Signals
The evidence layer behind every subnet decision.
This page is not a live signals feed. It is the proof layer — the historical record of which indicators have helped predict forward subnet returns over 1, 3, 7, and 14 days, and which have not. Live signal values live on the subnet pages. This page tells you which of those signals you should trust.
Measure
Information coefficient, t-stat, hit rate, sample size — across 1d / 3d / 7d / 14d horizons and the long-run window.
Classify
Surface promising signals, identify inverse relationships, and rule out the ones that look like noise.
Validate
Only strategies built from signals that survive fees, slippage, turnover, and capacity can graduate into system strategies.
What are STI, SMI, SHI, EYI, and BHI?
Signals are not recommendations. They are evidence tests: did an indicator help rank future subnet winners above losers?
BHI
Bittensor Health Index. Network regime context.
SHI
Subnet Health Index. Liquidity and tradability filter.
SMI
Subnet Momentum Index. Price momentum.
STI
Subnet Traction Index. Capital flow and reserve movement.
EYI
Emission Yield Index. Emissions relative to reserve size.
Value / Reserve Ratio
Whether a subnet appears cheap or expensive relative to pool reserves.
Strength × direction: verdict measures strength — Promising = |IC| ≥ 0.05 and |t-stat| ≥ 2; Watch = some signal but not yet significant; Weak = no clear edge; Too early = < 10 observations. Direction is separate: direct means higher value predicts higher forward return; inverse means lower value does. A signal can be both Promising and inverse.
Windows: the 1d / 3d / 7d / 14d ICs use the current evaluation window. Long-run IC uses the full available history. Divergence between the two columns is informative — it usually means the regime has shifted (e.g. momentum was working, has flipped to mean-reversion). Hit rate is direction-adjusted: for inverse signals we report the rate of directionally-correct calls, not raw up/down moves.