Synth
SN50Generates synthetic financial data for predicting market movements
Synth is trying to make market forecasting more useful for machines by turning price prediction into a competition over full probability distributions, not one-off calls.
// Synthetic price paths for market AI.
Synth tackles a specific problem inside the Bittensor ecosystem: Traders and AI agents often get a single forecast when what they really need is a range of plausible outcomes. Official sources describe it as a subnet where miners generate simulated price paths that try to match real market behavior, including volatility clusters and fat tails, while validators score those forecasts with Continuous Ranked Probability Score, then weight recent accuracy more heavily.
The simple version: It is like a weather model for markets, except it outputs many plausible price paths instead of one bold prediction.
Centralized equivalent: Think a probabilistic market data vendor, but built as a Bittensor competition.
How it works:
- Miners do generate simulated price paths that try to match real market behavior, including volatility clusters and fat tails
- Validators check score those forecasts with Continuous Ranked Probability Score, then weight recent accuracy more heavily
- The problem it solves: Traders and AI agents often get a single forecast when what they really need is a range of plausible outcomes.
- The opportunity: Better probabilistic data can improve trading, hedging, portfolio management, and agentic decision making.
- The Bittensor advantage: Bittensor turns forecasting into an always-on competition, so the model set can adapt continuously instead of waiting for a quarterly model refresh.
- Traction signals: Synth has official API docs, a live product site, and recent code activity. The subnet price is 0.01236, market cap is about 54,704 TAO, and the pool holds roughly 16,811 TAO, with 264 GitHub commits from 10 contributors in the latest snapshot.
Category: Financial Forecasting and Trading Signals | Centralized Competitor: Numerai, Kaiko, option analytics desks
Most market data products describe what already happened. Synth is trying to package what could happen next in a machine-readable way. That is useful if the end user is an AI agent, an options trader, or any workflow built around probability instead of narrative.
Mechanism:
According to the official README, miners are asked to generate multiple simulated price paths rather than one target price. Validators compare those distributions against actual market moves using CRPS, then update rankings with an exponential decay that favors recent performance. That design matters because it rewards calibration, not just lucky directional calls.
TAO.app and Supabase snapshot data show current pricing and demand conditions, but the mechanism and product-status claims above were kept anchored to official repos, docs, and websites. On the market side, Synth trades around 0.01236 TAO with roughly 54,704 TAO in market cap and about 16,811 TAO in pool depth. Seven day net flow sits near 2,126 TAO, which suggests recent demand has been constructive.
- Execution: Forecasting markets is brutal. A few good weeks do not prove durable edge, and probabilistic outputs are only useful if downstream users can actually integrate them.
- Competition: Financial prediction is crowded, both inside and outside Bittensor. Synth needs to keep improving calibration and distribution quality, not just storytelling.
- Market: The token is up 30.4% over 7 days. Momentum helps, but it also raises the bar for continued delivery.
