overnight — when was the money actually made: overnight or during the day?
For the top 20 most popular US-listed stocks right now (measured: highest 20-session average
dollar volume = close × volume, from the house 545-ticker store), this page splits every close-to-close move
into its two halves — what $1 became held only overnight (bought at each close, sold at the next open;
jargon: the overnight return component, the split the "overnight effect" studies measure) vs
held only intraday (bought at the open, sold at the close). A third line shows the best fixed clock
window (in-sample, modeled) — the one buy-time→sell-time range that grew $1 the most when repeated every
session over the chosen lookback (1q · 1y · 2y · 3y). Buy and sell times sit on the minute (not snapped
to 10:30 / 11:30) when 1-minute bars exist. Everything is past tense: what each window DID in this
history. Price returns only — dividends (which land overnight), spreads and fees are excluded (stated below).
Loading…
Zoom in → the detail view:/apps2/hourality cuts the same
trading day into 10-minute blocks and hunts contiguous anomalous ranges across the
86-name Hyperliquid / Ostium universe, over just over a year of sessions.
This page is the macro end:session-level segments — overnight · morning · afternoon
· full session — for the top 20 most-popular US names, over a deeper daily history.
Finer and wider there; coarser and longer here.
loading the pack…
checking how fresh this is…
Across the top 20 — who paid overnight and who paid during the day?
computing…
Which parts of the trading day closed higher more than 55% of the time?
One row per stock, with every stock pooled on the pinned top row. Columns are the
11 parts of the session in clock order — overnight first (the previous close to the open), then each
hour block of the regular session, then the morning, the afternoon and the whole day session. A cell is the
share of sessions that part closed higher than it started. Set your own line in the box below: cells at
or above it are marked ▲, cells at or below its mirror ▼. The second display shows the same
cells as the average move per session in basis points (100 bp = 1%) — the two readings disagree where a
part of the day moves a long way without moving up more often.
Resolution: this grid reads the pack's hourly boundaries
(09:30 · 10:30 · 11:30 · 12:30 · 13:30 · 14:30 · 15:30 · 16:00) plus the overnight.
10-minute blocks are not in this pack and live in /apps2/hourality, which owns that
granularity.
Judged at…
Your time frames
computing…
computing…
Loading…
—
Click any of the 3 lines on the chart to see what it measures
(overnight, intraday, best window). Click empty chart space for a
dateline with every value at that session; the same spot again, Esc or Reset clears. Shift-click = every value too.
SPX & NAS100 — which hours made the indices money, and which hours lost it?
$1 held ONLY during each index's most bullish clock hours vs ONLY during its most
bearish — in-sample extremes of the same 64-window search, on the ETF proxies (SPY / QQQ). Shares the
lookback, View, $1/% unit and linear/log controls above; datelines sync with the stock chart.
All 20 ranked — overnight vs intraday vs the best clock window
Every number is growth of $1 over that window. Click a column header to
sort; click a row to open its chart above. The window strip draws WHERE in the session clock the best window sat
(orange = held; wraps past the right edge = held overnight).
Costs hurdle: the overnight side traded TWICE every session — its avg bp/session must clear two spread
crossings plus any commissions before it was real profit (costs are not modeled here).
Loading…
Hyperliquid & Ostium — every name, five ways $1 was held
Every equity tradable as a perp on Hyperliquid or Ostium — no 55% filter, no top-20 cut. Five strategies, clearly separate, last 6 months: overnight (bought the close, sold the next open) · intraday (bought the open, sold the close) · best hours (the one minute-resolution clock window that grew $1 the most) · worst hours (the one that lost the most) · hold only, never sell (buy & hold). Cells are green if $1 grew, red if it shrank; more opaque = a bigger move. Best hours also show % true bullish (how often that window ended higher); worst hours show % true bearish (how often it ended lower). Click a cell to chart that path. Price returns only — DEX funding, spreads and fees are not modeled.
Overnight · Intraday · Best hours · Worst hours · Hold only, never sell
One row per Hyperliquid / Ostium name. Each cell is what $1 became under that strategy over the last 6 months. Green = gain, red = loss; opacity scales from the smallest move (faint) to the largest (solid). Best hours: % true bullish. Worst hours: % true bearish. Click a cell to chart it below; click a column header (heatmap or table) to sort.
—
Click a heatmap cell or a board row to chart that strategy's $1 path over the last 6 months. The five lines are overnight, intraday, best hours, worst hours, and hold only never sell. Click a line for what it measures; empty chart space for a dateline.
Loading…
Overnight-edge tape — who was an overnight stock, when?
Rows = each stock's overnight-edge index (the overnight growth line ÷ the intraday
growth line): rising = the overnight side was pulling ahead. Click any date column to sort the rows by that
column (click again to reverse). The picker re-colors the same rows six ways.
Notes
How "most popular" is measuredTop 20 by 20-session average dollar volume (close × volume) among the US-listed names of the
house stocks store.
The two headline strategies
Overnight only: bought every close, sold the next open — each session multiplied $1 by open ÷ previous close.
Intraday only: bought every open, sold the same close. By construction the two lines multiply to buy & hold
exactly — they split the whole close-to-close move between them. A session with a missing open or close left
BOTH sides flat, so the split stays honest.
Session map — what one cell is
Session map — the pinned pooled row
Session map — which sessions count
Session map — what the levels in the banner are
Session map — choosing the range
Session map — your criteria do the judging
Session map — what counts as random
Session map — resolution, and where 10-minute blocks live
The best fixed window is IN-SAMPLE
Costs & dividends
Which sessions count
Intraday depth varies per stock (the store's hourly history is ~2–3 years,
young listings less) — when it covers less than the requested lookback, the window search reports the span it
actually used. The daily overnight/intraday lines always run the full lookback the store covers.
Hyperliquid & Ostium — every name, five ways $1 was held
Data
The house 545-ticker stocks store (daily + hourly bars, refreshed 3×/day by its own cron) — this app only reads
it. Rankings and boundary prices are packed server-side (6h cache, stale the moment the store updates) and all
strategy math runs live in your browser; the measured cost is in the line under the header. Session boundaries
are computed in America/New_York time, so summer and winter sessions both line up.
Machine-readable print-out (ASCII) — every number on this page as plain text
Fixed-width columns, units and n stated, every canvas and table by name.
This is the read path for a language model: it restates every chart on this page without reading a single pixel.
loading…
Zoom in → the detail view:/apps2/hourality cuts the same
trading day into 10-minute blocks and hunts contiguous anomalous ranges across the
86-name Hyperliquid / Ostium universe, over just over a year of sessions.
This page is the macro end:session-level segments — overnight · morning · afternoon
· full session — for the top 20 most-popular US names, over a deeper daily history.
Finer and wider there; coarser and longer here.