Market Regime Update — Mar 28, 2026

Fourteen days. 137,826 settlements. The market has not flipped positive once.

Data covers March 14–28, 2026 across Binance, Bybit, Gate, KuCoin, and Hyperliquid.

The Regime in Numbers

Every single day in this window has printed a negative aggregate average funding rate. That's not a slow drift. It's a sustained directional posture — shorts have been structurally favored since mid-March.

March 17 is a data collection anomaly — 78 settlements is not representative of a full trading day. Strip that out and the regime picture is clear: negative averages throughout, positive settlement percentages ranging from 53% to 68%, with no day achieving a genuine bullish posture.

Today's Reading Is the Worst in Two Weeks

March 28 is early, but the early signal is notable.

48.6% positive settlements. That's the lowest ratio in the entire 14-day window, excluding the March 17 outlier. The last time positive settlements fell this far was March 22 (53.3%), and that episode resolved within two days as rates recovered toward 62-65%.

The average rate of -0.0138% today matches March 26 for the most negative reading since March 18.

Two sessions in three days printing at -0.013% or worse. That's not noise.

What's Driving the Deterioration

Look at the 48-hour price data for three tickers: SIREN, Q, and VCX. All three moved more than 190% within the observation window.

SIREN is the most instructive case. Funding on Bybit averaged -0.136% over 48 hours — short holders were collecting 49% APR equivalent while the price rose 193%. The market was structurally short on the way up, and is now structurally confused on the way down.

The SIREN funding rate story from March 26 covered the initial surge in detail. What's happened since: SIREN peaked around $2.14, dropped back to $0.73, and by this morning recovered to approximately $1.73. Each phase printed extreme funding — deeply negative on the way up, flipping to strongly positive as price reversed, then mixed as traders re-established positions.

These events don't happen in isolation. When a token with significant open interest across multiple exchanges swings 190% in 48 hours, the settlement counts drop (traders are closing, opening, getting liquidated) and the rate averages get dragged by the outlier readings. March 26's 6,402 settlements and March 27's 2,345 both reflect this — settlement counts collapsed while the regime readings deteriorated.

The OU Model View

For tokens where the Ornstein-Uhlenbeck model fits well, the implied mean-reversion structure gives a forward read on where rates are likely to go.

A few things stand out.

BABYUSDT on Hyperliquid has a half-life of 15.8 hours — the shortest in the high-R² set. Rates that deviate from the equilibrium (-0.003%) revert within hours. Fast reversion means Settled markets on this pair resolve quickly. Low half-life is where the probability of correct directional reads is highest on short time horizons.

WIFUSDT has a positive settlement rate of just 2.8%. That's not statistical noise — it's a token that almost never pays long holders. The OU equilibrium at -0.009% is deeply negative. Anyone holding WIF longs on Hyperliquid perps is paying roughly 7.9% APR in funding, continuously. That's the exposure Settled is designed to hedge.

DOTUSDT has a 124-hour half-life. Rate deviations on this pair take five days to resolve on average. Settled markets that resolve in 8 hours on DOT perps are essentially noise trades from the OU perspective — neither side has statistical edge on that time horizon.

The Structural Pattern Holding

Two weeks of negative aggregate rates, a positive percentage that has drifted down from 68% to 48% this morning, and a cluster of extreme-volatility events compressing settlement counts.

This is not a regime flip. Negative average rates persist because the distribution has a heavier left tail — a minority of deeply negative outlier settlements drag the mean while the majority cluster near the floor. The same structure we observed in the March 21 regime update is still in place.

What's changed: the chaos events (SIREN, Q, VCX) are creating transient dislocations where funding swings from -0.14% to +0.06% within hours. These are not mean-reversion opportunities from the OU model perspective — they're structural breaks driven by forced position changes. The OU model was not designed for these events, and neither side of a Settled market during a 194% price swing has statistical edge from the model. They're coin flips dressed up as rates.

The settled regime resumes when the chaos tokens stop printing 48-hour moves above 150% and settlement counts return to the 13,000–15,000/day range seen before March 26.

---

Until then, the actionable reads are in the slow-moving, high-R² pairs where the statistical structure is intact: BABY (fast reversion), VIRTUAL (medium-term negative bias), and INJ (positive equilibrium with 77.8% historical positive rate).

Related reading:

• The Crowd Was Short. SIREN Ran 175%. — the full SIREN crowd-positioning story
• Market Regime Update — Mar 21, 2026 — the prior two-week read
• How the OU Model Predicts Funding Rates — the statistical model behind Settled's predictions
• Browse Live Rates — current funding rates across 3,700+ pairs
• Predict Markets — trade directional views on where rates settle next