Subnet Ranking System

    Developed by Yuma, a DCG Company

    Yuma is a longtime, trusted validator on the Bittensor Network. We believe in the collective mission to decentralize machine intelligence.

    While there are changes in motion to further decentralize Bittensor with Dynamic TAO, validators currently determine the flow of emissions throughout the network with each block. We support builders with a longer-term view of decentralized AI development across the stack and have therefore developed a process to assign subnet weights in an objective and transparent manner.

    The Process

    Collecting Information

    Our process begins with discovery. In order to evaluate subnets, we must understand the purpose, the execution strategy, and the current state of their operation. This is increasingly hard to gather on an individual basis as the number of subnets expands.

    Yuma hosts a form available to all Subnet Owners to share information about the value proposition, roadmap, and longer-term goals of their project. After receiving the completed form, we often reach out to clarify gaps in our understanding and provide feedback to steer the team towards success.

    Collecting Information

    Our process begins with discovery. In order to evaluate subnets, we must understand the purpose, the execution strategy, and the current state of their operation. This is increasingly hard to gather on an individual basis as the number of subnets expands.

    Yuma hosts a form available to all Subnet Owners to share information about the value proposition, roadmap, and longer-term goals of their project. After receiving the completed form, we often reach out to clarify gaps in our understanding and provide feedback to steer the team towards success.

    Subnet Evaluation

    We translate our subnet research to objective scores for a set of 9 criteria which capture 3 themes: the idea (25%), technical implementation (35%), and team execution (40%). The scoring system incorporates a geometric progression scale (scores 1, 3, 9) to effectively rank the impact. Each criterion has an associated weight to signify their importance relative to each other.

    ThemeWeightCriteria
    The Idea5%Innovative - Is the task and/or digital commodity new to the ecosystem?
    10%Benefits from decentralized architecture - Is a decentralized solution necessary?
    10%End-user proposition - Will customers want to use the digital commodity?
    The Technical Implementation10%End-user interface - Does the subnet provide a high quality interface?
    5%Distribution of miner rewards - Is the mining incentive distribution fair?
    15%Rewards model - Does the evaluation metric use an objective ground truth?
    5%Barriers to mining - How easily can one deploy and sustain a miner?
    The Team Execution20%Team credibility - Is the subnet owner and development team reputable?
    20%Team velocity - How frequent does the team update the codebase?
    The Idea (25%)
    5%
    Innovative - Is the task and/or digital commodity new to the ecosystem?
    10%
    Benefits from decentralized architecture - Is a decentralized solution necessary?
    10%
    End-user proposition - Will customers want to use the digital commodity?

    Subnet Evaluation

    We translate our subnet research to objective scores for a set of 9 criteria which capture 3 themes: the idea (25%), technical implementation (35%), and team execution (40%). The scoring system incorporates a geometric progression scale (scores 1, 3, 9) to effectively rank the impact. Each criterion has an associated weight to signify their importance relative to each other.

    ThemeWeightCriteria
    The Idea5%Innovative - Is the task and/or digital commodity new to the ecosystem?
    10%Benefits from decentralized architecture - Is a decentralized solution necessary?
    10%End-user proposition - Will customers want to use the digital commodity?
    The Technical Implementation10%End-user interface - Does the subnet provide a high quality interface?
    5%Distribution of miner rewards - Is the mining incentive distribution fair?
    15%Rewards model - Does the evaluation metric use an objective ground truth?
    5%Barriers to mining - How easily can one deploy and sustain a miner?
    The Team Execution20%Team credibility - Is the subnet owner and development team reputable?
    20%Team velocity - How frequent does the team update the codebase?
    The Idea (25%)
    5%
    Innovative - Is the task and/or digital commodity new to the ecosystem?
    10%
    Benefits from decentralized architecture - Is a decentralized solution necessary?
    10%
    End-user proposition - Will customers want to use the digital commodity?

    Scores to Weight Conversion

    Weight assignments are one vehicle to support projects on an annual basis. As we move towards the implementation of dynamic TAO, we believe it will become increasingly important for every TAO-holding individual to understand the value produced by each subnet to inform their decisions within the Bittensor Network.

    Our system has four tiers: "Series A", "Seed", "Pre-seed" and "No Funding." We mapped a range of scores to each tier as outlined in the sample histogram below. The final score for each subnet directly determines the tier and therefore the suitable range of emissions for that project.

    Series A$2M - 3M / project
    Seed$1M - 2M / project
    Pre-Seed$100,000 - 1M / project

    The allocation pool for each tier relies on the number of subnets and predetermined funding for a given tier, the current TAO price, and the total stake on the Yuma, a DCG Company's validator as compared to the whole network. For example, if there were 10 subnets in the "Seed" tier, we would disperse a total of $20M amongst these subnets. Individual subnets receive an allocation which reflects the weighted average of their final score as compared to other subnets within the same tier. We incorporate the price of TAO and the percentage of staked funds on our validators to translate the desired funding to weights.

    Scores to Weight Conversion

    Weight assignments are one vehicle to support projects on an annual basis. As we move towards the implementation of dynamic TAO, we believe it will become increasingly important for every TAO-holding individual to understand the value produced by each subnet to inform their decisions within the Bittensor Network.

    Our system has four tiers: "Series A", "Seed", "Pre-seed" and "No Funding." We mapped a range of scores to each tier as outlined in the sample histogram below. The final score for each subnet directly determines the tier and therefore the suitable range of emissions for that project.

    Series A$2M - 3M / project
    Seed$1M - 2M / project
    Pre-Seed$100,000 - 1M / project

    The allocation pool for each tier relies on the number of subnets and predetermined funding for a given tier, the current TAO price, and the total stake on the Yuma, a DCG Company's validator as compared to the whole network. For example, if there were 10 subnets in the "Seed" tier, we would disperse a total of $20M amongst these subnets. Individual subnets receive an allocation which reflects the weighted average of their final score as compared to other subnets within the same tier. We incorporate the price of TAO and the percentage of staked funds on our validators to translate the desired funding to weights.