Subnet 79: MVTRX
MVTRX is Bittensor Subnet 79. Its on-chain description is “Building a SOTA Exchange for dTAO and Beyond,” and the project site is taos.im. Rather than running a live exchange today, the subnet operates a detailed simulation of financial markets and rewards miners for building trading strategies that perform well inside it. The data this generates is aimed at market research and AI model training, with live-exchange features planned for later. The MVTRX README source describes the live subnet as two connected components: the τaos market simulation and the GenTRX training layer.
What MVTRX Rewards
At the center of the subnet is an agent-based market simulator: many automated participants trade in a realistic order book, reproducing the fine detail of how real markets behave. Miners take part as trading agents in that simulation, and they are rewarded for risk-adjusted performance — making sound, well-managed trades across many simulated market conditions rather than getting lucky once. Because the trading happens in simulation, the goal is high-quality strategy and the valuable market datasets it produces, not real-money returns.
To keep the output meaningful, the design rewards active, intelligent trading. A miner has to keep trading at a reasonable level to be fully rewarded, which rules out passive buy-and-hold approaches, and faster, more efficient responses are advantaged because slow instructions are executed later in the simulated market. The subnet also includes an optional second component in which miners can help train a shared AI model built on top of the simulation data, earning additional rewards for contributing useful training work.
Simulation Data Context
MVTRX treats the simulator as an agent-based market environment. Miner agents are not scored against a static prompt; they act inside simulated order books where their orders can interact with the background market and with other participants. The useful output is therefore both miner performance information and market-behavior data generated by the simulation.
Detailed order-book data matters because high-frequency market research often needs more than candle prices or end-of-day returns. A simulated market that records individual order-book events can support strategy research, surveillance-style analysis, and model training on market microstructure. This is why the subnet’s evaluation is tied to active trading behavior rather than a single prediction about price direction.
GenTRX Context
GenTRX is the training component built on top of simulation data. In that track, miners can contribute training work for a shared order-book generative model, while the core trading pool continues to reward performance in the simulated market. The two components are connected: simulation produces the data, and GenTRX uses that data to improve a model of order-book behavior.
For readers, this makes MVTRX more than a trading-game subnet. The trading competition creates market data and tests strategy agents, while the training component turns that data into model improvement work. Both parts support the broader goal of building exchange and market-intelligence infrastructure.
Miner and Validator Roles
Miners develop and run trading-strategy agents. Each agent receives the current state of the simulated market and responds with orders to place or cancel, aiming to maximize risk-adjusted performance over many market scenarios. A registered hotkey on netuid 79 is what ties a miner’s agent and its results back to them, and miners may optionally extend their agent to take part in the shared model-training component.
Validators run the market simulation itself. They maintain the simulator, push the evolving market state out to miners, feed the miners’ trading instructions back into the simulation for execution, and score each agent on how well it performed. From those scores they set weights on the network, which is how reward flows through Yuma Consensus. At the article level the split is straightforward: miners supply the trading strategies, while validators run the market and weight each miner by how well its agent performed.
Source and Live Data
The codebase is maintained in the taos-im/sn-79 repository. Live SN79 data is available on TaoStats. The mechanism details in this article are tied to the public README rather than to live identity fields.
Relationship to Yuma Consensus
Subnet 79 uses Yuma Consensus to convert the trading-performance weight vectors that validators submit into the emission shares distributed to miners and validators within the subnet each tempo. The Yuma Consensus documentation describes how validator weight submissions are aggregated into consensus weights for each miner registered on the subnet.
In MVTRX’s context, validators maintain the τaos market simulator, push evolving market state to miners, execute miners’ trading instructions inside the simulation, and score each agent on risk-adjusted performance across many simulated market scenarios. Miners that trade actively and efficiently score higher than passive or slow strategies. Validators submit weight vectors reflecting those per-agent performance scores. The Emission documentation describes how those consensus weights determine each participant’s share of the subnet’s accumulated emission each tempo.
Development Stage Context
The Introduction to Bittensor describes subnet development as moving from localnet to testnet and then mainnet. For Subnet 79, that sequence applies to the standard Bittensor lifecycle: localnet for isolated development, testnet for shared non-production testing, and mainnet for live operation with real emissions.
On mainnet, Subnet 79 is registered as the live production subnet at netuid 79. The Bittensor Networks reference separates mainnet, testnet, and localnet. Participation examples or emission outcomes from one environment should not be read as representing production subnet performance in another environment.
Reader Boundary
Subnet 79 MVTRX should not be read as generic Bittensor subnet documentation, proof of live exchange trading profits, or a substitute for the project’s simulator documentation. The public MVTRX README source describes miners as trading agents inside an agent-based market simulation scored on risk-adjusted performance metrics.
Simulation Scores Reflect In-Simulator Performance
The README evaluates miners on intraday Kappa-3-style risk-adjusted measures together with a requirement to maintain sufficient cumulative round-trip trading volume. Miners must also respond within a reasonable timeframe because slower responses are executed later in the simulated market. Readers should treat rewarded performance as strategy quality inside the τaos simulation environment, not as audited returns from a production trading venue.
Latency and Market Impact Stay Inside the Simulator
Miner instructions are processed like orders on a simulated exchange, interacting with background agents and other miners so market impact is part of the score. That design supports research and strategy development on detailed order-book data, but it does not by itself establish that any strategy would perform the same way outside the subnet’s simulation parameters.
Validator weights still flow through Yuma Consensus to determine emissions each tempo (Yuma Consensus, Emission).