How accurate is Cracku's XAT percentile prediction tool?

Cracku's XAT percentile predictor lands within ±0.5 percentile points of the actual result for most candidates, making it the most reliable free third-party tool for XAT score estimation. The tool skews conservative, typically underestimating by 0.2 to 0.3 percentile rather than inflating your score, which is the safer direction to err when XLRI calls are at stake.

How the Tool Works

Cracku aggregates self-reported scores from test-takers immediately after the exam, building a normalization curve across the three scored sections: Decision Making, Verbal and Logical Ability, and Quantitative Ability and Data Interpretation. Your raw score gets mapped against this crowd-sourced distribution. In recent cycles, Cracku has collected 15,000+ responses within 48 hours, which gives the model enough density to be genuinely useful. XAT's negative marking structure (0.25 per wrong answer, 0.10 per unattempted question in DM) complicates raw score reporting, so the quality of what people input matters as much as sample size.

Accuracy Across Percentile Bands

The predictor is sharpest in the mid-to-upper range, where the distribution curve is steepest and small score changes produce clear percentile shifts. At the extremes, the model loses grip because data points thin out.

If you are targeting XLRI BM, which historically requires 95+ percentile, a Cracku estimate of 94.5 should be treated as genuinely borderline, not a near-miss. If you are targeting XLRI HRM, where the cutoff sits closer to 92+ percentile, the tool's accuracy in that band is high enough to guide your college list with reasonable confidence.

What Reddit Reports

Multiple Reddit threads from XAT 2023 and XAT 2024 show candidates comparing Cracku predictions against final XLRI scorecards. The consensus: predictions in the 92-98 band were off by no more than 0.3 to 0.4 percentile points. Above 99, a handful of users saw gaps of nearly 1 full percentile point, which at that level can separate a XLRI BM call from no call at all. The tool consistently underestimates in the top tail, so if Cracku gives you 99.2, your actual result may be closer to 99.3 or 99.4.

Where to Be Careful

The predictor is only as good as its inputs. Three factors reduce accuracy in any given year:

• A significant share of respondents misreport unattempted questions as wrong (or vice versa), skewing the raw-to-percentile curve.
• Low participation from a specific city or test slot can create blind spots in the normalization model.
• XAT changes its scoring emphasis or section weights, as it has done before, making prior-year curves less applicable.

Use Cracku as a directional signal, not a guarantee. If your predicted percentile sits within 2 percentile points of a program's historical cutoff, apply and do not self-screen out based on the predictor alone.