IV rank and IV percentile, done properly

Rank uses two days. Only the lowest and the highest reading of the window enter the formula. It says where today sits on the line between them. Percentile uses every day. It counts the days when implied volatility was below today's, strictly below, and divides by the days in the window. A day that tied today's counts as not lower.

A small invented case shows the split. Ten daily readings, nine of them between twelve and fourteen and one at thirty. Today is the last reading and it is fourteen. This is a teaching box with made up numbers, not market data.

The same fourteen is near the bottom of the range and well above the middle of the crowd. Neither is wrong. They answer different questions, and the thirty is why they disagree.

IV rank or IV percentile: what is the difference? Rank is a position in a range. Percentile is a position in a crowd. If one wild day sets the high of the window, rank squeezes every ordinary day into a thin band near the bottom. Percentile hardly notices, because that one day is one day. When the two disagree widely, the gap itself is the message: a single spike is setting the high.

Three choices change either number. Say them whenever you quote one.

A year of sessions holds one Budget, four results seasons and whatever else came along. If the year held a shock, every later reading is measured against it until it ages out of the window, and that takes a full window of sessions. Both measures carry that memory, and rank carries more of it.

To build either one yourself you need one column of daily readings from one series, and four steps.

Then change N and watch what moves. That is the fastest way to see which of the two depends on a spike.

What do IV and IVP mean in the option chain? On the NSE chain, the IV column is the volatility solved from each strike's own price. The rank and percentile are not part of that chain. Many broker and analytics screens add them next to it, and they are built on a series of the index's implied volatility, not on the strike's own history. Which series, which window and which of the two they quote differ from screen to screen. The question page has the short version and the other one sets the two side by side.

Neither number says options are cheap or dear. Both compare implied volatility with its own past. They do not compare it with the volatility that comes after, which is a different question and the subject of the realised volatility module.

Look at how the 30 Mar 2026 high reshapes the rank. It stretches the range to 20.60 points, so today's reading above the low counts for little. But 159 of 252 days were below it, which is why the percentile sits above the middle. On the line chart the spike dwarfs everything around it.

The window matters too. Shorten it, and both numbers move, but not equally.

The 60 session window leaves the spike out, so rank at 71.6 and percentile at 76.7 land within a few points of each other. The two longer windows include it. Rank drops to 16.9 on 120 sessions and 18.5 on the full window, while percentile reads 40.8 and 63.1. Series matter as well. India VIX closed at 13.43 that day, with a rank of 22.8 and a percentile of 63.9 over the same 252 sessions.

Read the two together. When both are low, implied volatility is low against its own history however you cut it. When both are high, the reverse. When rank is low and percentile is middling, as here, look at the chart for a spike before you read anything else into the pair. A screen that shows only one of the two hides that clue. Also note what the inputs are. The series is solved from settlement prices, so a stale or wide quote on one expiry can move a single day's reading, and that day then sits in every window that holds it.

Switching from the at the money series to India VIX moved the rank by a few points here. On a screen that sorts or filters by rank, a few points can move a name across a line. Pick one series and one window, state both, and keep them.