Who's accumulating and who's dumping, by cohort × sector. Bubble/tile area = size of the net flow.
🏆 Voices — Caller Track Record
Curated voices ranked by realised call accuracy & ROI. Private · not a public or verified rating.
Voices linked when they post the same ticker. Colour = cluster (same colour ≈ co-moving together); grey = solo. Bold glowing link = ≥2 shared = possible coordinated push. Scroll/pinch to zoom in (reveals handles → stats → which tickers bond each pair). Drag a bubble to untangle; tap for profile.
🎒 My Bags — Paper Portfolio
Tickers you're tracking at your own entry. Paper / simulated — marked to the latest price (live in-session, else last close). Private · not advice.
PSX Pulse
What Pakistani markets are actually talking about right now — voice velocity, NCCPL flows, target tracking.
Market canvas + cap-weighted treemap of the 67 tracked names
Honest detection — named observations with auditable receipts, never a blended score or a call
Calls — explicit price targets resolved against current price
You'll need a contributor cookie to post calls — claim an invite first.
How PSX Pulse works
PSX Pulse aggregates public posts from a curated list of Pakistani finance Twitter handles and joins them with NCCPL daily flows and PSX prices. Everything shown is detection, never advice: named observations (the Decision Card, convergence counts, regimes, filings) with auditable receipts. It is not investment advice.
No blended score
There is deliberately no composite 0–10 score anywhere. Earlier versions carried a multi-factor “Heat” score; it was retired (2026-06-10) because any blended number quietly implies a proven edge. Rankings use one stated fact at a time — voices in the window, |move| today, top voice tier, or recency.
Targets
Targets are extracted from post text via regex (target 250, TP 100, SL 90, etc.) or set explicitly through the Call it form. Resolved against current price with a 7-day expiry window. Reference price pinned on first observation.
Halaal mode
Hides commercial banks, insurance, and tobacco — based on the broad Meezan-style screening universe. Updated quarterly. Toggle persists per device.
Behavioral tags
COOKING: top-3 by voices in the window (never on a silent name). SENDING: ≥3 voices and |Δ%| ≥ 3. BEEF: ≥2 voices with conflicting bull/bear language. LURKING: 0 voices but |Δ%| ≥ 2. DEAD: voices exist but none in the recent window.
Data sources
PSX market summary · NCCPL FIPI/LIPI · public X/Twitter accounts (no scraping of private data). Snapshot frequency is currently daily; tape and prices refresh on snapshot reload.
Proven layer (forward-validated)
Most of what you see here is detection — observations of chatter, flows, filings and divergences. A separate, stricter track asks: does a signal actually lead real, dated outcomes out-of-sample? A signal only earns the word "proven" after it clears a pre-registered statistical gate on a forward window — firings recorded going forward in time, then scored once their outcome window has elapsed. In-sample evidence can only ever reach candidate; it never mints "proven".
The honest, expected status for roughly a year (or indefinitely) is accruing — N below the minimum. We will not widen the frozen outcome horizon or loosen the thresholds to force the count up. The current candidate is book_closure_lead — does a clustered filing signal lead a real book-closure date? — validated against payout filings, not price moves.
Disclaimers
This is a research and exploration tool. PSX Pulse is not registered with the SECP as an investment advisor. Always do your own research and consult a licensed advisor before trading.