The NSE lot size everyone publishes is wrong about half the time — and you can check this in two minutes

Disclosure: I work on OptionLab Pro. No pitch in here, the whole thing is one correction and you can verify it against NSE’s own file.

A lot size belongs to the contract, not to the symbol and not to the day you look it up.

During an NSE revision both lots are live at the same time: contracts that already exist run off at their original lot, and newly introduced ones carry the revised one. So on 15 December 2025, a NIFTY weekly expiring 23 December 2025 was 75 units, and one expiring 6 January 2026 was 65. One trade date. Both correct.

Now open any “NIFTY lot size” page. It gives you one number. That page is wrong for one of those two contracts and gives you no way to tell which — and if you size off it, every premium, margin and P&L figure you derive is off by the same ratio, with nothing on screen looking broken.

The exception is quarterly and half-yearly: the exchange converts those in place at a stated end of day, so for those the trade date decides and the expiry does not. And “at EOD” applies from the next session — trading on the stated day is still the old lot.

You can check the whole thing yourself. NSE publishes fo_mktlots.csv, and it is not one number per symbol — it is a symbol × expiry-month matrix, a column per expiry. That shape IS the answer: it is contract-keyed because the exchange publishes it that way.

We model it that way because order placement cannot be wrong about it, so I published the model: the revision history with the circular numbers, and a lookup keyed on the contract. It refuses where the per-symbol boundary has not been read from the circular rather than interpolating, because a stale-but-plausible lot size is worse than an error — it is silent.

Happy to be corrected on any of this if someone has the circulars to hand.