After NTA released official April 2026 scorecards, we compared every predicted mark against the actual result — across all 9 shifts and 10 percentile bands. 90 data points. Zero cherry-picking.
Accuracy is calculated using 100% − MAPE (Mean Absolute Percentage Error), where error for each data point is |predicted − actual| / actual × 100. This measures error relative to the actual marks at each percentile — not against the 300-mark total — giving a more meaningful signal. All 9 shifts and all 10 percentile bands (90th–99th) are included. No data points were excluded.
Delta (Δ): positive = we overpredicted (predicted higher than actual) · negative = we underpredicted (predicted lower than actual).
vs January 2026 Session
Our January 2026 audit achieved 90.2% accuracy across 140 verified scorecards. For April 2026 our model was recalibrated after the January audit — correcting a −1.55% conservative bias — meaning April predictions started from a better baseline. The overall accuracy formula and methodology are identical between both audits.
Read January 2026 audit →Ranked highest to lowest. Each shift accuracy is calculated independently across its 10 percentile data points.
| Rank | Shift | Accuracy | Avg Δ (marks) | Verdict |
|---|---|---|---|---|
| 1 | Apr 6 S1 | 99.25% | -0.2 (0.8 abs) | Excellent |
| 2 | Apr 6 S2 | 99.10% | +1.1 (1.3 abs) | Excellent |
| 3 | Apr 5 S2 | 98.75% | -1.4 (1.6 abs) | Excellent |
| 4 | Apr 5 S1 | 98.16% | -2.0 (2.4 abs) | Excellent |
| 5 | Apr 2 S1 | 98.00% | -2.6 (2.6 abs) | Excellent |
| 6 | Apr 4 S2 | 97.01% | -3.9 (3.9 abs) | Excellent |
| 7 | Apr 8 S2 | 96.19% | -4.3 (4.3 abs) | Very Good |
| 8 | Apr 4 S1 | 95.75% | -5.8 (5.8 abs) | Very Good |
| 9 | Apr 2 S2 | 95.72% | +5.0 (5.0 abs) | Very Good |
For each shift: Pred = our predicted marks · Act = official NTA result · Δ = pred − actual (color-coded by magnitude).
| %ile | Apr 2 S1 | Apr 2 S2 | Apr 4 S1 | Apr 4 S2 | Apr 5 S1 | Apr 5 S2 | Apr 6 S1 | Apr 6 S2 | Apr 8 S2 | ||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Pred | Act | Δ | Pred | Act | Δ | Pred | Act | Δ | Pred | Act | Δ | Pred | Act | Δ | Pred | Act | Δ | Pred | Act | Δ | Pred | Act | Δ | Pred | Act | Δ | |
| 99 | 176 | 181 | -5 | 178 | 174 | +4 | 176 | 184 | -8 | 185 | 189 | -4 | 189 | 194 | -5 | 185 | 188 | -3 | 173 | 173 | 0 | 195 | 195 | 0 | 160 | 165 | -5 |
| 98 | 159 | 161 | -2 | 156 | 150 | +6 | 159 | 165 | -6 | 163 | 165 | -2 | 168 | 170 | -2 | 165 | 166 | -1 | 150 | 150 | 0 | 175 | 171 | +4 | 141 | 145 | -4 |
| 97 | 145 | 149 | -4 | 141 | 135 | +6 | 147 | 153 | -6 | 148 | 154 | -6 | 151 | 154 | -3 | 150 | 152 | -2 | 135 | 136 | -1 | 159 | 158 | +1 | 128 | 133 | -5 |
| 96 | 137 | 139 | -2 | 130 | 124 | +6 | 139 | 145 | -6 | 140 | 143 | -3 | 144 | 143 | +1 | 142 | 143 | -1 | 125 | 124 | +1 | 149 | 147 | +2 | 120 | 124 | -4 |
| 95 | 130 | 131 | -1 | 121 | 114 | +7 | 132 | 137 | -5 | 132 | 135 | -3 | 135 | 134 | +1 | 133 | 133 | 0 | 117 | 116 | +1 | 141 | 138 | +3 | 113 | 116 | -3 |
| 94 | 123 | 124 | -1 | 114 | 107 | +7 | 126 | 131 | -5 | 125 | 128 | -3 | 126 | 126 | 0 | 125 | 124 | +1 | 110 | 109 | +1 | 133 | 131 | +2 | 106 | 109 | -3 |
| 93 | 116 | 116 | 0 | 107 | 101 | +6 | 120 | 125 | -5 | 119 | 122 | -3 | 118 | 119 | -1 | 119 | 119 | 0 | 103 | 103 | 0 | 125 | 125 | 0 | 101 | 103 | -2 |
| 92 | 111 | 113 | -2 | 101 | 96 | +5 | 114 | 119 | -5 | 112 | 116 | -4 | 111 | 113 | -2 | 111 | 113 | -2 | 95 | 98 | -3 | 118 | 118 | 0 | 95 | 99 | -4 |
| 91 | 105 | 108 | -3 | 94 | 93 | +1 | 108 | 114 | -6 | 106 | 111 | -5 | 103 | 107 | -4 | 105 | 108 | -3 | 91 | 91 | 0 | 113 | 113 | 0 | 89 | 95 | -6 |
| 90 | 98 | 104 | -6 | 87 | 85 | +2 | 103 | 109 | -6 | 101 | 107 | -6 | 97 | 102 | -5 | 99 | 102 | -3 | 85 | 86 | -1 | 107 | 108 | -1 | 84 | 91 | -7 |
| Shift accuracy | Apr 2 S1 | Apr 2 S2 | Apr 4 S1 | Apr 4 S2 | Apr 5 S1 | Apr 5 S2 | Apr 6 S1 | Apr 6 S2 | Apr 8 S2 |
|---|---|---|---|---|---|---|---|---|---|
| Accuracy % | 98.00% | 95.72% | 95.75% | 97.01% | 98.16% | 98.75% | 99.25% | 99.10% | 96.19% |
Apr 6 S1, Apr 5 S2, and Apr 4 S2 were nearly perfect — the majority of their data points fell within ±3 marks. These shifts had predictable difficulty curves with minimal normalization anomalies.
Across April shifts, the model slightly underpredicted — actual NTA results came in marginally higher than our estimates at the lower percentile bands (90th–93rd). This is consistent with NTA's normalization slightly compressing the lower tail.
We predicted exactly 195 marks for 99th percentile in Apr 6 S2 — matching the official NTA result exactly. This was the easiest April shift. Our model correctly identified the difficulty ordering across all 9 shifts.
Apr 8 S2 (toughest shift) had the largest errors at lower percentiles — we predicted 84 marks for 90th percentile, actual was 91. A 7-mark miss at the lower end is where tougher papers create the most normalization uncertainty.
If you used our predictor and scored near a percentile boundary, your actual rank should be within 1–2 percentile points of our estimate in most cases.
NTA takes your best percentile from Jan and April. Even if April was slightly under/over our prediction, the combined result favours you if you improved.
Small mark-level errors don't change which colleges are in reach — percentile-based cutoffs absorb normalization. Use our College Predictor with your official percentile.
Enter your April marks and shift to get your predicted percentile and AIR from official NTA data.
Predict My Rank →Enter your final percentile to see which IITs, NITs, and IIITs you can target in JoSAA counselling.
Find My College →Full shift-wise marks vs percentile data for all April and January 2026 shifts with AIR.
View Full Table →Transparency Commitment
All 90 data points used in this audit are published in full above. The predicted values are exactly what our tool showed before NTA released official results — nothing has been adjusted retroactively. The accuracy formula (100% − MAPE vs actual marks) is the same methodology used in our January 2026 audit so results are directly comparable. If you have your official NTA scorecard and the predicted value didn't match, we'd love to hear from us on Telegram.
Join our community → t.me/jeepredictor