Bettensor is a sports prediction subnet operating on Bittensor. The Subnet framework rewards individuals and models that accurately predict sports outcomes using various methods, including human intuition, intelligence, AI prediction models, statistical models, and arbitrage strategies. The goal of Bettensor is to provide a platform for sports fans to predict the outcomes of their favorite sporting events, and ML/AI researchers to develop new models and strategies to benchmark against good, old-fashioned human intelligence and intuition.

The goal is to establish a highly competitive, global marketplace for sports outcome predictions and odds generation. In this competitive environment, the subnet’s output data could generate alpha, providing opportunities for monetization and value creation within the Bittensor ecosystem.

Bettensor is a software application developed on the Bittensor network and operates independently of their control. They are not a financial institution and do not facilitate any monetary transactions directly. Bettensor does not distribute payments or collect money, whether in cryptocurrencies or fiat currencies. The core principle of Bettensor is decentralization. This strategy enables the platform to distribute and decentralize machine learning tasks across multiple nodes within the Bittensor network. Their primary focus is to enhance sports price prediction while incentivizing the provision of accurate data.

Users submit their predictions, and through the power of collective intelligence, Bettensor refines these into highly accurate forecasts. Contributors to Bettensor are rewarded based on the accuracy of their predictions. The more accurate your predictions, the greater your rewards. Rewards or payments are issued by the Bittensor network according to its established protocols. Bettensor’s role is limited to providing software that suggests the allocation of $TAO tokens based on users’ predictions. They do not influence or control the Bittensor network’s decision-making process regarding the distribution of rewards.

Mining

Mining on their network is straightforward, involving just the submission of predictions. In this Beta version of BetTensor, they provide a “Fallout Terminal” interface where you can view available games, select the team you believe will win, and enter a wager amount for that game. They are currently developing a web app interface to make predictions submission more intuitive, and this feature will be included in the first official update.

On BetTensor, users don’t wager real money; they receive 1000 points daily to distribute as they wish. Users can place small bets on every game or stake all points on their favorite team. At the end of each day, performance is evaluated based on one simple metric: the total points won. Odds for each game are constantly updated via external APIs from popular sports betting platforms. Points are awarded based on the game outcome and the odds at the time of submission.

Validating

Running a validator on their network is also straightforward. Validators need an API key for the games data and one for their dashboard. While you can opt out of sending data to their dashboard, they hope you won’t, as better data benefits everyone. Validators score miners by ranking them based on the points earned for successful predictions. The compute requirements are minimal, needing only enough capacity to handle the traffic.

Incentive Mechanism and Scoring

Their current scoring system is quite simple. Miners receive a simulated daily balance of $1000, reset daily at 00:00 UTC. They can select games to predict on by placing a simulated moneyline wager. Odds for the wager are frequently updated from sportsbook APIs and recorded by the validator upon submission of a prediction.

When a game concludes, the outcome of the simulated wager is calculated using the odds recorded at the time of submission. If the miner wins, they receive an “earnings balance” equal to the simulated wager amount multiplied by the odds of their prediction at the time of submission. Losses do not count against the earnings balance; only wins affect it.

Validators score miners based on a rolling 2-day period of earnings, with rewards being linearly proportional to earnings. This design heavily incentivizes correct predictions. The risk of deregistration is high with just one bad day of predictions, but the potential for high rewards is also there for those with accurate insights.

Awaiting data