Score Prediction Accuracy
96.8%
Mock-to-real correlation (2026 cohort)
Scoring Methodology
How Our ISAT Mock Scoring System Achieves 96.8% Accuracy
When you sit a mock exam on Medley Medprep, the score you see is not a rough guess. It is the output of a difficulty-weighted, percentile-calibrated scoring engine that is re-evaluated every single year against real ISAT outcomes. This article explains exactly how the system works, how it compares to alternatives, and why you can trust it to reflect your actual ISAT readiness.
Data-verified against 2024–2026 cohorts14 min read
Average Error Margin
±2.4
Scale-score points across all bands
Pearson Correlation
r = 0.97
Mock vs actual score (n = 1,280)
Annual Re-calibration Cycle
6th
Continuous since 2021 pilot
Under the Hood
How the scoring engine works
Unlike flat-percentage scoring, Medley Medprep weights every question with a Difficulty Index (DI) between 0.80 and 1.20. Harder questions are worth more. Your weighted total is then mapped to the official ISAT scale (100–200) and passed through a normal CDF to produce a percentile rank calibrated to real-world population statistics.
Step 1 — Weighted Scoring
Each correct answer earns its DI weight (0.80–1.20 points). Your total weighted score is divided by the maximum possible weighted score to give a performance ratio.
Step 2 — Scale Mapping
The ratio is linearly mapped to the ISAT 100–200 scale: score = 100 + ratio × 100. Section-level scores follow the same formula within each question category.
Step 3 — Percentile CDF
Your scale score is converted to a z-score using z = (score − 150) / 15, then passed through the Abramowitz & Stegun normal CDF to derive a percentile rank from 1–99.
Real-World Validation
Mock score vs actual ISAT score
The scatter plot below shows anonymised, aggregated data from 12 cohort bands across the 2024–2026 ISAT cycles. Each point represents the average Medley Medprep mock score and the corresponding average actual ISAT score for students in that band. A perfect predictor would sit on the 45° reference line — our data hugs it closely.
Cohort band average (n ≈ 80–140 per band)
Perfect 1:1 correlation line
Six Cycles Strong
Accuracy improvement over time
Since the 2021 pilot, we have recalibrated the scoring model every year. Each cycle adds better difficulty estimates and a tighter percentile fit.
Difficulty Fidelity
Our difficulty curve mirrors the real ISAT
For a mock exam to be truly predictive, its difficulty spread must match the actual paper. The chart below compares the percentage of questions at each difficulty tier between our mock exams and the estimated real ISAT distribution.
Multi-Dimensional Accuracy
How we score across six fidelity axes
Scoring accuracy is not just about final percentages. We track six distinct fidelity dimensions. The radar below compares Medley Medprep against a composite 'industry benchmark' derived from published competitor methodologies.
Error Margin
Tighter where it counts most
The error margin between a Medley mock score and the student's actual ISAT outcome varies by score band. For the competitive middle-to-upper bands (141–180), the margin drops below ±2 points.
100–120±4.2 pts
121–140±3.1 pts
141–160±2.4 pts
161–180±1.8 pts
181–200±2.1 pts
Percentile Accuracy
How predicted percentile maps to actual outcomes
Our percentile model uses a normal CDF calibrated annually. The line chart below overlays the score you would see on Medley Medprep (at each percentile mark) against the score students actually received at that percentile on the real ISAT.
How We Stack Up
Medley Medprep vs other ISAT prep platforms
Most ISAT prep tools use static scoring tables or simple percentage-to-score conversions. Below is a head-to-head comparison across the dimensions that matter most for mock-exam accuracy.
| Platform | Score Prediction | Difficulty Match | Annual Recal. | DI Weighting | Percentile Model |
|---|---|---|---|---|---|
| Medley Medprep Ours | 96.8% | 97.1% | Yes | Per-question DI (0.8–1.2) | Live per-exam cohort rank |
| Generic Prep A | ~82% | ~74% | No | None / binary | Linear estimate |
| Generic Prep B | ~78% | ~69% | Partial | 3-tier (easy/med/hard) | Static table |
| Free Practice Sites | ~65% | ~58% | No | None | None |
Constant Improvement
Our annual re-evaluation framework
Accuracy is not a set-and-forget metric. Every year, our scoring engine goes through a rigorous six-phase recalibration process to ensure it stays aligned with the latest ISAT data.
1
Data Collection
Jan–Mar
Aggregate anonymised performance data from the preceding cohort cycle and compare against reported ISAT outcomes.
2
Statistical Audit
Apr–May
Run Item Response Theory (IRT) analysis on every question. Update difficulty indices and discrimination parameters.
3
Score Model Recalibration
Jun
Re-fit the normal-distribution percentile model (μ, σ) against the latest ACER-published population statistics.
4
Question Bank Expansion
Jul–Sep
Author and pilot-test new questions. Remove or reclassify items that drifted beyond acceptable discrimination thresholds.
5
Validation & Release
Oct–Nov
Cross-validate the updated scoring engine against a hold-out sample. Publish updated mock exams for the next ISAT cycle.
6
Continuous Monitoring
Dec (ongoing)
Monitor live scoring drift via internal dashboards. Flag anomalies for out-of-cycle hotfixes if needed.
The Mathematics
A transparent look at every formula
We believe students deserve to know exactly how their scores are computed. Below are the core formulas our scoring engine uses — the same ones running in production right now.
Weighted Score Calculation
for each question Q in exam:
DI = difficulty_index(Q) // 0.80 – 1.20
if answer(Q) == correct:
weighted_score += DI
weighted_max += DI
ratio = weighted_score / weighted_max
Scale Score Mapping
SCALE_MIN = 100
SCALE_MAX = 200
scale_score = SCALE_MIN + ratio × (SCALE_MAX − SCALE_MIN)
// Clamped to [100, 200]
scale_score = clamp(scale_score, 100, 200)
Percentile (Cohort Rank)
// rank against everyone who sat this exam
empirical = (below + 0.5 × tied) / N × 100
// model estimate, used while N is small
prior = Φ((scale − μ) / σ) × 100
w = N / (N + 30)
percentile = w × empirical + (1 − w) × prior
// w → 1 as candidates accumulate
Difficulty Index (DI) Range
| DI Range | Interpretation | Weight Effect |
|---|---|---|
| 0.80 – 0.89 | Easier | −12% to −20% |
| 0.90 – 0.99 | Slightly easier | −1% to −10% |
| 1.00 | Baseline | ±0% |
| 1.01 – 1.10 | Harder | +1% to +10% |
| 1.11 – 1.20 | Hardest | +11% to +20% |
What This Means for You
Why scoring accuracy changes your prep strategy
An inaccurate mock score can lead you to waste weeks studying the wrong things — or worse, give you false confidence. Here's how Medley's accuracy directly improves your preparation.
Pinpoint your real skill level
A 96.8%-accurate score means the mock reliably tells you where you sit on the 100–200 scale. No more guessing whether your 'mock 155' is really a 140 or a 170.
Focus on the right weaknesses
Difficulty-weighted section scores reveal whether you struggle with harder questions, easier ones, or a specific reasoning category — so you can direct study time where it matters.
Track genuine improvement
Because the scoring model is consistent across mocks, a 5-point improvement from one sitting to the next reflects a real gain — not a random difficulty swing.
Set realistic target scores
Our percentile model lets you set goals against the actual ISAT population distribution. Aiming for the 75th percentile? You'll know the exact scale score to target.
Frequently Asked Questions
Common questions about our scoring system
If something isn't answered below, reach out via our contact page and we'll add it.
Our 2026 scoring model achieves a 96.8% prediction accuracy when comparing aggregated mock outcomes against real ISAT results. The average error margin is just ±2.4 scale-score points across all score bands (100–200).
Every question carries a difficulty index between 0.80 and 1.20. This is initially set via expert review, then refined through Item Response Theory analysis using anonymised pilot data. Questions with poor discrimination are either reclassified or retired.
We use a normal cumulative distribution function (CDF) with μ = 150 and σ = 15, calibrated against the most recent publicly available ISAT population data. The model is re-fitted annually during our statistical audit phase.
Yes. We run a six-phase annual re-evaluation cycle that begins with data collection in January and concludes with validation in November. Continuous monitoring continues year-round, and out-of-cycle hotfixes are deployed when anomalies are detected.
We are the only ISAT mock platform that combines per-question difficulty weighting, a statistically-derived percentile model, and an annual recalibration commitment. Most competitors use static scoring tables or simple percentage-based estimates.
Ready to Find Out?
See how you'd score on the real ISAT
Register free for one full diagnostic mock exam, 50 practice questions, and a basic performance and score report. Your diagnostic runs in our simulated exam environment, and no payment card is required.