Cohort stretch — every instrument vs its own history (lines)
Each line = one instrument (its color = the legend), normalized so
differently-priced instruments compare on ONE axis. Impulse z (default) = σ vs the
visible window mean — best for divergence/alpha (+2σ means the same rarity for every line);
full stretch to fit = each line 0→1 across the window (shape); % change = from
the window start. SPX + NAS100 baselines are solid, dark and NEVER muted (both default
on; Bold baselines thickens them). Adjust reprices every line (and the baselines) into
REAL terms before normalizing: ÷CPI = inflation-adjusted purchasing power, ÷M2 =
vs money printing (each divided by the CPI/M2 of its observation month, fwd-filled to daily —
a historical viz, not a tradeable signal). Individual tickers only — no composite/basket line.
Click a line to identify it (dims the others; its tooltip alone); click empty space to
dismiss; shift-click for the all-series dateline. The stretch tape below is this chart's
companion heatmap. The Retail P/C reference (default on) rides its own right scale:
>1 = puts/fear, <1 = calls/greed. Window & zoom = the shared View(bars) control on the
tape (both charts zoom together).
Cohort stretch tape — every instrument vs its own history (heatmap)
Each row = one instrument's stretch: today's percentile vs its own last ~3 years.
DEEP RED = top decile vs own history — stretched RICH; DEEP BLUE = bottom decile — stretched CHEAP; pale = normal.
Rows: every member individually (that is the point — no aggregate hides its parts; composites were removed
2026-07-03), plus the Retail P/C row — the retail crowd's 0DTE put/call volume ratio, stretched the
same way (red = fear extreme, blue = greed extreme). Korea names carry a "(t-1 lag)" chip — their close
prints hours before the US close, so their returns are time-shifted one bar to line up.
Click a row's cells for its detail chart + stretch history below (the tooltip shows THAT row; click empty
space to dismiss; shift-click for all rows). Sort rows re-orders the tape by the dateline-selected
column — click any cell to pick the column; with none picked it uses the latest bar.
loading universe…
Select allDeselect all
Retail P/C overlay
Slow money backdrop — liquidity plumbing at native cadence
Net liquidity impulse = weekly change in (Fed balance sheet − reverse repo − Treasury cash),
z-scored vs the last 2 years. Above 0 = plumbing ADDING fuel to risk assets; below 0 = draining.
Strip below: each plumbing series individually (M2, WALCL, RRP, TGA, CPI, FINRA margin debt) at its own
native weekly/monthly cadence — never blended into the fast cohort math, and each print is only shown from
its real publication date (release-lag shifted), so nothing here peeks into the future. Amber = ramping
(younger than the stats window).
loading plumbing…
apps2/cohorts2 · redesign of /cohorts' main chart: change over time, extremes vs own history, per-member divergence (rz vs cohort peers).
RETAIL = NVIDIA · Tesla · Nasdaq-100 ETF · Russell 2000. VOL COMPLEX (MM-dealer proxy) = VIX · VVIX · SKEW — an honest proxy label: 2–3 vol indices are not dealer positioning.
BANK = USD crosses (JPY·CHF·CNY·KRW) + 10Y/30Y Treasury futures + gold + oil — the LISTED slice only; much of real bank flow is OTC.
CUSTOM = SK Hynix + your own tickers (analyzed alongside the built-ins).
RETAIL P/C (vol) = the retail crowd's 0DTE put/call VOLUME ratio (QQQ·TSLA·NVDA, ~2y of prints and accruing — ramping) — a tape row, a detail-chart overlay, and a backdrop overlay.
2026-07-03: composite/spread aggregates and the forward-return calibration panel were REMOVED at user directive (aggregates destroy granularity; bucketed forward returns compare narrow point samples) — the code is parked, not deleted.
Individual members only; internal ex-self baskets survive solely as the residual (rz) reference, never displayed.
Honesty notes: futures returns are roll-gap guarded (Yahoo =F is unadjusted); Korea names are time-shifted one bar ("(t-1) lag"); slow FRED/FINRA series are release-lag shifted (CPI +45d, M2/margin debt +28d, weekly +5d) so no chart peeks ahead;
FINRA margin debt is monthly with a ~1-month lag. Data: Yahoo + Databento GLBX + FRED + FINRA via ../decoup/data.php (per-asset src shown) + ../fundamentals/zerodte5y.php (P/C), 10-min cache, no auto-refresh.