What actually happens the day AFTER an upper lock

The most expensive belief in this market

The belief: a stock that hits the upper circuit will keep running, so buy it at the open tomorrow. It is the single most common retail action on PSX.

Here is the uncomfortable part: on the raw numbers, the belief looks right. Using the correct cap-rule definition from the previous lesson, across the whole panel:

[MEASURED 2026-07-31 · cap rule · whole panel 2016-2026]

And the clustering is not subtle. A stock that locked today locks again tomorrow 29.8% of the time. A stock that did not lock today locks tomorrow 0.70% of the time. That is a 42.6x lift. One of the largest conditional effects anywhere in this dataset.

So the median post-lock day pays +3.22% against a PSX round-trip cost of about 3.04 percentage points [LEDGER: reference_backtest_methodology]. It clears the toll booth. On these numbers, chasing locks works.

Now split the same rows by one question: could you have bought it?

A stock sitting at its upper cap has buyers and no sellers. That is what a lock is. If it opens locked again tomorrow, your buy order does not get filled, you are queued behind everyone else who read the same WhatsApp message. So separate the 10,162 post-lock days by whether the next session actually re-locked:

[MEASURED 2026-07-31]

The entire post-lock edge lives in the bars you cannot buy. Every one of those +10.00% medians is a day the stock ran straight back to its ceiling and stayed there. Strip them out and what is left. The 70% of cases where a real order would actually have been filled, has a median of exactly zero and a negative mean of -0.90%, before paying 3.04pp in costs.

The same split on the liquid universe only (60-day average volume ≥ 10,000 shares, the population ZEME will actually discuss): re-lock rate 31.7%, fillable median +0.00%, fillable mean -0.10%, n=4,617. The finding does not depend on thin, untradeable names. It survives exactly where you would trade.

The lesson, stated precisely

The momentum after a lock is real, huge and measured. It is also almost entirely unreachable, because the mechanism that produces the return, no sellers at the cap, is the same mechanism that prevents you from taking the other side of it. A backtest that does not model fills will show you a spectacular strategy that cannot be executed by anyone.

This is why ZEME's own rules forbid a same-day post-circuit entry and apply a 3-5 day cooling period, and it is why "can this order actually be filled?" is a question we ask before "is this return large?".

Evidence · History · Confidence · Counterargument

Evidence, 10,162 cap-rule locks with a following session, partitioned by whether that session also locked. Both partitions come from one pass over the same rows. [MEASURED 2026-07-31]

History. An earlier draft of this lesson used the loose >= 7.4% threshold and reported a median of +0.06%, concluding that post-lock momentum barely existed. The correct definition shows the opposite: the momentum is enormous. The conclusion, do not chase the lock, survives, but the reason changed completely, from "there is no edge" to "the edge is unfillable". A right answer for a wrong reason is still a wrong model, and it would have failed the moment it met a different question.

Confidence, high on the split; n is large in both partitions and the effect is an order of magnitude, not a rounding difference.

Counterargument. The strongest case against this lesson: "did the next session close locked?" is a proxy for fillability, not a measurement of it. Some re-lock days open unlocked and trade freely for part of the session before pinning at the cap, and a fast order genuinely gets filled. We do not yet have enough intraday tape, it starts 2026-07-10, to measure how often that happens. Until we do, treat the fillable column as the honest floor and the re-lock column as unreachable. Neither is a forecast.