Classifying Derivatives Regimes with Funding, OI, and Basis
Funding describes only one dimension of positioning cost. Combining its historical percentile with changes in open interest and perpetual basis separates new leverage, deleveraging, and low-conviction regimes.
Three variables answer different questions
Funding is the periodic transfer between perpetual longs and shorts; open interest measures outstanding contract exposure; basis is the mark-price deviation from the spot index. High funding does not imply an imminent decline—it indicates a higher current cost for holding long exposure.
A historical percentile is more adaptable than a fixed threshold. A 0.01% observation can be extreme in a quiet regime and ordinary during speculation. Kieran Lab reports both percentile and raw value so relative context does not hide absolute cost.
Four operational regimes
High-percentile funding with rising OI indicates new exposure increasing the cost of longs and is labeled crowded-long; low-percentile funding with rising OI is crowded-short. A sharp OI decline takes precedence as deleveraging because aggregate OI alone cannot identify which side exited. Remaining states are neutral rather than forced into a signal.
Basis checks whether funding aligns with the perpetual premium. Positive funding with positive basis is structurally consistent; divergence may reflect formula differences, local liquidity, or timing. Divergence is first a data and market-structure diagnostic, not an automatic contrarian trade.
Use and failure conditions
The regime label is useful for risk tiering, historical filtering, and execution priority—not as a standalone next-candle forecast. Crowding can persist through a trend, leaving early contrarian positions paying funding while losing on price.
OI notional is mechanically affected by price. Dollar OI can rise with unchanged contract count when the asset appreciates. The research view retains both contracts and notional, and material conclusions should test whether the change is merely a price effect.