Studies

How a chart read becomes a number — the warped-time twin engine that ships today (multi-channel DTW over 104 tickers / ~167k sessions), the stricter twin metrics, canonical curves, motif search, and day-character physics landing now, and the roadmap beyond. Every method mapped back to what your eye sees on the tape.

How many historical days until every day has a true twin? Measured on the finder’s own metric: trading days vary in ~15 independent ways, so match quality grows on a power law — what more data buys, where brute force ends, and why 1:1 mornings are findable when 1:1 days are not.

The strongest look-alikes across the liquid universe right now, and the most identical pairs of trading days ever recorded — different symbols, eras, and asset classes, ranked by exact DTW distance.

Every call the look-alike finder would have made on SPY since 2022, replayed with the corpus cut to what existed before each date and scored against what the day actually did — abstains counted, weak reads scored separately.

What trade management actually paid per setup and per canonical day shape: every stop/target/breakeven/trail/scale/re-entry combination replayed over historical detector hits, with a train/test split separating recommended policies from in-sample noise.

Fade the first retest of the first-hour HOD/LOD. Out-of-sample +0.144R per trade, 87% win rate, profit factor 2.08 across 15 liquid US large-caps and ETFs.

Do opening gaps fill intraday? Measured in ATR units (comparable across instruments): gaps under 0.25 ATR revert 75% of the time; gaps over 0.5 ATR extend. The fill curve crosses 50% cleanly at ~0.5 ATR.

After an inside day (today's range contained within the prior bar's), the next session expands 72.9% of the time — 26.6 pp above the 46.3% baseline across 1,762 events. Direction is a coin flip: compression tells you magnitude, not which way.

The mirror of the inside day. After an outside day (today's range engulfs the prior bar's), the next session contracts 74.3% of the time — 25.9 pp above the 48.4% baseline across 1,475 events. Same lesson in reverse: the blow-out reverts, but the close location doesn't call the direction.

The 'always-in' follow-through, measured. After a daily trend bar (big body, close on its extreme), the next session pushes past that extreme 75.1% of the time on bull days / 69.1% on bear days — ~22-23 pp over baseline, symmetric across 2,811 events. But it doesn't stick: next-day close direction is a coin flip (53.2%). The extreme is tagged early, then reverts.

Two inside days in a row — a breakout-mode coil. The next session expands 85.5% of the time vs 72.3% after a single inside day (non-overlapping CIs across 76 events). But the move is the same size (0.91 ATR), just more reliable off a tighter base. Depth of compression buys conviction, not magnitude — and direction stays a coin flip.

The 'who is making money doing what' — per-bar cohort P&L across the seven trader cohorts (always-in, scalpers, swing, responsive, trapped longs/shorts, profit-takers).

Does fading a failed breakout pay? A new 20-day extreme that closes back inside reverses the next session exactly 50.0% of the time (CI95 [47.5%, 52.7%]) — a coin flip. All three pre-registered verdicts fail: the blind daily fade is a myth, the edge is in the confirmation.

Does breakout context rescue the trend-bar coin flip? A trend bar that closes through the prior 20-day extreme tags a new extreme 75.6% of the time next session — but its close finishes in-trend only 48.7% of the time, no better than a trend bar stuck inside the range (50.5%). The breakout-vs-range distinction buys the early poke, not the close.