ASVAB SIM — AFQT 95 ✓

🔗 Live app → https://relie28.github.io/ASVAB-SIM/

A single-file, offline-capable web app built to simulate and reinforce performance on the ASVAB / PICAT (Pre-screening, Internet-delivered Computerized Adaptive Test). Built around one user's real verified test results — official AFQT 95 confirmed, March 2026 — and kept live as an ongoing drill tool for continued prep and score maintenance.

Official scores (CAT-ASVAB · March 2026)

Section	Correct	Standard Score
AR	13 / 16	65
WK	11 / 16	60
PC	9 / 11	68
MK	12 / 16	62
GS	9 / 16	51

AFQT: 95

---

What It Is

When you take the PICAT, your answers are saved. If you score high enough to qualify for enlistment, you are later given a Verification Test — a shorter, proctored version of the ASVAB that confirms your PICAT score was genuine. If your Verification score doesn't match your PICAT within a margin, your PICAT score is discarded.

This app was built to close that gap. It reconstructs the real PICAT session — every question, every section score — and turns it into a repeatable, adaptive drill tool. The target was AFQT 95. That target has been met and confirmed.

The app remains live as a maintenance and improvement tool across all AFQT subtests.

---

Modes

PICAT Verification Drill (30 questions · 25 minutes)

The primary mode. Pulls from a curated bank of 130 real-test and practice questions across four AFQT sections:

Section	Full Name	Questions in pool
AR	Arithmetic Reasoning	38
WK	Word Knowledge	26
PC	Paragraph Comprehension	21
MK	Mathematics Knowledge	45

Every session is intelligently assembled using weighted sampling:

• Questions you missed on the real PICAT → always included (guaranteed)
• Questions you got right on the real PICAT → high weight (rehearse your hits)
• Harder fresh questions → medium weight
• Easier fresh questions → lower weight

Questions and answer options are shuffled on every session so you can't memorize position.

Full ASVAB Simulation (71 questions · 90 minutes)

A complete ASVAB simulation in real test order: GS → AR → WK → PC → MK. Draws from a pool of 118 original questions — only 71 are selected per session, so the question mix changes every time. AFQT score is calculated from AR + WK + PC + MK.

---

Difficulty Flavors

Every session (both modes) is randomly assigned one of three difficulty profiles:

Flavor	Effect
BALANCED	Equal weight across easy / medium / hard
CHALLENGING	Strongly favors hard questions
REVIEW	Strongly favors easy questions

The active flavor is shown as a small badge in the quiz header. This ensures that even if you see a repeated question, the surrounding session context — and the difficulty pressure — varies.

---

Adaptive Engine (PICAT mode)

Within the PICAT drill, the order of questions adapts live as you answer:

• Answer 2 consecutive questions correctly in the same section → the next hard question for that section is promoted to appear immediately, escalating the challenge
• Miss a question → streak resets, pressure eases

---

Real-Test Question Behavior

Questions pulled directly from your real PICAT carry special behavior:

• A REAL Q badge (green) or REAL MISS badge (red) identifies them
• When you answer a REAL Q, the app shows exactly what you selected on the real test (highlighted in gold), letting you compare your original response to your current one
• If you matched your original wrong answer, the correct answer is intentionally withheld — so you're forced to reason it out again next time, not just memorize the reveal

---

Scoring

PICAT Drill

Reports per-section accuracy and a raw score percentage at the end.

Full ASVAB — How the AFQT Is Actually Calculated

The AFQT is not a raw percentage. It's a percentile rank derived from a pipeline that normalizes per-section accuracy into standard scores, combines them into a composite, and maps that composite onto a national reference population.

---

Step 1 — Raw Correct Answers

Accuracy is computed per section. The verified real-test scores this app is anchored to:

Section	Correct	Accuracy	Standard Score
AR	13 / 16	81.25%	65
WK	11 / 16	68.75%	60
PC	9 / 11	81.8%	68
MK	12 / 16	75.0%	62

---

Step 2 — Item Response Theory (Difficulty Weighting)

The PICAT uses adaptive-style scoring. Each question has three parameters:

• Difficulty — how hard the question is
• Discrimination — how well it separates high vs. low ability test-takers
• Guess probability — the likelihood of getting it right by chance

These are combined to estimate your ability level (θ, "theta") per section — a continuous measure of true ability, not a raw count.

How this app approximates it:

Every question in the bank has a diff field (1 easy, 2 medium, 3 hard). The diff value drives two things:

1. Session flavor sampling — CHALLENGING sessions draw proportionally more hard questions; REVIEW sessions draw more easy ones.
2. Result badges — correct/incorrect hard questions are highlighted in the question review.

For the standard score calculation itself, raw accuracy (correct / total) is used with calibrated per-section slopes. Applying bonus weights for hard questions creates a systematic downward bias — realistic test-takers concentrate misses on harder items — so difficulty weights are excluded from scoring.

---

Step 3 — Standard Scores (20–80 scale)

Raw accuracy per section is converted to a standard score on a 20–80 scale using per-section slopes calibrated directly from the verified real-test data:

std = 50 + (pct − 0.5) × slope

Section	Slope	Derivation
AR	48	13/16 (81.25%) → SS 65
WK	53	11/16 (68.75%) → SS 60
PC	57	9/11 (81.8%) → SS 68
MK	48	12/16 (75.0%) → SS 62

Each slope is solved algebraically from the real score: slope = (SS − 50) / (accuracy − 0.5). PC uses a steeper slope because fewer questions (11 vs. 16) means each carries more weight on the real test.

---

Step 4 — Build VE (Verbal Expression)

VE is derived from combined WK+PC raw performance — mirroring how the real ASVAB computes it from the sum of correct WK and PC answers before any scaling:

VE_pct = (WK_correct + PC_correct) / (WK_total + PC_total)
VE_std  = 50 + (VE_pct − 0.5) × 58

Using the real test: (11 + 9) / (16 + 11) = 0.741 → VE = 64

The slope 58 is calibrated so that this combined-raw approach reproduces the verified composite:

VE = 64   →   2×VE = 128   →   128 + 65 + 62 = 255   →   AFQT 95 ✓

Why not just average the two standard scores?
Averaging (WK_std + PC_std) / 2 treats WK's 16 questions and PC's 11 questions as equally weighted. On the real ASVAB, VE is derived from the combined raw sum — so a miss on PC hurts more than a miss on WK. The combined-raw method captures this correctly.

---

Step 5 — AFQT Composite

AFQT = 2 × VE + AR + MK
     = 2(64) + 65 + 62
     = 128 + 65 + 62
     = 255

VE is doubled. Verbal (WK + PC) accounts for half of your total AFQT. Improving by 2–3 vocabulary questions has a disproportionately large effect on your final score.

---

Step 6 — Convert Composite → Percentile

The composite is mapped to a percentile using a normal distribution approximation of the DoD 1997 norming sample:

Percentile = Φ((composite − 200) / 33.4) × 100

• Mean = 200 — corresponds to 50th percentile (all sections at standard score 50)
• SD = 33.4 — derived so that composite 255 maps to exactly the 95th percentile

$$\Phi!\left(\frac{255-200}{33.4}\right) = \Phi(1.647) \approx 95.0%\ ✓$$

Cross-checks against known data points:

Composite	Calculated	Expected
200	50th	~50th ✓
235	~83rd	~80th ✓
245	~90th	~90th ✓
255	95th	95th ✓
265	~98th	~98th ✓
270	~98.5th	~99th ✓

This replaces the previous hand-crafted lookup table with a smooth, continuous, mathematically grounded function.

---

What a 95 Means for Enlistment

• Anything above 93 is effectively the same qualification tier
• Qualifies for nearly every enlisted MOS/rating across all branches
• At this level, line scores (GT, ST, EL, etc.) matter more to your recruiter than the AFQT percentile
• This score has been officially confirmed — the app continues to serve as a maintenance and improvement tool

---

Screens

Screen	Purpose
Home (Splash)	Shows your real PICAT section scores, launches either mode
Category Select	Toggle which AFQT sections to include in the drill; allocation scales proportionally
Quiz	Active question — section tag, difficulty badge, adaptive timer, progress bar
Results	Full breakdown by section with question-by-question review
Stats Dashboard	Tracks session history in localStorage — AFQT trend, per-section accuracy bars, last 50 sessions

---

How to Use

1. Open index.html in any modern browser (Chrome, Safari, Firefox, Edge)
2. No server required — runs entirely offline
3. Tap BEGIN TIMED QUIZ for the adaptive PICAT drill
4. Tap ASVAB SIMULATION for a full-length practice test
5. Tap VIEW MY STATS to see your history across sessions

---

Technical Notes

• Single file — all HTML, CSS, and JavaScript in index.html (~6,200 lines)
• No dependencies — no build step, no npm, no frameworks
• Fonts — loaded from Google Fonts (Barlow Condensed, IBM Plex Mono); app is functional without them
• Persistence — session stats stored in localStorage under the key asvab_stats (up to 50 sessions retained)
• Responsive — media queries at 768 px, 600 px, and 400 px for tablet and mobile
• Scoring engine — per-section slopes (AR 48 · WK 53 · PC 57 · MK 48) + combined-raw VE (slope 58) + normal CDF percentile (μ=200, σ=33.4); all calibrated to verified AFQT 95 anchor
• Integrity audits — auditQuestionBank() and auditExplanations() run on page load and auto-patch conflicting answer keys; auditUserAnswers() runs before scoring to recover any points lost to stale answer-key bugs

---

File Structure

ASVAB SIM/
└── index.html     ← entire app (questions, logic, UI, styles)
└── README.md      ← this file

---

Question Bank Summary

Bank	Sections	Total questions	Used per session
PICAT_QUESTIONS	AR, WK, PC, MK	130	up to 30 (weighted)
FULL_QB	GS, AR, WK, PC, MK	118	71 (flavor-weighted)

Questions in PICAT_QUESTIONS carry metadata:

• forced — the exact text of the answer you selected on the real PICAT (or null for fresh questions)
• initiallyWrong: true — you missed this question on the real test
• diff — difficulty tier: 1 easy · 2 medium · 3 hard