Is FII buying followed by a next-day gap up? (Nifty 50)
Day-T FII/DII cash figure (released ~18:00–20:30 IST) vs the T+1 opening gap (T+1 open ÷ T close − 1) · 3,154 T→T+1 pairs, T = 2014-01-01 → 2026-10-29 · Built 30 Oct 2026 22:43 IST
Verdict: only slightly. FII buying gives a small, statistically real but economically weak lift, and it is not something to trade on its own.
Nifty gaps up on 66.8% of all days (n=3,154). After FII net buying it is 69.0% (n=1,315); after FII net selling it is 65.1% (n=1,838). The 3.9 pp difference has z-test p=0.0211 and permutation p=0.0225. The average gap size barely differs: +0.120% vs +0.102% (permutation p=0.402).
Size matters more than sign, mostly on the selling side.
FII > +5,000 cr: 80.9% (n=47, p=0.0437), with a mean gap of +0.29%.
FII < −2,000 cr: gap-up odds fall to 58.0% (n=500, p<0.001).
Most of the raw effect is the day's own move. FII buying coincides with Nifty rising on day T (68.5% of buy days vs 43.2% of sell days; Spearman ρ=0.334). Comparing like with like:
Within T-return quintiles, the buy-minus-sell difference shrinks to about 2.0 pp.
Controlling for T return and the US overnight move, 1 SD more FII buying adds about 3.401 bps to the gap (≈8 pts at 23,053; HAC t=1.85) and about 1.63 pp to P(gap up) (p=0.0338).
The US overnight move dominates.
S&P up overnight: Nifty gaps up about 83% of the time, whether FIIs bought or sold.
S&P down overnight: 51% after FII buying vs 45% after selling.
S&P down ≥1%: only 23.3% gap up, and FII buying does not rescue it.
It does not hold out of sample. In 2024–26 the controlled FII coefficient is ≈0 (t=-0.52). A logit trained on 2014–23 scores 64.3% on 2024–26 with FII and 64.3% without it, so FII adds nothing. “FII bought → buy calls at open” would have been right less often than “always expect a gap up”.
For an options buyer, gap-ups after FII buying mostly fade.
T+1 closes above its open only 47.8% of the time (mean open→close -0.048%).
The gap is fully filled intraday on 66.5% of those days.
Bigger signals hold a little better: after FII > +2,000, gap-ups close above the open 51.7% of the time, which is still roughly a coin flip.
Key table: next-day gap after day-T flows
Condition on day T
Days
P(gap up T+1)
Mean gap %
Median gap %
Mean gap pts
Median gap pts
Binomial p vs baseline
ALL days (baseline)
3154
66.8
0.11
0.141
14.0
17.4
1.0
FII net buy (>0)
1315
69.0
0.12
0.14
16.0
16.9
0.0841
FII net sell (<0)
1838
65.1
0.102
0.141
12.6
17.7
0.1374
FII > +2,000 cr
213
70.9
0.158
0.182
23.4
31.3
0.2165
FII > +5,000 cr
47
80.9
0.292
0.298
46.4
45.8
0.0437
FII < -2,000 cr
500
58.0
0.002
0.094
1.0
19.8
0.0
FII < -5,000 cr
115
57.4
0.009
0.118
7.1
29.3
0.0374
DII net buy
2127
65.6
0.093
0.137
13.0
18.0
0.2498
DII net sell
1026
69.2
0.144
0.145
16.1
16.9
0.1044
FII+DII combined > 0
1869
67.4
0.108
0.137
14.8
17.1
0.6061
FII+DII combined < 0
1285
65.9
0.112
0.145
12.9
17.6
0.5148
FII 5-day sum > +5,000
391
71.6
0.182
0.182
24.2
27.2
0.0468
FII 5-day sum < -5,000
802
62.3
0.072
0.122
11.3
23.2
0.0087
FII buy & Nifty up on T
901
72.8
0.162
0.176
22.1
21.0
0.0001
FII buy & Nifty down on T
412
60.9
0.029
0.069
2.6
7.8
0.0139
FII sell & Nifty up on T
794
68.9
0.136
0.165
20.2
22.1
0.2138
FII sell & Nifty down on T
1041
62.2
0.076
0.116
6.8
14.8
0.0022
Binomial p tests P(gap up) in the group against the 66.8% all-days baseline. Points are historical index points; the index ran from ~6,000 to ~26,000 over the sample, so the % columns are the comparable ones.
Size buckets and quintiles (FII, FII 5-day sum, DII)
Typical FII day sizes (std dev) have grown about 4× since 2014 (≈₹870 cr → ≈₹3,700 cr). The “within year” quintiles rank each day against its own year, which removes that drift. Under that view the lift almost disappears: top quintile 69.2% vs bottom 66.4%.
Bucket
Range (cr)
Days
P(gap up)
Mean gap %
Median gap %
P(gap > +0.25%)
P(gap < −0.25%)
FII net quintile (all years): Q1 (biggest sell)
-21106 .. -1614
631
60.9
0.069
0.125
38.8
22.8
FII net quintile (all years): Q2
-1605 .. -557
631
67.5
0.127
0.144
39.0
16.0
FII net quintile (all years): Q3
-557 .. 51
630
66.8
0.116
0.145
36.3
14.8
FII net quintile (all years): Q4
53 .. 738
631
70.7
0.093
0.137
33.8
12.7
FII net quintile (all years): Q5 (biggest buy)
739 .. 28739
631
68.0
0.143
0.148
38.2
12.4
FII net quintile within year: 1
-21106 .. -286
636
66.4
0.119
0.169
42.9
20.0
FII net quintile within year: 2
-4673 .. 80
628
66.6
0.098
0.145
38.1
16.9
FII net quintile within year: 3
-2346 .. 416
629
65.7
0.099
0.131
35.6
15.4
FII net quintile within year: 4
-819 .. 1585
628
66.1
0.084
0.123
31.5
14.6
FII net quintile within year: 5
344 .. 28739
633
69.2
0.148
0.152
37.9
11.7
FII 5-day sum quintile: 1.0
-48805 .. -6914
617
60.6
0.05
0.113
37.6
22.5
FII 5-day sum quintile: 2.0
-6900 .. -2667
616
66.9
0.143
0.164
39.9
15.3
FII 5-day sum quintile: 3.0
-2666 .. 20
617
66.9
0.082
0.119
33.9
16.2
FII 5-day sum quintile: 4.0
22 .. 2923
616
68.2
0.109
0.145
36.5
13.0
FII 5-day sum quintile: 5.0
2939 .. 29249
617
70.8
0.158
0.16
37.8
11.8
DII net quintile: 1
DII -6908 .. -330
631
68.6
0.154
0.166
40.6
12.4
DII net quintile: 2
DII -329 .. 169
631
69.9
0.126
0.141
34.2
13.0
DII net quintile: 3
DII 169 .. 692
630
65.7
0.066
0.137
35.9
16.8
DII net quintile: 4
DII 692 .. 1870
631
68.8
0.157
0.153
40.1
14.9
DII net quintile: 5
DII 1871 .. 16764
631
60.9
0.045
0.114
35.3
21.6
Significance tests
z = two-proportion z-test. Permutation = 10,000 shuffles of the FII label across days (5,000 for the mean gap). Welch t-test p values are in results/summary.json.
Comparison (A vs B)
n A
n B
P(up) A
P(up) B
Diff (pp)
z-test p
Permutation p (P up)
Mean gap A %
Mean gap B %
Permutation p (mean gap)
FII buy vs sell
1315
1838
69.0
65.1
3.9
0.0211
0.0225
0.12
0.102
0.402
FII>+2000 vs FII<-2000
213
500
70.9
58.0
12.9
0.00118
0.0015
0.158
0.002
0.0244
FII>+5000 vs rest
47
3107
80.9
66.6
14.3
0.039
0.0445
0.292
0.107
0.0446
DII buy vs sell
2127
1026
65.6
69.2
-3.6
0.0435
0.0459
0.093
0.144
0.0292
Combined>0 vs <0
1869
1285
67.4
65.9
1.4
0.396
0.403
0.108
0.112
0.856
FII5>0 vs <0
1240
1843
69.4
64.8
4.6
0.00794
0.0097
0.131
0.093
0.104
Controls: day-T return and US overnight
FII × S&P 500 overnight
FII on T × S&P 500 overnight
Days
P(gap up)
Mean gap %
Median gap %
FII buy & S&P up overnight
701
83.7
0.257
0.242
FII buy & S&P down overnight
573
51.1
-0.047
0.004
FII sell & S&P up overnight
942
82.9
0.326
0.28
FII sell & S&P down overnight
837
45.0
-0.144
-0.043
FII > +2,000 & S&P up overnight
117
84.6
0.272
0.256
FII > +2,000 & S&P down overnight
90
52.2
0.011
0.01
Same T-return bucket, FII buy vs sell
Nifty T-return quintile
T return range
n FII buy
n FII sell
P(up) FII buy
P(up) FII sell
Diff (pp)
1
-12.98% .. -0.56%
124
507
66.1
61.3
4.8
2
-0.56% .. -0.11%
214
415
59.8
63.1
-3.3
3
-0.11% .. 0.25%
282
349
68.8
65.3
3.5
4
0.25% .. 0.71%
330
300
71.8
71.3
0.5
5
0.71% .. 8.76%
364
267
73.4
68.2
5.2
Weighted average buy−sell difference across T-return quintiles: 2.0 pp, compared with 3.9 pp raw.
Regressions (partial effect of FII)
Model
n
FII effect (bps of gap, or pp of P(up))
HAC t
p
R² / pseudo-R²
OLS gap_bps ~ fii_z
3034
2.9
1.85
0.0645
0.003
OLS gap_bps ~ fii_z + retT
3034
2.991
1.66
0.0976
0.003
OLS gap_bps ~ fii_z + retT + spx
3034
3.401
1.85
0.0638
0.2573
OLS gap_bps ~ fii_z + dii_z + retT + spx
3034
3.271
1.91
0.0559
0.2573
OLS gap_bps ~ fii_k(1000cr) + retT + spx
3153
1.629
2.19
0.0285
0.257
Logit up ~ fii_buy
3154
3.92 pp (AME)
0.0214
0.0013
Logit up ~ fii_buy + retT
3153
2.76 pp (AME)
0.118
0.0031
Logit up ~ fii_buy + retT + spx
3153
3.05 pp (AME)
0.057
0.1446
Logit up ~ fii_z + retT + spx
3034
1.63 pp (AME)
0.0338
0.1429
fii_z = FII net ÷ its trailing 250-session standard deviation, using past data only, so 1 unit = 1 “normal-size” FII day for that era. OLS uses HAC (Newey–West, 5 lags) errors. Logit effects are average marginal effects on P(gap up) in percentage points. T return and S&P are in %.
Stability by year and train/test
Year
Days
Base P(up)
n buy
P(up) after buy
n sell
P(up) after sell
Diff (pp)
Mean gap after buy %
Mean gap after sell %
Fisher p
2014.0
239.0
67.8
159.0
69.8
80.0
63.7
6.1
0.121
0.082
0.38
2015.0
244.0
66.8
108.0
65.7
136.0
67.6
-1.9
0.073
0.166
0.785
2016.0
243.0
68.3
116.0
69.0
126.0
67.5
1.5
0.074
0.129
0.89
2017.0
246.0
74.8
88.0
72.7
158.0
75.9
-3.2
0.143
0.174
0.646
2018.0
243.0
70.8
83.0
65.1
160.0
73.8
-8.7
0.056
0.106
0.181
2019.0
245.0
71.0
114.0
74.6
131.0
67.9
6.6
0.155
0.143
0.263
2020.0
251.0
70.9
134.0
71.6
117.0
70.1
1.6
0.222
0.151
0.889
2021.0
248.0
72.2
98.0
70.4
150.0
73.3
-2.9
0.153
0.209
0.665
2022.0
248.0
60.5
78.0
71.8
170.0
55.3
16.5
0.158
-0.027
0.0172
2023.0
246.0
68.3
107.0
68.2
139.0
68.3
-0.1
0.149
0.11
1.0
2024.0
248.0
65.7
92.0
75.0
156.0
60.3
14.7
0.143
0.073
0.0191
2025.0
249.0
59.4
79.0
63.3
170.0
57.6
5.6
0.045
0.034
0.41
2026.0
204.0
48.5
59.0
50.8
145.0
47.6
3.3
-0.046
0.015
0.758
Split
Days
From
To
Base P(up)
P(up) after buy
P(up) after sell
Diff (pp)
FII z coef (bps, controlled)
HAC t
“Buy→up / sell→down” rule accuracy
“Always gap-up” accuracy
train ≤2023
2453
2014-01-01
2023-12-29
69.1
70.0
68.5
1.5
3.553
1.59
48.5
69.1
test 2024–2026
701
2024-01-01
2026-10-29
58.5
64.8
55.4
9.4
-0.807
-0.52
51.2
58.5
Out-of-sample 2024–26 hit-rate of a logit trained on 2014–23: FII only 58.5% · T return + S&P (no FII) 64.3% · FII + T return + S&P 64.3% · always predict the majority 58.5%. Buy-days beat sell-days in 8 of 13 years, and the gap is significant only in 2022 and 2024.
Do gap-ups after FII buying hold or fade? (options-buyer view)
T+1 situation
Days
Mean gap %
Mean open→close %
Median open→close %
P(close > open)
P(gap fully filled intraday)
P(close still beyond prev close)
Mean open→high %
Mean open→low %
All gap-ups
2106
0.373
-0.069
-0.051
46.5
66.1
65.8
0.433
-0.617
Gap-up after FII buy
908
0.328
-0.048
-0.02
47.8
66.5
68.5
0.401
-0.554
Gap-up after FII sell
1197
0.406
-0.084
-0.074
45.5
65.9
63.7
0.458
-0.665
Gap-up after FII > +2,000
151
0.362
0.01
0.011
51.7
58.3
76.8
0.401
-0.544
Gap-up ≥0.25% after FII buy
473
0.515
0.001
0.011
51.8
49.5
82.0
0.439
-0.557
All gap-downs
1047
-0.419
-0.063
-0.052
46.4
62.7
70.4
0.542
-0.629
Gap-down after FII buy
406
-0.345
0.006
0.047
51.7
68.5
63.8
0.548
-0.555
Gap-down after FII sell
641
-0.466
-0.107
-0.09
43.1
59.0
74.6
0.538
-0.676
Any day after FII buy
1315
0.12
-0.032
-0.012
49.0
67.1
67.1
0.446
-0.554
Any day after FII sell
1838
0.102
-0.092
-0.078
44.7
63.5
67.5
0.486
-0.669
“Fully filled” means the T+1 low reached T's close (or the high, for gap-downs). “Close still beyond prev close” means T+1 closed above T's close for gap-ups (below, for gap-downs). Open→high and open→low are the average excursion from the open, i.e. the room a 09:15 option buyer had in either direction.
Context: Monday 2 Nov 2026
Day-T inputs: 30 Oct FII +3,599 cr (z≈1.01, ~92nd percentile of all days on the size-adjusted scale), DII -818 cr. Last-5-session FII sum is +6,305 cr: 10-26: -1,462, 10-27: +4,240, 10-28: +872, 10-29: -944, 10-30: +3,599. Nifty closed 23,053.10 (+0.51% on the day).
On flows alone:
Past days with FII > +2,000 and a 5-day sum > +5,000 gapped up 78.6% of the time (n=126).
FII > +2,000 into a weekend or holiday: 75.6% (n=45).
Logit (FII + T return, no US input): 69.7%.
But 2026 is a weak-gap year: only 48.5% of 2026 days gapped up, and 57.1% after FII > +2,000 (n=14).
The US Friday close will matter more than the FII print. At ~22:35 IST the S&P 500 was about −1.6% intraday (7,577 vs 7,699.6). Brent was ~$114–118 after Hormuz tanker attacks, and weekend Iran/oil headlines are an extra, unmodelled risk. If the S&P closes ≤ −1%, history gives only ~14–23% odds of a gap up even after FII buying.
Scenario for Mon 2 Nov (FII 30 Oct = +3,599 cr)
Past days
P(gap up)
Mean gap %
Median gap %
Median gap ≈ pts at 23,053
FII buy, S&P up ≥ +0.5%
344
92.2
0.375
0.317
73.0
FII buy, S&P −0.5%..+0.5%
733
70.8
0.112
0.119
27.0
FII buy, S&P ≤ −1%
104
13.5
-0.476
-0.342
-79.0
FII sell, S&P ≤ −1%
226
27.9
-0.543
-0.397
-92.0
Any FII, S&P ≤ −1%
330
23.3
-0.522
-0.382
-88.0
FII > +2,000, S&P ≤ −1%
11
0.0
-0.615
-0.32
-74.0
FII buy, S&P ≤ −1%, weekend/holiday gap
25
20.0
-0.596
-0.425
-98.0
These are base rates, not forecasts. They ignore weekend news, GIFT Nifty and crude oil, which on a weekend like this can outweigh everything above.
FII/DII daily cash (provisional, NSE+BSE+MSEI combined, ₹ cr): Tapetide monthly FII/DII flow reports (example month), 3,156 sessions 2014-01-01 → 2026-10-30. buy − sell = net on every row. Cross-checks: all 30 overlapping days with our live NSE/Moneycontrol feed matched exactly (0 mismatches); 118 of 118 overlapping real days from the Mr. Chartist NSE API (30 Mar → 30 Oct 2026) matched exactly. Its Jan–Mar 2026 rows are rounded “seeded/estimated” values, including 3 NSE holidays, and were not used. Spot checks against news reports for 24 Aug 2015, 13 Jun 2022 and 4 Jun 2024 (−12,436 / −3,319) also matched.
Nifty 50 daily OHLC: NSE official index history API (2014-01-01 → 2026-10-30, 3,174 sessions). This is the primary source for the full 2014–2026 span. It was checked against daily bars derived from the 1-minute parquet (first bar 09:15 open, last bar close; 2022-01-03 → 2026-10-06, read-only). Opens match exactly on 1,178 of 1,179 days. Closes differ by a median of 8 points because NSE's official close is a 30-minute average while the parquet uses the last traded tick. Gap direction agreed on 94.65% of days. With 1m-based gaps, P(up | FII buy) was 67.1% vs 67.6% official, and P(up | FII sell) was 55.9% vs 57.6%, so the conclusion does not change.
US overnight: S&P 500 (^GSPC) daily closes via Yahoo Finance. Each move runs from the last US close before Indian session T to the last US close before T+1. The 30 Oct 2026 US session was still open at build time and is excluded.
Missing days are never filled. 19 Nifty sessions have no FII print (mostly 2014–2018 special or partial sessions) and are dropped. One FII row (2024-03-02, a Saturday DR-test session) has no Nifty bar and is dropped.
Caveats
The FII figures are the provisional exchange prints, which is what traders see that evening. NSDL custodian-confirmed FPI numbers differ and are not used.
Gap = T+1 official open ÷ T official close − 1. The 09:15 open comes after the pre-open auction, and ~0.03% of days have exactly zero gap, counted as “not up”.
US overnight uses S&P cash closes. GIFT Nifty, crude and news are not modelled. The results show association, not causation.
Sample size is not the problem: 3,154 days, with 1,315 FII-buy days and 1,838 FII-sell days. The effect is small at that size, and the last 30 sessions alone (1 of 7 FII-buy days gapped up) say nothing either way.