Trust is established through data, not claims. We are publishing a comprehensive statistical review of our Session 1 algorithms, comparing our pre-exam models against the official NTA normalization curves.
Mean accuracy across all 10 shifts (90-99th percentile band).
Achieved in Jan 21 Shift 2.
Mean Absolute Error (Percentile Points) in Hard shifts.
Individual percentile nodes analyzed for this report.
Our audit methodology utilizes Mean Absolute Error (MAE) calculation to measure the deviation between our predicted percentile curves and the actual NTA distribution.
We categorize shifts into three difficulty clusters (Easy, Moderate, Hard). The accuracy percentage is derived by calculating the proximity of our predicted rank node to the actual rank node across the critical 90th to 99th percentile spectrum.
Models showed a conservative tendency (under-predicting percentiles).
Variance in "Moderate" difficulty models.
Highest correlation coefficient.
| Shift Identification | Difficulty Classification | Accuracy Rating | Performance Band |
|---|---|---|---|
| Jan 21 Shift 2 | Easy | 95.5% | Optimal |
| Jan 21 Shift 1 | Hard | 94.8% | Optimal |
| Jan 24 Shift 1 | Easy | 94.4% | High |
Analyzing the deviation between predicted and actual outcomes strictly for the top 1% of candidates.
| Shift Date | Percentile Deviation | Accuracy | Statistical Verdict |
|---|---|---|---|
| Jan 24 Shift 2 | 0.00 pts | 100% | Perfect Match |
| Jan 21 Shift 2 | +0.30 pts | 99.7% | Minimal Deviation |
| Jan 28 Shift 2 | -0.42 pts | 99.6% | Within Margin |
| Jan 22 Shift 2 | -0.77 pts | 99.2% | Standard Deviation |
Prediction models were frozen on January 20, 2026. Actual results were aggregated from NTA public scorecards and verified user submissions (n=140 verified nodes).
All MAE calculations were derived using predicted vs actual percentile node comparison across 14 reference points (90.0 – 99.9 percentile). Independent verification can be performed using the publicly archived prediction tables available on jeepredictor.in.
The percentile prediction tables used in this audit were publicly available on jeepredictor.in prior to the release of official Session 1 results (Jan 20–29, 2026). Archived snapshots of these prediction pages can be independently verified via search engine cache or timestamped third-party indexing.
"Building predictive models for JEE is a high-stakes responsibility. When the NTA releases results, most platforms go silent. We choose to speak up, especially about where we missed. This audit isn't just a report; it is our commitment to radical transparency. We don't just want to predict your rank; we want to earn your trust for the upcoming Session 2."
— Lead Analyst, jeepredictor.in
