On a normal day between May and September 2026, the Nifty's high and low were about 170 points apart, or 0.72% of the previous close. The India VIX close the evening before gave a rough idea of how wide the day would be, and almost no idea of how far it would close from the day before.
Both numbers shape what I expect each morning, so I checked them against 96 sessions of my own market log.
What I measured, and how
The question. How big is a normal Nifty day? And does India VIX, read the evening before, tell you whether the next day will be bigger or smaller than normal?
The data. 96 Nifty 50 sessions, from 14 May to 29 September 2026, from the daily record my bot keeps, built from exchange data. It has one row per session and no duplicates. The three weekdays missing from it (28 May, 26 June and 14 September) have no NSE daily file either, so I've treated them as market holidays.
I checked every row against NSE's official daily index file. The day's range agreed to within a point on 66 of the 96 days, but my log's closing price often differed from the official close, by as much as 75 points. So I used the exchange's figures for all prices, and took the India VIX close, which my log didn't hold, from the same files. The 29 September file wasn't out yet, so that day's prices come from the data behind my post-market report for 29 September.
Three measures for each day:
- Range: the day's high minus its low, in points and as a percentage of the previous close.
- Close-to-close move: how far the close was from the previous close, ignoring direction.
- VIX the evening before: India VIX's close on the previous session, known before the day started.
Turning VIX into a one-day move. NSE's white paper describes India VIX as an annualised percentage, and its formula measures time in minutes out of 525,600 in a year, which is 365 calendar days. The index describes expected volatility over the next 30 calendar days.
To scale an annual figure down to one day, you divide by the square root of the number of days in a year. On NSE's calendar clock, that's VIX ÷ √365, or VIX ÷ 19.1. A VIX of 12 then means about 0.63% for one day, or about 145 points with the Nifty at 23,000.
Many traders divide by √252 instead (about 15.9), counting only trading days, which turns a VIX of 12 into about 0.76%, or 174 points. I use √365 because it follows NSE's own clock, and show the other versions beside it. None changes the conclusions.
How big a normal day is
| Day's range | Points | % of previous close |
|---|---|---|
| Smallest (6 August) | 72.9 | 0.30% |
| 10th percentile | 109.9 | 0.46% |
| 25th percentile | 131.3 | 0.54% |
| Median | 170.4 | 0.72% |
| Average | 192.5 | 0.81% |
| 75th percentile | 220.0 | 0.93% |
| 90th percentile | 313.1 | 1.34% |
| Biggest (29 May) | 518.0 | 2.17% |
Put simply: 5 days had a range under 100 points, 55 were between 100 and 200, 24 were between 200 and 300, and 12 were 300 or more.
The close-to-close move was much smaller. The median was 0.38% (about 93 points) and the average 0.50%. Only 14 of the 96 days closed more than 1% away from the day before. A typical day's range was about 1.9 times its close-to-close move: the market travelled further during the day than it ended up going.
VIX as a ruler for the day
Over the sample, VIX ÷ √365 came to a median of 0.66%, about 160 points. How often the actual close-to-close move beat it:
| One-day yardstick | Median size | Close-to-close move bigger | More than twice as big |
|---|---|---|---|
| VIX ÷ √365 (NSE's calendar clock) | 0.66% | 24 of 96 days (25%) | 5 days (5.2%) |
| VIX × √(calendar days since the last close ÷ 365) | 0.70% | 18 days (19%) | 3 days (3.1%) |
| VIX ÷ √252 (trading-day convention) | 0.80% | 16 days (17%) | 5 days (5.2%) |
The middle row counts a Monday as three calendar days. For a benchmark: if daily moves followed a bell curve and VIX were exactly right, about 31.7% of days would beat the yardstick and about 4.6% would beat twice it.
Two things stand out.
Most days were quieter than VIX implied. The median close-to-close move was 0.57 times the √365 yardstick. Over the whole sample, the daily moves work out to an annualised volatility of about 10.7% on a trading-day basis, against an average VIX of 13.3 the evening before.
The big days still came. These are the five days that moved more than twice the √365 yardstick:
| Day | VIX the evening before | One-day yardstick | Close-to-close move | Multiple |
|---|---|---|---|---|
| 12 June | 15.61 | 0.82% | +1.99% | 2.4× |
| 8 July | 11.65 | 0.61% | −2.12% | 3.5× |
| 3 August | 11.76 | 0.62% | +1.60% | 2.6× |
| 24 September | 10.35 | 0.54% | −1.64% | 3.0× |
| 28 September | 12.16 | 0.64% | −1.56% | 2.4× |
Four of the five came after a VIX close of 12.16 or lower, and 24 September followed the lowest VIX close in the sample.
What I didn't expect: the yardstick matched the range better than the close. The day's range was bigger than VIX ÷ √365 on 53 of 96 days, and the median range was 1.04 times the yardstick.
Does a higher VIX mean a bigger day?
I split the 96 days into thirds by the VIX close the evening before:
| VIX the evening before | Days | Median range | Median range (points) | Median close-to-close move | Days with a range above 1% |
|---|---|---|---|---|---|
| 10.35 to 11.93 | 32 | 0.64% | 152 | 0.38% | 4 |
| 12.01 to 13.39 | 32 | 0.60% | 143.5 | 0.38% | 5 |
| 13.43 to 19.63 | 32 | 0.89% | 213.5 | 0.38% | 12 |
The range column shows a link: 12 of the 32 high-VIX days had a range above 1%, against 4 and 5 in the other thirds, and six of the ten widest days came from that third. The close-to-close column shows none. The median move was 0.38% in every third.
Correlations say the same. (A correlation runs from −1 to 1, and 0 means no straight-line link.) Between the VIX close and the next day's range it was 0.38, with a range from resampling the 96 days of roughly 0.19 to 0.57. That explains about 14% of the variation in the range. Between the VIX close and the size of the close-to-close move it was −0.02, with a range of −0.22 to 0.22.
For comparison, yesterday's range had a correlation of 0.30 with today's range: nearly as much as VIX, from a number anyone can read off a chart.
There's a catch. The high-VIX third is mostly May and June (25 of its 32 days), and the low-VIX third is entirely July to September, so part of the link is simply early summer against late summer. Within each half of the sample the correlation was weaker: 0.26 for the first 48 sessions and 0.11 for the last 48.
And a low VIX didn't rule out a wide day. Two of the three days with a range above 2%, 8 July and 15 September, came after VIX closes of 11.65 and 12.29.
Expiry days
My log marks expiry days. In this sample that's every Tuesday, 20 in all, five of them monthly expiries. How expiry works is in my post on expiry days.
| Days | Median range | Median range (points) | Median close-to-close move | Days with a range above 1% | |
|---|---|---|---|---|---|
| Expiry Tuesdays | 20 | 0.80% | 187.1 | 0.47% | 4 (20%) |
| Other days | 76 | 0.70% | 166.7 | 0.35% | 17 (22%) |
The medians were a little higher on expiry days, but wide days were no more common, and the VIX going into them was almost the same (a median of 12.75 against 12.68). Twenty days is too few to call this a difference, and since every Tuesday was an expiry, I can't separate expiry from the weekday.
What this doesn't show
- It's 96 sessions. The resampling range on the main correlation, 0.19 to 0.57, is wide, and busy days cluster, so the days aren't independent.
- It's one market regime. VIX fell from 19.43 at the 13 May close to between about 10 and 14 in August and September. The Nifty's highest close in the sample was 24,774 on 3 August, and it ended at 22,716 on 29 September, the lowest close of the 96 days. On days like 28 September, oil prices and news of the war involving Iran were driving the market. A calmer or rising market could look quite different.
- It's daily data only. A range says nothing about when the high and low came, or whether the day trended or swung back and forth.
- VIX is a 30-day measure. Using it for one day assumes expected movement is spread evenly, and weekends, holidays and known events make it uneven.
What I take from it
VIX works for me as a rough ruler, not a forecast. In this period, VIX ÷ 19.1 was about the size of a typical day's full range, and the close-to-close move was usually a little over half of that. The middle half of days spanned roughly 130 to 220 points.
A calm VIX isn't a promise. Four of the five days that moved more than twice the yardstick came after a VIX close of 12.16 or lower.
It rhymes with what I found about the pre-market read: GIFT Nifty describes the gap well, VIX loosely describes the size of the swing, and neither says which way the day goes.
For a trading bot, it's a sense of scale: the opening range breakouts I tested all live inside a day that typically spans about 170 points, and some days far more. I'll repeat this when the sample doubles.
Sources
- NSE: India VIX white paper (annualised percentage; time measured in minutes out of 525,600 a year)
- NSE: India VIX computation methodology (expected volatility over the next 30 calendar days; near and mid-month Nifty options)
- NSE daily index closing files for 13 May to 28 September 2026, for official Nifty 50 and India VIX prices, for example 28 September 2026
- Session list and expiry marks: my own market log, built from exchange data. Correlation ranges are from resampling the 96 days; all figures are my calculations.
This post describes an analysis of my own records, for information only. It is not investment advice or a recommendation to trade. Past patterns say nothing certain about the future. I am not registered with SEBI as an investment adviser or research analyst.