1) AION Smart Crowd -
@aion5100
AION Smart Crowd is an inference model built on elite trader behavior rather than average market participation. Instead of sampling the crowd, it derives probabilities from a curated set of consistently top-performing traders.
The model identifies traders with strong historical performance using internal performance signals. For each market, it observes their live exposures and positioning, aggregates those signals using weighted methods, and applies mathematical transformations to convert that aggregate into a probabilistic forecast.
The resulting probability can be interpreted as what the strongest traders, taken collectively, are implying about an outcome, expressed as a single number.
Coverage: broad and applies to most
@Polymarket markets where tracked traders are active. AION Smart Crowd is currently in an experimental phase, and its weighting logic and transformations are actively being refined. It should be used as a directional signal rather than a standalone decision engine.
2) Outcome Learned Inference (OLI)
Outcome Learned Inference is an external forecasting engine optimized specifically for probabilistic prediction. It is trained using reinforcement learning, where learning signals are derived exclusively from resolved real-world outcomes.
The model is trained on recent prediction questions combined with contemporaneous news, and it only updates when events settle. This allows it to optimize for calibration and accuracy rather than for narrative or explanation.
Research results in simulation suggest >10% ROI across a test set in a practical Polymarket-style trading setup, and improved probabilistic calibration with a compact ~14B reasoning model. At inference time, it uses stabilized sampling techniques to reduce noise and improve consistency across markets.
On futurefun, OLI is integrated as a baseline signal within our aggregation framework. It is not proprietary to futurefun, but is treated as a high-quality external inference component.
Coverage: All markets across Polymarket.
3) Allora Network -
@AlloraNetwork
Allora is a model coordination network where many independent machine learning models generate forecasts on the same objective. These models are continuously evaluated, dynamically reweighted based on performance, and aggregated into a single evolving forecast.
Rather than relying on a fixed model, Alloraโs output reflects the weighted consensus of the best-performing models in the network at any given time.
On futurefun, Allora forecasts are integrated specifically for crypto price prediction markets on Polymarket.
Coverage: Bitcoin, Ethereum, and Solana markets.
4) Synth Network -
@SynthdataCo
Synth provides probabilistic price path data generated through a decentralized network of models on Bittensor. Instead of extrapolating solely from historical prices, Synth simulates a wide range of potential future price trajectories under different market conditions.
Hundreds of models contribute simulations, performance is continuously evaluated, and higher-performing models are rewarded. The output is a probabilistic distribution over future price paths rather than a single point estimate.
On futurefun, Synth forecasts are integrated for selected crypto, commodities, and equity markets on Polymarket.
Coverage: BTC, ETH, SOL, Gold, SPY, NVDA, GOOGL, TSLA, and AAPL markets.
/-
OVERALL Notes and limitations
Model availability depends on the specific market. Not every inference system applies everywhere!
All forecasts are probabilistic estimates, not guarantees, and unexpected events or thin liquidity can quickly invalidate assumptions. Also, at the moment, we are not displaying reasoning.
All models should be considered experimental, reflecting the natural learning curve of artificial intelligence, and will continue to improve over time as both we and their respective teams train, evaluate, and refine them.
app.future.fun