends.
Reader Boundary
Conviction and locked stake are commitment vocabulary for subnet alpha, not a separate token, a validator recommendation, or a reward forecast (Conviction and locked stake, Glossary: Alpha Tokens).
Conviction names the time-based signal built from locked stake; stake names the amount in the position. Root Subnet TAO staking and ordinary unlocked alpha stake are separate paths (Glossary: Root Subnet, Staking into Subnets).
Movement Boundary
Locked stake should be read as a movement constraint on an existing subnet alpha staking position, not as a new wallet or a separate token balance. The conviction-staking documentation describes the lock in terms of how far the staked alpha balance can fall, while subnet staking keeps the position inside subnet-local alpha context (Conviction and locked stake, Staking into Subnets).
That distinction keeps the commitment signal separate from custody language. The lock constrains movement of the staked alpha position; it does not rename the asset or make conviction a wallet authority term.
Surplus Stake Above the Lock Stays Movable
Official Conviction and locked stake documentation describes the lock as a floor on how low total staked Glossary: Alpha Tokens on a subnet can fall for one coldkey. Stake above that locked floor is not itself locked and can leave through ordinary Glossary: Unstaking paths (Staking into Subnets).
The same guide also states that a coldkey can keep staking additional alpha on the subnet after a lock exists. The lock blocks only reductions below the locked amount; it does not cap how much new alpha can be added to the position.
That reading separates the commitment floor from the full staking position. Conviction names the signal built from the locked portion, while surplus alpha above the floor remains ordinary movable Glossary: Stake on the same subnet position.
References: Conviction and locked stake, Staking into Subnets
Exiting a Lock Starts With a Public Signal
New locks begin in decaying mode by default under official conviction documentation. A holder who wants the locked mass to stop shrinking must explicitly switch to perpetual mode; returning to decaying mode starts the unwind path instead (Decaying and perpetual modes, Conviction and locked stake).
The documentation treats that mode change as visible commitment vocabulary rather than a silent local setting. Switching from perpetual back to decaying is recorded on chain immediately, and the following decay period exists so other stakers can observe a large holder winding down before the lock fully clears.
That timing keeps conviction aligned with its role as a public commitment signal. A reader watching subnet stake should treat a decaying-mode switch as advance notice that locked mass will shrink over time rather than as an instant release of the whole position.
References: Decaying and perpetual modes, Conviction and locked stake
Decaying Conviction Can Outlast the Remaining Lock
In decaying mode, official conviction documentation describes two linked curves: locked mass erodes over time while conviction first rises and then falls. Because conviction accumulates maturing time from the lock, it can peak partway through the decay and later remain higher than the alpha still locked at that moment (The conviction score, Decaying and perpetual modes).
That behavior makes decaying conviction a record of past commitment rather than a live mirror of current locked holdings. After the lock has been shrinking for long enough, the score can reflect duration already served even when much of the original locked mass is gone.
Readers comparing conviction to wallet balances should therefore keep the two separate. Stake names what is held now; decaying conviction can still carry weight from earlier locked periods until both values finish falling toward zero.
References: The conviction score, Conviction and locked stake