Allora Documents Hyperliquid Testnet Topic for Three-Minute Forecasts

Allora Forge documents testnet Topic 90 as a competition for forecasting three-minute log returns across Hyperliquid perpetual-futures markets, giving model builders a multi-asset task to evaluate short-term prediction performance.

The topic is labeled “Hyperliquid perps — 3-minute log returns” and supports up to 100 asset labels per worker. Its documented example uses three workers. A log return measures the logarithm of the future price divided by the current price.

Workers submit predictions for the selected assets, and Forge compares those forecasts with subsequently available ground truth during recurring evaluation cycles. Rankings and rewards are updated after each round of scoring.

Allora Forge is a model competition platform for the Allora Network. For return and volatility topics, promotion-readiness is assessed using effective sample size, directional accuracy, Pearson correlation, WRMSE improvement, WCZAR improvement, log aspect ratio and participation.

The directional-accuracy requirement calls for a one-sided 95% lower confidence interval greater than 50%. Participation must be greater than 90%.

Allora Atlas supplies the timeseries data platform used by Forge competitions and workers. Its functions include market-data discovery, historical downloads and real-time streaming.

Topic 90 is a testnet activity, not a new token, a change to Hyperliquid’s trading rules or a direct alteration to trading on the platform.