Sportstensor is a decentralized artificial intelligence (AI) network designed for sports prediction that operates as Subnet 41 on the Bittensor blockchain. The project aims to create a competitive environment where developers, data scientists, and AI enthusiasts, referred to as "miners," can build, deploy, and monetize predictive models. By harnessing the principles of collective intelligence and ensemble modeling, Sportstensor aggregates numerous independent forecasts into a single, more accurate "meta-model" to discover and exploit persistent, predictable patterns in sports betting markets. ^{[1] [2]}​

Overview

Sportstensor addresses the high barrier to entry in the field of algorithmic sports prediction, an industry historically dominated by specialized quantitative firms with significant computational and financial resources. The project's core premise is that sports competitions function as "mathematical proving grounds" where, despite apparent chaos, exploitable patterns emerge. It provides a decentralized platform where individual developers can compete to create superior predictive models while protecting their intellectual property. ^{[3] [4]}

The network operates as a continuous, real-time algorithmic contest. Miners submit predictions from their proprietary models, which are then evaluated by a network of independent validators against real-world outcomes. Performance is rewarded with the network's native token, SN41, creating a direct financial incentive for continuous model improvement and innovation. ^{[5] [6]} The ultimate goal is to create a collective intelligence that outperforms any single contributing model and traditional market odds, a concept the project summarizes with the slogan, "Beating the odds through collective intelligence." ^[7]

History

The development of Sportstensor began in early 2024, with the initial commit to its public GitHub repository occurring on March 31, 2024. ^[6] The project established its social media presence with the creation of its official X (formerly Twitter) account in June 2024 and officially launched as a Bittensor subnet on July 15, 2024. ^[7] The foundational phase of the project took place in the third quarter of 2024, which included its launch on the Bittensor testnet (Subnet 172) and the development of baseline prediction models for Major League Soccer (MLS) and Major League Baseball (MLB). ^[8]​

A significant technical overhaul, "Incentive v2," was implemented on October 11, 2024, which introduced key features such as league commitment handling for miners and a scoring system based on "closing edge" value against market odds. ^[6] By February 2025, the project's treasury demonstrated early success, generating $17,842 in trading profits from its intelligence models, representing a 178.42% return on capital for the month. These profits were subsequently reinvested into the project's native token. ^[7]​

Sportstensor gained institutional recognition in April 2025 when it was featured in a research report on Bittensor published by Grayscale and authored by FS Insight. ^[7] Development continued throughout 2025, with updates to support new seasons for leagues like the English Premier League and to maintain compatibility with the evolving Bittensor protocol. On September 11, 2025, the project released another major update to its incentive mechanism, "IM v2.5 Sortino," which integrated the Sortino Ratio and Profit and Loss (PnL)-based calculations to better reward consistent, risk-managed performance. ^[6]​

On September 14, 2025, Sportstensor announced a strategic partnership with the prediction market platform Polymarket to expand the accessibility of sports prediction markets and develop new liquidity-enhancing methods. ^[7] The project also reported progress on its computer vision technology in October 2025, noting a focus on building an improved dataset for model training. ^[7]​

Technology and Architecture

Sportstensor's technical foundation is built upon the Bittensor network, where it operates as a specialized, self-contained economic market for sports intelligence. ^[3]​

Bittensor Subnet Foundation

As Subnet 41 (SN41) on the Bittensor mainnet, Sportstensor leverages the parent network's decentralized infrastructure for peer-to-peer machine learning. This allows it to utilize Bittensor's core functions for incentive distribution, consensus, and secure communication between network participants without needing to build a blockchain from scratch. The project also uses Subnet 172 for testing purposes. ^{[6] [5]}​

Miner-Validator Mechanism

The network's architecture is composed of two primary roles: miners and validators. This structure creates a "digital survival-of-the-fittest" environment where performance is directly tied to rewards. ^[3]​

• Miners are the AI developers and data scientists who create, deploy, and operate the predictive models. They receive computational tasks (prediction requests for specific matches) from validators, execute them using their proprietary models and data, and return the results. This system allows miners to monetize their models while keeping their underlying code and intellectual property private. ^{[3] [6]}
• Validators are responsible for orchestrating the network and maintaining its integrity. They generate prediction tasks for miners, assess the quality and accuracy of the submitted predictions against actual game outcomes, and allocate token rewards based on performance. The use of multiple independent validators is designed to ensure transparent performance tracking and make cheating difficult. ^{[3] [1]}

Scoring and Incentives

The incentive mechanism is designed to reward "genuine intelligence, not just lucky streaks." ^[3] The primary metric for evaluating miner performance is "closing edge," which measures a prediction's value against the final, most efficient market odds available before a match begins. This encourages models that can find true alpha rather than simply predicting obvious outcomes. The "IM v2.5 Sortino" update further refined this by introducing the Sortino Ratio, a risk-adjusted metric that penalizes downside volatility more heavily than upside volatility, rewarding miners for generating consistent, stable returns. ^[6]​

Sportstensor Meta Model (STMM)

The STMM is an advanced ensemble architecture that aggregates the predictions from the entire network of competing miner models. The core principle is that a synthesized forecast from a diverse set of intelligent models will be more accurate and robust than any single model. A graph included in the April 2025 Grayscale report demonstrated the STMM significantly outperforming the NBA market over a three-month period. ^[7]​

Computer Vision Initiative

Sportstensor is actively developing computer vision (CV) technology to democratize and annotate sports video data. The objective is to analyze video footage to extract novel insights and granular data points, such as frame-level positional data for players and the ball. This enriched data can then be used by miners to train more sophisticated and accurate predictive models. ^{[7] [8]}​

Products and Ecosystem

Sportstensor's technology powers a growing ecosystem of products and serves various audiences within the sports analytics and decentralized finance sectors.

Almanac

Almanac is a prediction-market layer built on top of the Polymarket platform. It functions as the primary vehicle for monetizing the collective intelligence of the Sportstensor network. Almanac aggregates the highest-quality predictions from Sportstensor miners into its meta-model and uses these signals to trade on Polymarket with its own bankroll. It charges a 1% fee on winning trades, and the revenue generated is used to buy back the subnet's SN41 token, with the potential to burn the tokens to reduce supply. ^[8]​

Supported Markets and Data Consumers

The platform is designed to be model-agnostic, supporting various approaches like neural networks and simulations. It covers major global sports leagues, including the National Football League (NFL), National Basketball Association (NBA), Major League Baseball (MLB), National Hockey League (NHL), English Premier League (EPL), and Major League Soccer (MLS). ^{[4] [6]}​

The predictive data generated by the network is consumed by various entities. One notable example is BillyBets, an autonomous AI sports betting agent that operates by analyzing real-time data directly from Sportstensor to inform its strategies. ^[1]​

Future Developments

The project's roadmap includes the development of a proprietary Large Language Model (LLM) chatbot linked to miner predictions to provide conversational insights. Other planned initiatives include the expansion of the computer-vision pipelines and the exploration of a hedge fund that would trade on Almanac's aggregated, positive expected value signals. ^[8]​

Tokenomics (SN41)

The native utility token of the Sportstensor subnet is the SN41 Alpha Token, identified by the ticker SN41. It is a derivative of Bittensor's primary cryptocurrency, TAO, and is central to the network's incentive mechanism. ^[3]​

• Ticker: SN41
• Network: Bittensor
• Max Supply: 21,000,000 SN41
• Total Supply (as of Nov 6, 2025): 3,085,453 SN41
• Circulating Supply (as of Nov 6, 2025): 3,085,453 SN41

The primary utility of SN41 is to reward miners and validators for their valuable contributions to the network's intelligence. Miners who consistently deliver accurate, high-edge predictions receive a greater share of token emissions. ^[3] The token's value is also supported by the project's treasury strategy, which reinvests profits from its trading activities back into the token. Furthermore, the fee mechanism of the Almanac product creates a continuous buy-back pressure on the open market. ^{[7] [8]}​

The token's price history includes an all-time low of 16.73 on November 2, 2025. ^[5]​

Partnerships and Recognition

Sportstensor has formed key relationships to enhance its ecosystem and gain credibility within the broader AI and blockchain industries.

• Polymarket: The project has a strategic partnership with Polymarket, one of the largest prediction market platforms. This collaboration aims to enhance liquidity and pioneer new incentive structures. The partnership was formalized after the Sportstensor team had already driven significant trading volume to the platform, demonstrating the value of its predictive signals. Sportstensor's Almanac product is built as a layer on top of Polymarket. ^{[7] [8]}
• Grayscale / FS Insight: In April 2025, Sportstensor was featured in the "Bittensor: The Internet of AI" research report from Grayscale and FS Insight. This inclusion provided the project with significant institutional recognition and validated its position within the Bittensor ecosystem. ^[7]

Team

Sportstensor was founded by Leo Lucian (X handle: ). The team also includes a member known as Stephen (handle: Neuromancer). As of October 2025, the core team consisted of five people. ^[8]​

The project also acknowledges contributions from the broader community. Notable contributors mentioned include for work on the computer vision technology and GitHub users such as xzistance, conidig, and Cardoso-topdev for their activity in the project's open-source repository. ^{[7] [6]}​