Sportstensor (Subnet 41 on Bittensor) is redefining the economics of prediction markets. In an in-depth discussion with Ventura Labs, Stephen Rothwell, the Head of AI at Sportstensor, shared how the project is transforming sports betting and prediction markets through data, incentives, and decentralized intelligence.
Check the full discussion below:
The Broken Betting Market
Traditional betting markets are notoriously hostile to smart players. According to Rothwell, anyone who consistently wins or demonstrates predictive skill is quickly banned or restricted by centralized bookmakers.
“These bookmakers aren’t in the business of paying out to smart bettors,” he explains. “They only want the losing money. That’s what keeps their business profitable.”
This creates an uneven playing field for data scientists, quant funds, and machine learning experts who provide valuable “signal” insights that improve price discovery.
Partnering with Polymarket: Betting Without Friction
To solve this, Sportstensor partnered with Polymarket, a fast-growing onchain prediction platform known for its zero-fee model.
Unlike traditional bookmakers who charge a “vig” or house fee (typically around 7 to 9%), Polymarket allows users to trade outcomes freely, turning prediction markets into transparent systems of collective intelligence.
Rothwell explains, “When you remove the fee, you remove the incentive to ban winners. Now, both smart and casual bettors can contribute to shaping the truth of markets.”
Polymarket earns revenue not through fees, but by selling high-quality market data to institutions, investors, and government partners. This is a radically different model that rewards information, not losses.
Sportstensor’s Role: The Meta-Market for Truth
Sportstensor sits on top of Polymarket as what Rothwell calls a “meta market.”
While Polymarket ensures price efficiency (meaning the odds reflect what participants are betting), Sportstensor ensures signal efficiency, measuring how accurate those predictions actually are over time.
In essence, Sportstensor turns prediction accuracy into a measurable, tradable asset.
It tracks the performance of signal providers (miners) across thousands of markets, rewards consistent accuracy with $TAO emissions, and creates a new layer of intelligence for decentralized prediction systems.
Building a Store of Value Through Signal
Subnet 41 operates on a simple but powerful principle: the better your signal, the more valuable your stake.
Sports Tensor charges a 1% fee on bets routed through its dApp (a fraction of traditional costs) and then recirculates 100% of those fees back into its token system.
This feedback loop transforms the subnet into a store of value mechanism.
Bettors who generate consistent returns not only earn rewards in the short term but also hold tokens that grow in value as more participants join and improve the network.
“If you lose your edge,” Rothwell explains, “your token still earns from others who gain one. You’re storing the future value of intelligence itself.”
A Principled Incentive Mechanism
At the heart of Sportstensor’s design is what Rothwell calls a “principled” approach to emissions.
Subnet 41’s reward structure is built on convex optimization, a mathematical framework that guarantees fair and optimal distribution of emissions, thereby, preventing waste, gaming, and dilution.
Rather than rewarding one winner, Sportstensor distributes rewards across multiple high-performing miners, ensuring long-term sustainability and signal diversity.
“We don’t want to leak $TAO to bad actors,” Rothwell says. “Emissions should only go to those creating real market intelligence.”
This design allows the subnet to grow without inflating or diluting token value which is a key problem across many AI and DeFi systems.
Why Data Is the Real Edge
In sports and political betting, Rothwell notes, data scarcity is the biggest challenge.
While AI thrives on large datasets, most sports markets only have around 8,000 usable data points, far too small for deep learning systems to function efficiently.
That’s why Sportstensor is actively sourcing proprietary datasets (from e-sports to live in-game statistics) and plans to make them available to its miners through APIs.
“We’re leveling the playing field,” Rothwell says. “The little guy should also have access to high-quality data.”
This includes potential partnerships with major e-sports data providers and analytics platforms that can offer real-time feeds directly from game servers.
The Bigger Picture: Training the Truth
When asked what the ultimate goal of Sportstensor is, Rothwell’s answer was clear:
“Our goal isn’t just to make the perfect betting line. It’s to train the truth.”
By aggregating thousands of decentralized predictions and rewarding accuracy over time, Sportstensor is building a self-improving system of collective intelligence. This system has the potential to redefine how we price risk, predict events, and understand probability itself.
In his words, “We’re training prediction markets the way AI trains on data, except our data is human intelligence.”
A $2.4 Trillion Opportunity
The global betting industry processes over $2.4 trillion in annual volume, with 66% of that flowing through regulated markets.
If Sportstensor can capture even a fraction of that by routing smart money through on-chain systems, it could generate millions in sustainable yield for Bittensor’s ecosystem while creating a new class of AI-powered prediction economies.
Rothwell estimates that onboarding just three institutional-scale betting desks could produce over $10 million in annual fee volume, all of which cycles back into TAO’s decentralized economy.
Key Lessons for Subnet Builders
In closing, Rothwell offered advice for other Bittensor subnet developers:
“Be principled. Build constraints into your incentive mechanism. Don’t leak $TAO without value creation.”
He emphasizes that the most sustainable subnets are those that treat emissions like a business budget; ensuring that every token emitted generates real work, intelligence, or insight.
Conclusion
Sportstensor isn’t just a betting platform. It’s a new kind of AI-economic engine that rewards intelligence, not speculation.
By combining decentralized incentives, transparent data, and onchain prediction markets, Subnet 41 is pioneering how truth itself can be mined, priced, and stored on the Bittensor network.
Useful Links
Official Website: https://www.sportstensor.com/
X (Formerly Twitter): https://x.com/sportstensor
GitHub: https://github.com/sportstensor/sportstensor
Whitepaper: https://docs.sportstensor.com/mining-comprehensive-guide
Be the first to comment