Latest note
When is the market actually awake? Intraday activity in IST
The intraday activity profile of 49 perps on the Indian clock: where liquidity lives, where it dies, and why the quiet hours are quieter rather than dead.
Crypto perpetual-futures research: market structure, order flow, and systematic strategies. Negative results kept.
The intraday activity profile of 49 perps on the Indian clock: where liquidity lives, where it dies, and why the quiet hours are quieter rather than dead.
When the market is open, how deep it is, and how that changes across the clock and the calendar.
The intraday activity profile of 49 perps on the Indian clock: where liquidity lives, where it dies, and why the quiet hours are quieter rather than dead.
BTC realized volatility in 2025 against 2026 year-to-date, decomposed into close-to-close, intraday, Parkinson and range measures.
A suspected BTC–ZEC inverse relationship that turns out to live entirely in the levels of two non-stationary price series. ETH as control.
Hypothesis-driven tests of cumulative volume delta: absorption, failed sweeps, spot–perp divergence and flow-confirmed breakouts.
The original research brief: formulate falsifiable CVD–price hypotheses, implement them rigorously, and keep the negative results.
Combined dashboard across the H-series order-flow hypotheses.
Spot versus perp CVD divergence: frozen design.
First result in the series — no edge from spot–perp flow divergence.
H1, price absorption: design and repository assessment.
H1 exploratory pass, 2026 in-sample.
H1 verdict: all seven preregistered criteria fail. Retained as a negative.
H2, failed liquidity sweeps: frozen design.
H2 exploratory pass, 2026 in-sample.
H2 verdict: aggressive flow after a failed sweep does not predict reversal.
H3, flow-confirmed breakout continuation: frozen design.
H3 on 62 crypto perps.
H3 on the full 100-symbol universe including TradFi.
H3 verdict: aligned flow adds nothing to the breakout itself.
Signal and parameter sweeps on 1-hour candles, with walk-forward selection and explicit execution costs.
EMA crossover study design.
143 EMA pairs × 6 variants on 10 perps: no configuration survives walk-forward selection.
EMA9/20 crosses as a directional signal on ten coins, 1h.
EMA/SMA cross study design.
EMA/SMA crossover results across the 1h universe.
EMA momentum study design.
EMA momentum results.
EMA crosses filtered by flow (MFA) — design.
Does CVD improve EMA cross entries? Results.
Range expansion, confluence filters, and the frozen-method re-run on a genuine temporal split.
Breakouts, confluence filters and volatility regimes on BTCUSDT — the prior negative that later flow studies have to beat.
Out-of-sample order-flow hypotheses: design 2025, test 2026.
The frozen H-series methods re-run on a genuine 2025-design / 2026-test split.
Quiet upward grinds after a bearish impulse — do they break up or break down?
How often does a small, low-volatility upward consolidation break down rather than up?
Low-volatility grind, v1 results.
V2: does the same structure break upward and continue, beyond ordinary drift?
Low-volatility grind, v2 results.
What the lab needs next to answer questions it currently cannot.
