Count it, do not feel it
On any given day, WhatsApp groups make it feel as though upper circuits are everywhere. So count them. But before you can count a thing you have to define it, and this is where almost every PSX number you will ever be quoted goes wrong.
First, the definition, because it changes the answer by 3.7x
A PSX upper lock is not "a big green day". The exchange sets a hard ceiling for each session:
cap = previous close + MAX(10% of previous close, Rs 1.00)The Rs 1 floor is the part everyone forgets. For a share at Rs 3.00 the cap is Rs 4.00, that is +33%, not +10%. For a share at Rs 200 the cap is Rs 220, exactly +10%. So a single percentage threshold cannot describe the circuit at all: it is too loose for expensive shares and far too tight for cheap ones.
Count the same 1,029,026 daily bars three ways and you get three completely different markets:
| "Upper lock" defined as | Locked bars | Rate | Off by |
|---|---|---|---|
chg_pct >= 7.4% | 37,676 | 3.661% | 3.7x too many |
chg_pct >= 9.4% | 20,435 | 1.986% | 2.0x too many |
| the cap rule above | 10,172 | 0.989% | , |
[MEASURED 2026-07-31 · close >= prev_close + MAX(0.10*prev_close, 1.00) - 0.005]
The true rate is 0.989%, about 1 bar in 101, not 1 in 27. Across the panel's 2,756 sessions that is 3.7 real locks somewhere in the whole market on a typical day, and the median symbol locks 14 times in ten and a half years, roughly twice a year, not nine times.
Why this matters more than it looks
A percentage threshold does not just mis-count; it mis-counts unevenly. Split the same panel by share price and the size of the error changes by an order of magnitude:
| Previous close | chg_pct >= 7.4% says | Cap rule says | Overstated by |
|---|---|---|---|
| Under Rs 5 | 4.07% | 0.33% | 12.2x |
| Rs 20 and above | 1.24% | 0.97% | 1.3x |
[MEASURED 2026-07-31 · n = 133,442 sub-Rs5 bars and 580,860 Rs20+ bars]
For an expensive share the loose rule is roughly right. For a share under Rs 5 it is wrong by twelve times, because the Rs 1 floor means a cheap share needs a far bigger percentage move to actually lock, and an 8% day is nowhere near it. Cheap shares are exactly the ones that dominate tips, screeners and group chats. The error is concentrated precisely where you were most likely to act on it.
Why the base rate is the first number you should ever ask for
If a method claims it "finds upper circuits", the only question that matters is: how often does the thing happen anyway? A method that flags stocks which lock 1% of the time has found nothing if 1% is the base rate. It has found the market.
This is why ZEME publishes the base rate next to every phase, and why a hit rate quoted without a base rate, a definition and a window is not a number, it is a sentence.
Evidence · History · Confidence · Counterargument
Evidence. One query over the whole panel, all three definitions computed side by side on identical rows so they cannot drift: 10,172 cap-rule locks of 1,029,026 bars with a prior close = 0.989%. [MEASURED 2026-07-31]
Second source. The cap figure is not trusted from one calculation. Our formula was checked against a separately maintained table that stores each session's cap independently: 100.0% agreement on 5,949 overlapping rows, and 99.5% agreement on the resulting locked/not-locked flag. A number that only one piece of code can produce is not yet a fact. [MEASURED 2026-07-31]
History. This Academy's own draft of this lesson taught 3.660% using the >= 7.4% threshold. It was measured correctly and defined wrongly, and it is corrected here in public rather than quietly. Our internal constraints file already recorded the rule, "believe the cap formula", and the lesson had not caught up with it.
Confidence, high. The cap formula is the exchange's own arithmetic, not an estimate, and the three counts come from one pass over the same rows.
Counterargument, two real ones. First, a lock is defined here on the close, so a stock that touched its band intraday and fell back is not counted; the "touched the band" rate is higher and this number is the wrong one for an intraday trader. Second, PSX band rules have changed over the decade and a handful of symbols trade under different bands, so a small number of bars are judged by a cap that did not apply on that date.