Practice exam strategy and error logs

By week eight, Maya had answered more than four hundred practice questions, and her scores had climbed from 64 percent on the diagnostic to the high seventies. Then they stopped climbing. For two weeks she sat at 78 or 79 no matter how much she studied. Luis asked for her error log, read it for ten minutes, and handed it back. "You're not missing knowledge any more," he said. "Half of these you knew. You're missing them in the reading."

This lesson is about the part of exam performance that is not knowledge: reading questions accurately, eliminating options, managing time over a long exam, deciding when to guess and when to change an answer, and keeping an error log that turns every wrong answer into a specific fix. These skills are worth several marks to almost every candidate, and they are learned the same way as the trade, by practice with feedback.

How a multiple-choice question is built

Every question has a stem, the question itself, and four options: one key, the correct answer, and three distractors. Good distractors are not random. Each one is the answer a candidate would reach by a specific mistake: the general rule instead of the exception, the closest table value instead of the next one up, a gauge pressure instead of an absolute, a units slip, or a correct statement that does not answer the question asked. Knowing this changes how you read the options. When your answer matches one, ask what mistake would produce each of the others. If you can name the mistakes, you can be confident.

Stems also carry signal words that change the question completely. Maya circled them in every practice question until it became automatic.

A method for every question

Luis taught Maya a sequence of five steps for each question. It sounds slow, but with practice it takes only seconds on easy questions, and it saves the most time on hard ones because it stops her rereading.

• Read the stem to the end, and circle the signal words and the numbers.
• Say or write your own answer before looking at the options, if you can.
• Read all four options, even if the first one matches your answer.
• Eliminate the options you can rule out, and say why.
• Choose, flag if unsure, and move on within your time budget.

Reading all four options matters because exam writers sometimes put a partly correct answer before a fully correct one. Saying why an option is wrong matters because it exposes guesses that feel like knowledge.

Elimination and guessing

Since there is no penalty for a wrong answer, every question should be answered. The only question is how good the guess is. A blind guess among four options is right a quarter of the time. Eliminating one option raises that to a third; eliminating two raises it to a half. Across many unsure questions, that is a lot of marks.

Some patterns help when elimination runs out. Options with absolute words such as always and never are wrong more often than options with qualified words. When two options say opposite things, one of them is often the key. When one option is much more specific and complete than the others, it is often the key, because writers are careful to make the key fully correct. None of these is reliable enough to override knowledge, but they beat a blind guess.

Changing answers

Many candidates believe that the first answer is usually right and that changing it is a mistake. Research on multiple-choice tests has generally found the opposite: when people change an answer, the change is more often from wrong to right than from right to wrong. The useful rule is not to avoid changes but to change only for a reason: a misread word found on review, a calculation error found, or a later question that reminds you of a rule. Changing because of a vague feeling is where the losses come from.

Maya tested this on herself. In her error log, she marked every answer she changed in the review pass. Over four mock exams, she changed 23 answers: 15 from wrong to right, 5 from right to wrong and 3 from wrong to wrong. All five losses had the same note: "changed on a feeling". She kept reviewing, and stopped changing without a reason.

Pacing a long exam

Three hours is a long time to concentrate, and on an open book exam the book itself can eat the time. Maya's plan used two passes. On the first pass she answered every question she could within her time budget, about 2 minutes 15 seconds, and flagged any that needed a long search or a plan-reading calculation. She did not skip questions completely; she always marked a best answer before flagging, so that nothing was blank if time ran out. On the second pass, in the time she had banked, she returned to the flagged questions in order of how close she thought she was.

Concentration drops in the second half of any long exam, and mistakes from misreading rise. Maya practiced the full length at least three times, so that her pace and her reading held up at the end. She also planned short resets: at each checkpoint, ten seconds with her eyes closed, a slow breath, a sip of water. They cost less than a minute in total and helped more than she expected.

The error log

The error log is the single most useful tool in this course. Maya's log had one row for every wrong answer and every lucky guess, with six columns: the date, the source, the topic, the cause, the fix and a review date. The cause column was the important one. Luis insisted that every error be put into one of six causes, because each cause has a different fix.

When Maya sorted her log, the result explained her plateau. Knowledge gaps, which she had been studying hard, were only about a fifth of her errors. Misreads and lookup errors together were more than half. She had been fixing the wrong problem. For the next two weeks, she spent less time on new content and more on reading discipline and lookups, and her scores rose into the mid-eighties.

The program below does the sorting. It counts errors by cause and by topic, ranks the causes by their share, and prints the review dates for each knowledge-gap error using a spaced schedule of 1, 3, 7 and 14 days.

# Error log review: causes, topics and spaced review dates
from collections import Counter
from datetime import date, timedelta

log = [  # (date, topic, cause)
    ("2026-09-07", "venting", "misread"), ("2026-09-07", "hydronic", "knowledge gap"),
    ("2026-09-08", "water supply", "lookup error"), ("2026-09-09", "drainage", "lookup error"),
    ("2026-09-10", "backflow", "misread"), ("2026-09-12", "water heaters", "calculation slip"),
    ("2026-09-13", "hydronic", "knowledge gap"), ("2026-09-13", "venting", "misread"),
    ("2026-09-14", "rigging", "knowledge gap"), ("2026-09-15", "water supply", "lookup error"),
    ("2026-09-16", "drainage", "misread"), ("2026-09-17", "storm", "time pressure"),
    ("2026-09-19", "fixtures", "second-guessing"), ("2026-09-20", "water supply", "lookup error"),
    ("2026-09-21", "venting", "misread"), ("2026-09-22", "steam", "knowledge gap"),
]
causes = Counter(c for _, _, c in log)
total, running = len(log), 0
print(f"{total} errors by cause")
for cause, n in causes.most_common():
    running += n
    print(f"  {cause:16s} {n:2d}  {100 * n / total:5.1f} %   cumulative {100 * running / total:5.1f} %")
print("Topics:", ", ".join(f"{t} {n}" for t, n in Counter(t for _, t, _ in log).most_common(4)))
print("Spaced review for knowledge gaps:")
for d, topic, cause in log:
    if cause == "knowledge gap":
        start = date.fromisoformat(d)
        days = ", ".join((start + timedelta(k)).strftime("%b %d") for k in (1, 3, 7, 14))
        print(f"  {topic:9s} from {d}: {days}")

16 errors by cause
  misread           5   31.2 %   cumulative  31.2 %
  knowledge gap     4   25.0 %   cumulative  56.2 %
  lookup error      4   25.0 %   cumulative  81.2 %
  calculation slip  1    6.2 %   cumulative  87.5 %
  time pressure     1    6.2 %   cumulative  93.8 %
  second-guessing   1    6.2 %   cumulative 100.0 %
Topics: venting 3, water supply 3, hydronic 2, drainage 2
Spaced review for knowledge gaps:
  hydronic  from 2026-09-07: Sep 08, Sep 10, Sep 14, Sep 21
  hydronic  from 2026-09-13: Sep 14, Sep 16, Sep 20, Sep 27
  rigging   from 2026-09-14: Sep 15, Sep 17, Sep 21, Sep 28
  steam     from 2026-09-22: Sep 23, Sep 25, Sep 29, Oct 06

The cumulative column is the key. When the top two or three causes add up to most of the errors, they are where the next week's effort goes. The topic count matters too: venting shows up three times, all as misreads, which suggests that Maya reads venting questions too quickly because she thinks she knows them.

Spaced review

Knowledge-gap errors need more than one fix. Material studied once is mostly forgotten within days; material reviewed at growing intervals sticks. Maya's schedule reviewed each gap 1, 3, 7 and 14 days after she first studied it, using a summary card and two or three new practice questions on the topic. If she got the review questions right, the topic moved to the next interval. If she got them wrong, it went back to the start. The schedule is simple to keep in a notebook or a calendar, and it ensures that early topics are not forgotten by exam day.

Mixing topics in practice sessions also helps. Practicing one topic at a time feels easier, but mixed practice, where each question could be from any area, is closer to the exam and trains the skill of recognizing what kind of question it is. Maya did focused practice when learning a topic and mixed practice for everything else.

Practice exams under exam conditions

A practice exam only tells you something if it is taken like the real one. Maya's rules for mock exams were strict: the full length, a timer, no phone, no notes beyond what the exam allows, no stopping to look things up, and marking only after the end. Within a day, she logged every error with its cause, and within a week, she had reviewed every topic that produced a knowledge gap.

She also tracked her scores on a chart, with a line at the pass mark and a second line at her target of 80 percent. The target gave her a margin for a bad day, a hard set of questions or nerves. A candidate who scores 72 percent on practice exams may pass, but the margin is thin; a candidate who scores 82 percent consistently will almost certainly pass.