{"id":"score","title":"Score","content":"**Score** is a decentralized computer vision platform and a vision AI research company built as Subnet 44 on the Bittensor network. The project's core mission is to \"make every camera intelligent\" by creating an open, permissionless infrastructure that transforms raw video into structured, actionable data. It combines a proprietary \"Programmable Vision\" framework with a decentralized network of miners and validators to create and evaluate adaptable AI models for visual analysis. Initially focused on sports analytics, Score has expanded its scope to serve various enterprise sectors, including retail, agriculture, and logistics. [\\[1\\]](#cite-id-wPwFEpM3B14ksggf) [\\[2\\]](#cite-id-IzTp6QIPJAO8RpWQ)​\n\n[YOUTUBE@VID](https://youtube.com/watch?v=xFUuG_ZjD2w)\n\n## Overview\n\nScore addresses the high cost and inefficiency of traditional video analysis, which often relies on time-consuming manual annotation. [\\[3\\]](#cite-id-nX7x9XNfP8JY5O5Y) By leveraging the [Bittensor](https://iq.wiki/wiki/bittensor-tao) network, Score created a competitive marketplace where a global network of contributors (miners) process video data using their own machine learning models. These models are then evaluated by [validators](https://iq.wiki/wiki/validator), creating a self-improving ecosystem that aims to reduce video analysis costs by a factor of 10 to 100 while increasing speed and accuracy. The project posits that this decentralized approach democratizes access to advanced computer vision capabilities, allowing developers and businesses to build and deploy vision systems without needing specialized AI/ML teams. [\\[4\\]](#cite-id-ZSwfoHDYbIyaV1PM) [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)​\n\nThe project began with a focus on sports, using the complex, multi-agent environment of football as a training ground for its foundational models. This strategy was based on the premise that a system capable of understanding the chaotic and nuanced activity of a live sport could be generalized to other real-world scenarios. [\\[1\\]](#cite-id-wPwFEpM3B14ksggf) Since its inception, the company has broadened its vision to become a generalized vision layer for the physical world, developing commercial applications for multiple industries. The project's suite of products is designed to make this technology accessible, most notably through Manako, a vision AI assistant that allows users to create custom vision systems using simple text prompts. [\\[2\\]](#cite-id-IzTp6QIPJAO8RpWQ)​\n\n## History and Development\n\nScore was founded in New York City in 2024 by Maxime Sebti, Tim Kalic, and Nigel Grant. Initially conceived as a \"Moneyball\" protocol, the company's first focus was on sports analytics. During this period, it established a key partnership with Reading Football Club to provide recruitment analytics, demonstrating the technology's potential in professional sports. [\\[2\\]](#cite-id-IzTp6QIPJAO8RpWQ)​\n\nTowards the end of 2024, the company pivoted its strategy, expanding from a sports-specific protocol to a more generalized vision layer applicable to a wide range of industries. This shift marked a significant evolution in the project's ambition. In 2025, co-founder Maxime Sebti attended a private side forum at Davos focused on AI and data center infrastructure, participating in discussions with representatives from the Opentensor [Foundation](https://iq.wiki/wiki/foundation) and other industry leaders. [\\[2\\]](#cite-id-IzTp6QIPJAO8RpWQ)​\n\nThe project launched its [mainnet](https://iq.wiki/wiki/mainnet) as Subnet 44 on the [Bittensor](https://iq.wiki/wiki/bittensor-tao) network in the first quarter of 2025. This deployment was a critical milestone, transitioning the project from a centralized research and development entity to a decentralized AI network powered by a global community of contributors. The roadmap following the mainnet launch included plans to enhance the platform with features like \"action spotting\" for identifying specific events, match event captioning, and the release of integration APIs for third-party developers. [\\[3\\]](#cite-id-nX7x9XNfP8JY5O5Y)​\n\n## Technology and Architecture\n\nScore operates on [Bittensor's](https://iq.wiki/wiki/bittensor-tao) Subnet 44, leveraging the network's decentralized structure to create, validate, and improve computer vision models continuously. [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)​\n\n### Network Participants\n\nThe ecosystem functions through the interaction of key participants who are incentivized with [Bittensor's](https://iq.wiki/wiki/bittensor-tao) native token, TAO:\n\n* **Miners:** These are network participants who contribute computational resources to process and analyze video streams. They run their own computer vision models to perform tasks like object detection, tracking, and annotation on video clips provided by the network. Miners are rewarded in TAO based on the accuracy, consistency, and speed of their work. [\\[4\\]](#cite-id-ZSwfoHDYbIyaV1PM)\n* **Validators:** [Validators](https://iq.wiki/wiki/validator) are responsible for securing the network and ensuring the quality of the work submitted by miners. They create challenges, such as providing video data for analysis, and then evaluate the miners' outputs using proprietary validation protocols. Validators earn TAO rewards for accurately assessing miner performance and contributing to the network's consensus. [\\[3\\]](#cite-id-nX7x9XNfP8JY5O5Y)\n* **Subnet Owner:** The owner of Subnet 44, the Score team, manages the overall health and direction of the subnet. This includes optimizing network parameters and adjusting incentive mechanisms to encourage high-quality contributions. [\\[3\\]](#cite-id-nX7x9XNfP8JY5O5Y)\n\n### Core Concepts and Innovations\n\nScore's platform is built on several key technological innovations:\n\n* **Programmable Vision:** Instead of developing large, monolithic models for single tasks, Score creates a library of small, specialized \"vision skills\" (e.g., \"detect person,\" \"track ball\"). These skills act as building blocks that can be programmatically combined to create complex, customized vision systems for new applications without requiring retraining from scratch. [\\[1\\]](#cite-id-wPwFEpM3B14ksggf)\n* **Turbo Vision:** This system uses advanced vision-language models (VLMs) to automatically generate labels for a small sample of video frames. This process creates a \"pseudo ground truth\" dataset that is then used to train smaller, more efficient models to process entire videos at high speed and low cost. [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)\n* **Lightweight Validation:** To minimize computational overhead for validators, Score developed a two-step validation method. It first filters irrelevant video frames and validates the accuracy of keypoints (like player positions). Then, for key frames, it uses CLIP-based object checks to verify the classification of detected objects, ensuring a high degree of accuracy without processing every single frame intensely. [\\[3\\]](#cite-id-nX7x9XNfP8JY5O5Y)\n* **Dual-Track System:** The platform operates two parallel tracks to serve different needs. A Public Track offers a transparent environment where models are publicly available. A Private Track is designed for enterprise clients with sensitive data, running computations inside [Trusted Execution Environments](https://iq.wiki/wiki/trusted-execution-environments) (TEEs) to ensure data privacy. [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)\n\n### Annotation and Incentive Pipeline\n\nThe process of turning raw video into data follows a structured pipeline. Full video files are programmatically sliced into 30-second clips and distributed to miners. Miners process the clips, returning structured data in a JSON format. Validators then perform semantic and geometric checks to verify accuracy. Miners who provide accurate submissions are scored and rewarded, with a \"streak multiplier\" that increases payouts for consistently high performance. The validated clips are then aggregated to create a complete, data-rich timeline of the original video. [\\[4\\]](#cite-id-ZSwfoHDYbIyaV1PM) [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)​\n\n## Products and Ecosystem\n\nScore's technology is delivered through a suite of products and specialized entities catering to different markets.\n\n### Core Products\n\n* **Manako:** The flagship application layer, Manako is a vision AI assistant with a no-code interface. Users can connect a camera feed and use simple text prompts to define an analysis task. Manako then automatically selects and combines the necessary vision skills from Score's library to create a functional vision system. [\\[1\\]](#cite-id-wPwFEpM3B14ksggf) [\\[2\\]](#cite-id-IzTp6QIPJAO8RpWQ)\n* **AVIA:** A commercial solution for the retail fuel and convenience store sector. It uses a facility's existing cameras to provide real-time intelligence on customer behavior, workflow efficiency, and security alerts for anomalies like spills or unauthorized entry. [\\[1\\]](#cite-id-wPwFEpM3B14ksggf)\n* **Two-a-Day:** A solution designed for agricultural and food production environments. It analyzes camera feeds on production lines to monitor quality control, identify bottlenecks, and reduce waste. [\\[1\\]](#cite-id-wPwFEpM3B14ksggf)\n\n### Consumer Applications\n\n* **Score App:** A mobile application that allows football fans to act as \"human miners\" by making match predictions. Users earn points for correct predictions and can win rewards, including TAO. [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)\n* **Fantasy Sports App:** The project plans to launch a fantasy sports application powered by the subnet's deep analytical data, with a target launch around the 2026 World Cup. [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)\n\n### Ecosystem Entities\n\n* **Mettle:** A specialized branch focusing on sports analytics, which powers partnerships like the one with Reading F.C. [\\[1\\]](#cite-id-wPwFEpM3B14ksggf)\n* **Lavance:** An application for the automotive services industry, used to monitor operations at car and heavy-vehicle wash stations. [\\[2\\]](#cite-id-IzTp6QIPJAO8RpWQ)\n* **Sire:** An [AI agent](https://iq.wiki/wiki/ai-agents) developed to participate in prediction markets, leveraging Score's analytical capabilities. [\\[1\\]](#cite-id-wPwFEpM3B14ksggf)\n\n## Use Cases and Applications\n\nThe structured data generated by Score's platform enables a wide range of applications across various industries.\n\n### Sports Analytics\n\nThe initial and foundational use case for Score is sports. The platform provides data for:\n\n* **Football:** Real-time tactical analysis, objective player performance evaluation, and scouting across more than 280 leagues. This data is used by clubs, coaches, and sports betting syndicates. The system is designed to generate insights like expected goals (xG) and team pressing intensity. [\\[4\\]](#cite-id-ZSwfoHDYbIyaV1PM)\n* **Cricket:** The platform is expanding into cricket, with the goal of replicating and improving upon Hawkeye-style ball-tracking using standard broadcast footage for use in scouting and tactical development. [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)\n\n### Enterprise Solutions\n\nScore has expanded its technology to serve a variety of enterprise needs:\n\n* **Retail and Safety:** Score has a contract with a major European petroleum company to monitor and send alerts across thousands of automated gas stations. [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)\n* **Agriculture:** Through its partnership with Two-a-Day Group, a major fruit exporter, Score's technology is used to enhance fruit sorting, grading, and quality analysis on production lines. [\\[2\\]](#cite-id-IzTp6QIPJAO8RpWQ)\n* **Logistics and Insurance:** The technology is used to automatically generate home-move inventory lists by analyzing video walkthroughs of properties, streamlining processes for logistics and insurance companies. [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)\n* **Security:** The platform can enhance home and CCTV security systems by providing more intelligent and accurate alerts. [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)\n\n## Tokenomics\n\nScore's native subnet token is SN44. As a subnet on the [Bittensor](https://iq.wiki/wiki/bittensor-tao) network, its economic model is deeply integrated with Bittensor's primary token, TAO.\n\n* **Ticker:** SN44\n* **Max Supply:** 21,000,000 SN44\n* **Role:** The SN44 token represents a stake or ownership share in Score's specific intelligence market (Subnet 44) within the broader Bittensor ecosystem.\n* **Incentives:** While SN44 is the native token, the primary rewards distributed to miners and validators for their contributions on the subnet are paid in TAO. [\\[6\\]](#cite-id-8IaEMuwcjrUXMBEF) [\\[4\\]](#cite-id-ZSwfoHDYbIyaV1PM)\n\nThe Score team is developing a formal economic model called \"Alphanomics,\" which aims to directly link the project's real-world enterprise revenue back to the Bittensor ecosystem. The project has stated its intention to use a portion of its revenue (approximately 20%) to execute [TAO](https://iq.wiki/wiki/bittensor-tao) buybacks and burns, a mechanism designed to create value for the network and its token holders. [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)​\n\n## Team and Key Personnel\n\n* **Maxime Sebti:** Co-founder & CEO [\\[2\\]](#cite-id-IzTp6QIPJAO8RpWQ)\n* **Tim Kalic:** Co-founder & CTO [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)\n* **Nigel Grant:** Co-founder & CSO [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)\n* **Matthew Cyzer:** Chairman. Cyzer has over 35 years of experience in global finance, having been a Partner at Goldman Sachs and holding senior roles at BTIG, Cowen, and Toronto Dominion Bank. [\\[2\\]](#cite-id-IzTp6QIPJAO8RpWQ)\n* **Other Team Members:** Other key members include Peter Cotton, Nick Compton (former professional cricketer), Arnaud Deffuant, Jack Devlin (Marketing Manager), Gaetan Lajeune (Validator & Miner Manager), and Alfie Grant (Football Analyst). [\\[2\\]](#cite-id-IzTp6QIPJAO8RpWQ) [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)\n\n## Partnerships\n\nScore has established a number of strategic partnerships across its target industries:\n\n* **Reading Football Club:** An early partner for football recruitment analytics. [\\[2\\]](#cite-id-IzTp6QIPJAO8RpWQ)\n* **Two-a-Day Group (Pty) Limited:** A major African fruit exporter using Score's technology for quality analysis in agriculture. [\\[2\\]](#cite-id-IzTp6QIPJAO8RpWQ)\n* **Lavance:** A partner in the automotive services industry, deploying Vision AI in car wash stations across France. [\\[2\\]](#cite-id-IzTp6QIPJAO8RpWQ)\n* **European Petroleum Company:** A signed contract with an unnamed major European petroleum company for safety and monitoring at gas stations. [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)\n* **Sports Betting Syndicate:** A strategic partnership with an unnamed $5 billion sports betting syndicate to provide predictive analytics. [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)\n* **Sports Data Provider:** A partnership to access video footage from 283 leagues, covering over 400,000 matches. [\\[5\\]](#cite-id-yz6qnbJELSYAOHFg)","summary":"Built on Bittensor's Subnet 44, Score is a decentralized AI computer vision protocol that transforms raw video into structured, actionable data, initially for sports analytics before expanding to broader enterprise applications.","images":[{"id":"QmVreNzKYzUoiHaLbc3ehTDep25PzZUZThEUmmsbBJr3Ec","type":"image/jpeg, image/png"}],"categories":[{"id":"dapps","title":"dapps"}],"tags":[{"id":"AI"},{"id":"Blockchains"},{"id":"Protocols"},{"id":"Developers"},{"id":"Entertainment"}],"media":[{"id":"QmTXckt5BacWaENDmTBLcPdtKisMgf2h1y4YZ7x9B1nDJj","type":"GALLERY","source":"IPFS_IMG"},{"id":"QmbY3RtHRLS3JEbUT23TPettjwn5VqmVQo8SuJkWqNk4KF","type":"GALLERY","source":"IPFS_IMG"},{"id":"QmSr3GUm6hvaoUa4Dq48ibX15YaKTg5YS9D9q7ADdrWxtm","type":"GALLERY","source":"IPFS_IMG"},{"id":"Qma1FhBFoRnHDKHqgDHGpR6KN9SphEDEECQVMoKH3F3iRy","type":"GALLERY","source":"IPFS_IMG"},{"id":"https://www.youtube.com/watch?v=xFUuG_ZjD2w","name":"xFUuG_ZjD2w","caption":"","thumbnail":"https://www.youtube.com/watch?v=xFUuG_ZjD2w","source":"YOUTUBE"},{"id":"QmbrnesmTkeQd3axgzRTadUJVZ4NbT9EbcaWj2ynY65JJz","name":"Cópia de Design sem nome (1).png","caption":"","thumbnail":"QmbrnesmTkeQd3axgzRTadUJVZ4NbT9EbcaWj2ynY65JJz","source":"IPFS_IMG"},{"id":"QmcGrkUeXoTr3KiUsvcuDmYupph7sZaa329TyVXBrN1yUw","name":"jOfQr8PE_400x400.jpg","caption":"","thumbnail":"QmcGrkUeXoTr3KiUsvcuDmYupph7sZaa329TyVXBrN1yUw","source":"IPFS_IMG"}],"metadata":[{"id":"references","value":"[\n  {\n    \"id\": \"wPwFEpM3B14ksggf\",\n    \"url\": \"https://www.wearescore.com/\",\n    \"description\": \"Score official website overview\",\n    \"timestamp\": 1775673557953\n  },\n  {\n    \"id\": \"IzTp6QIPJAO8RpWQ\",\n    \"url\": \"https://www.linkedin.com/company/webuildscore/\",\n    \"description\": \"Company mission and overview on LinkedIn\",\n    \"timestamp\": 1775673557953\n  },\n  {\n    \"id\": \"nX7x9XNfP8JY5O5Y\",\n    \"url\": \"https://github.com/score-technologies/score-vision\",\n    \"description\": \"Score Vision technical paper on GitHub\",\n    \"timestamp\": 1775673557953\n  },\n  {\n    \"id\": \"ZSwfoHDYbIyaV1PM\",\n    \"url\": \"https://medium.com/wearescore/score-deep-dive-from-vision-to-winning-c9d6b4b7de7e\",\n    \"description\": \"Score operational model and cost reduction\",\n    \"timestamp\": 1775673557953\n  },\n  {\n    \"id\": \"yz6qnbJELSYAOHFg\",\n    \"url\": \"https://subnetalpha.ai/subnet/score/\",\n    \"description\": \"Platform overview and cost-effectiveness\",\n    \"timestamp\": 1775673557953\n  },\n  {\n    \"id\": \"8IaEMuwcjrUXMBEF\",\n    \"url\": \"https://coinmarketcap.com/currencies/score/\",\n    \"description\": \"SN44 token supply details\",\n    \"timestamp\": 1775673557953\n  }\n]"},{"id":"website","value":"https://www.wearescore.com/"},{"id":"twitter_profile","value":"https://x.com/webuildscore"},{"id":"linkedin_profile","value":"https://www.linkedin.com/company/webuildscore/"},{"id":"github_profile","value":"https://github.com/score-technologies/score-vision"},{"id":"medium_profile","value":"https://medium.com/wearescore"},{"id":"coinmarketcap_url","value":"https://coinmarketcap.com/currencies/score/"},{"id":"coingecko_profile","value":"https://www.coingecko.com/en/coins/score"},{"id":"contract_url","value":"https://taostats.io/subnets/44/chart"},{"id":"commit-message","value":"\"Added project page for Score\""}],"events":[{"id":"e758f729-f2fd-4edd-adae-b96474dd1b3e","date":"2024-01","title":"Score Founded in New York City","type":"CREATED","description":"Score was founded, starting as a 'Moneyball' protocol focused on sports analytics and expanding to a generalized vision layer.","link":null,"multiDateStart":null,"multiDateEnd":null,"continent":null,"country":null},{"id":"5279e08e-90a1-452d-9b18-cbafa2a96095","date":"2025-01","title":"Mainnet Launch on Bittensor","type":"DEFAULT","description":"Score deployed to the Bittensor mainnet as Subnet 44 (SN44), transitioning from a centralized R&D company to a decentralized AI network.","link":null,"multiDateStart":null,"multiDateEnd":null,"continent":null,"country":null},{"id":"2b1c9eef-e98c-4057-8f4c-e07630a1677b","date":"2024-01","title":"Reading F.C. Partnership","type":"DEFAULT","description":"Score announced a partnership with Reading Football Club to provide football recruitment analytics, demonstrating its Mettle sports analytics branch.","link":null,"multiDateStart":null,"multiDateEnd":null,"continent":null,"country":null},{"id":"dcefc463-2948-49b0-a5f9-8a3f03ca8a39","date":"2025-01","title":"Participation in Davos Forum","type":"DEFAULT","description":"Co-founder Maxime Sebti attended a private side forum at Davos 2025 focused on AI and data centre infrastructure, alongside other industry leaders.","link":null,"multiDateStart":null,"multiDateEnd":null,"continent":null,"country":null}],"user":{"id":"0x8af7a19a26d8fbc48defb35aefb15ec8c407f889"},"author":{"id":"0x8af7a19a26d8fbc48defb35aefb15ec8c407f889"},"operator":{"id":"0xd5893989b9952c6568a99F854795AcC5Ae480D56"},"language":"en","version":1,"linkedWikis":{"blockchains":["bittensor-tao"],"founders":[],"speakers":[]}}