Every morning I publish a pre-market report: GIFT Nifty, crude, global cues. So does every market page in India. Almost nobody goes back to check whether any of it predicted anything, so I did, with 125 sessions from 24 March to 28 September 2026.
The answer turned out to be clear and useful. The pre-market read explains about 81% of the opening gap and about 0.1% of what happens after it. Knowing exactly where that line falls changed how I read my own morning notes.
What I tested
The inputs, all known before the open:
- GIFT Nifty: how far it had moved by 7 AM from where it stood when the Indian market closed the day before. (What it is and how accurate it is as a gap guide: GIFT Nifty explained.)
- Overnight crude: the overnight percentage move in MCX crude oil futures.
- The US close: this one I could not test. My records don't include it in a form I could line up with each date, and I'd rather leave it out than approximate it. It's a gap in this post.
Two things they were set against:
- The gap: where Nifty opened at 9:15 compared with its previous close.
- The day after the open: the move from the 9:15 open to the 3:30 close.
The distinction matters. The gap is the overnight part, and it has already happened by 9:15. The open-to-close move is the part still to come, and it's the part a trading bot actually trades.
Result 1: GIFT Nifty explains the gap
| Correlation with the gap | Share of the gap's variation explained | |
|---|---|---|
| GIFT Nifty move at 7 AM | 0.90 | 81% |
It pointed the right way on 87% of days. That's a strong relationship, and it's not surprising: GIFT Nifty is a Nifty futures contract trading while India is closed, so of course it prices the gap.
Result 2: it explains almost none of the day
| Correlation with the open-to-close move | Share explained | |
|---|---|---|
| GIFT Nifty move at 7 AM | 0.03 | about 0.1% |
The direction of the signal matched the direction of the open-to-close move on 55% of days. If you flipped a coin you'd expect around 50%. To be fair to the number: with 125 days, a 55% hit rate is well within chance, and the margin of error on the correlation is roughly −0.18 to +0.23. There is no relationship visible here, and the sample is too small to rule out a small one.
Even on days with a big signal, more than 100 points either way (53 days), the open-to-close direction matched only 55% of the time.
Result 3: crude has a weak link, and it's already in the price
| Overnight crude move | Days | Average gap | Days that gapped up |
|---|---|---|---|
| Biggest falls (lowest third) | 42 | +62 points | 69% |
| Middle third | 41 | +37 points | 59% |
| Biggest rises (highest third) | 42 | −75 points | 24% |
That's a clear pattern in the gap. Falling crude went with higher opens and rising crude with lower ones, which is plausible for a market where India imports most of its oil. The correlation is −0.29, and the range of plausible values, given the sample size, is roughly −0.46 to −0.14.
But it's a weak link, and here's the important part: when I put GIFT Nifty and crude together, the share of the gap explained went from 81.2% to 81.3%. Crude added nothing beyond what GIFT Nifty already showed. That's what you'd expect if the market has already priced crude by 7 AM.
Against the day after the open, crude's correlation was 0.08. Nothing there either.
One thing I noticed, and don't trust
I went looking for the popular idea that big gaps tend to reverse:
| Gap at open | Days | Closed above the open |
|---|---|---|
| Down by more than 100 points | 25 | 16 |
| Up by more than 100 points | 24 | 15 |
Both rows say the same thing: on big-gap days the market more often rose after the open (31 of 49), whichever way it had gapped. That isn't a reversal pattern, and at this size it isn't strong evidence of anything. It goes on the list to recheck when the sample doubles.
What this means
- The pre-market read is a description of the start, not a forecast of the day. It tells you how far the market has already moved overnight.
- Crude mostly comes through GIFT Nifty. It added nothing once GIFT Nifty was counted. I couldn't test US markets, so that one stays open.
- The part that matters for a trading bot happens after 9:15, and none of these inputs tell me much about it. That's why my morning reports describe the setup and never say where the day is going.
- Sample size limits everything here. 125 days is enough to confirm the obvious and not enough to trust the subtle. What counts is the number of days, not the number of data points.
I'll repeat this when I have twice as much data, and add the US close properly.
Sources
- Session data: my own compilation from the GIFT Nifty, Nifty 50 and crude oil futures price history for 24 March to 28 September 2026, obtained through the Kite Connect API. Correlations and ranges are my own calculations, with ranges from resampling the 125 days.
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.