Pruning Score
A pruning score expresses how a subnet decides which participant makes way when it is full and a newcomer wants in. It captures a simple principle of Bittensor: a place in a subnet is held by continuing to earn, and the participant that has earned the least is the one asked to leave. On a subnet such as netuid 1, that is the participant with the lowest pruning score.
References: Mining in Bittensor
A Place That Must Be Earned
A subnet has only so many places, so admitting a new participant means an existing one gives way. The pruning score is how that choice is framed: it stands for how much each participant has been earning, and the one earning the least is the one displaced. The effect is that holding a place is never permanent. It depends on continuing to contribute, which keeps a subnet’s limited capacity in the hands of those actively earning rather than those who simply arrived first.
References: Mining in Bittensor
What the Score Reflects
The pruning score reflects a neuron’s emissions, so it mirrors how much the network has been rewarding that participant rather than any outside rating. A participant earning well sits far from being displaced, while one earning little stays close to the edge. Reading the score this way makes its meaning clear: it is a standing that follows from performance, not a label applied independently of how a participant is doing.
References: Mining in Bittensor
A Grace Period for Newcomers
A newly admitted participant is shielded by an immunity period before this pressure applies. That grace window gives a newcomer time to establish itself and begin earning before its standing can put its place at risk, so the principle does not punish participants merely for being new. Once the grace period passes, the same expectation applies to everyone: a place is kept by earning.
References: Mining in Bittensor, Glossary: Immunity Period
Why It Matters
The pruning score turns a subnet’s limited places into a continuously contested resource. It is the idea behind participation in Bittensor being ongoing rather than a one-time entry: who holds a place is aligned with who is currently contributing, and a participant that stops earning gradually moves toward giving its place to someone who will. That keeps incentives pointed at sustained contribution.
References: Mining in Bittensor
Development Stage Context
The Introduction to Bittensor describes subnet development as moving from localnet to testnet and then mainnet. For pruning score, that sequence changes how readers should interpret subnet slot competition examples.
In localnet, pruning mechanics can be tested in an isolated environment. Localnet pruning scores do not represent production subnet participation.
On testnet, slot competition can be exercised in a shared non-production network. Testnet pruning outcomes are separate from mainnet subnet state.
On mainnet, pruning score describes live production subnet slot competition. Observed rankings depend on the selected subnet’s current emission and participation context (Mining in Bittensor).
The Bittensor Networks reference separates mainnet, testnet, and localnet. A pruning-score example from one environment should not be read as representing production slot competition in another environment.
Relationship to Yuma Consensus
Pruning Score 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, pruning score 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
This page describes the concept at a high level. Which participant actually stands lowest on a given subnet at a given time is live state that shifts as earnings change, and is best read from current chain data rather than inferred here. The durable point is the principle: a subnet place is earned, the least-earning participant makes way, and a grace period protects newcomers first.
References: Mining in Bittensor
Full Subnets Remove the Lowest-Scoring Incumbent
When a subnet is full, Mining: Miner deregistration identifies the existing participant with the lowest pruning score as the displacement candidate for an incoming registration. The newcomer receives the slot after that incumbent is removed, not because the newcomer itself held the lowest score.
That rule keeps pruning score on the incumbent side of registration pressure. A new miner can register only by displacing the weakest eligible participant already on the subnet (Glossary: Deregistration).
References: Mining: Miner deregistration, Glossary: Deregistration
UID Trimming Ranks by Emission Instead
UID Trimming describes a separate owner-initiated bulk reduction ranked by emission scores. Pruning score instead names the standing used when a full subnet must displace one participant to admit a new registration.
The two terms should not be swapped. Trimming is a manual resize with compression rules; pruning score is the performance signal cited in mining documentation for single-slot replacement on a full subnet (UID Trimming).
References: Mining in Bittensor, UID Trimming
Immunity Exceptions Can Affect Eligibility
Mining: Miner deregistration lists immunity-period exceptions among the eligibility rules that can affect which participant is removed when a subnet is full. A newly registered neuron may therefore be shielded for a time before it can be treated as a displacement candidate.
Immunity does not change what pruning score measures. It changes who is eligible for removal at a given moment while the score still reflects emission standing (Glossary: Immunity Period).
References: Mining: Miner deregistration, Glossary: Immunity Period