Step 324. Competing in a Domestic Qualifier — Chasing the Finals: The Total Score Beats the Hero

Level 4 — Professional | Difficulty ★★★★☆ | Estimated time: 2 days (half a day fixing strategy + 24-hour qualifier + results recording)

Prerequisites: Step 298’s past-exam analysis, the real-world check competitions of Steps 300 and 302, and Step 303’s Level 3 comprehensive evaluation complete. This chapter is not training — it’s competition day.

What you need: a registered domestic major competition (CODEGATE, CCE, etc.) qualifier, a team channel (Discord, etc.), a shared document, your exploit library and cheat sheets, and the problem-prioritization tool you’ll build today. Every competition scene and scoreboard is a screen example; the priority-calculation script is a measured tool you run with your own data.
Caution: the qualifier’s goal is not the honor of a single problem but the total score — reaching the finals (or setting the team’s best record). Every judgment in this chapter is organized by this one criterion.
⚠️ All exercises in this chapter are for your own lab and legal platforms only. Applying them to unauthorized systems is a crime. Read and follow the competition rules (team size, no external collaboration, no attacking challenge infrastructure) before the start.

With past-exam analysis (Step 298) you read the texture of the stage, and with two check competitions (Steps 300, 302) you finished the dress rehearsals. Now comes the real qualifier. For 24 hours, a dozen or so problems open, and teams across the country solve the same problems at the same time. Your skill here is already fixed — the only remaining variable is the rate at which that skill converts into total score.

What determines that conversion rate is operations. Who takes which problem, when to let go when stuck, who’s awake at dawn. Today you’ll turn that operation into rules and tools, walk into the qualifier, and come back out leaving results as records.

1. Learning Objectives

By the end of this chapter, you will be able to:

Explain and execute the four mechanisms of qualifier-day operations (team channel, problem assignment, time management, handoff rules)
Finish the initial enumeration of all problems within one hour, marking difficulty, category, and assignee
Compute problem priorities by score efficiency (expected score ÷ estimated time) and set the engagement order
Explain the feel of the finals cutline (comparing historical trends against expected total score)
Leave a results record right after the end (problems solved/times, unsolved-problem list, expected rank)

2. Background Knowledge — Today’s Tools and Concepts

Today’s Tools at a Glance

Category	Details
Language/environment	Python 3 (priority calculation), the competition platform, team channel (Discord), shared document
Today’s command	`python step324_qual_strategy.py` — compute problem priorities, expected total score, and handoff candidates
Concepts needed	Score efficiency, initial enumeration, handoff rules, cutline feel, rotating dawn sleep
Today’s deliverable	A qualifier-day strategy document + priority table + competition log + results record

2-1. The Structure of a Qualifier — Why "Sure Points Across Every Category"

A domestic major qualifier is usually a 24-hour online Jeopardy competition. Since its purpose is to select finalists, the problem composition is distinctive — easy problems most teams solve, medium problems that split pass from fail, hard problems only a few top teams open.

In this structure, the hard problems are not what decides the cutline. Most teams can’t solve the hard ones anyway, so who harvests the medium problems more, and faster is what makes the ranking. The essence of a qualifier is not "one hard problem" — the teams that advance are the ones sweeping up "sure points across every category."

So the qualifier’s enemy is not "problems you don’t know" but "missing problems you could have solved." A forensics problem nobody even opened because no one owned it, a crypto problem that went cold while the whole team clung to a stuck pwn — these leaks are what drops a team near the cutline. Every operational mechanism today is a device for blocking these leaks.

2-2. Score Efficiency — One Formula That Decides Engagement Order

When problems open simultaneously, the answer to "what do we grab first?" is calculation, not instinct. For each problem, estimate three numbers — point value, estimated time, and your team’s solving confidence (0–1). Then the order sorts itself with this formula.

expected score = points x confidence
score efficiency = expected score / estimated time
engagement order = highest score efficiency first

The reason confidence enters the formula matters. A 300-point problem with a 30% solve chance has an expected score of 90; a 100-point problem with 90% also has 90 — but the latter takes far less time. The early-qualifier goal is not a high scoreboard but the swift harvest of sure points. Being drawn to point values is instinct; seeing efficiency is training.

The cutline feel is born on top of this calculation too. If the historical finals cutline (estimable from competition retrospectives and public scoreboards) was, say, around 800 points, today’s question is "is our expected total score above that?" You’ll actually run this calculation in 3-3.

2-3. The Four Mechanisms of Team Operations — Channel, Assignment, Time, Handoff

On qualifier day, a team is a small organization, and an organization needs mechanisms. Here are the four verified across Level 3’s twelve competitions.

Mechanism	Rule	Accident it prevents
Team channel	One thread per problem, findings shared immediately, chat in a separate channel	Two people duplicating the same attempt
Problem assignment	One primary assignee per category; recruit/reinforce for uncovered categories before the start	Problems left untouched
Time management	Initial enumeration 1 hour, sync meetings every 3–4 hours, scoreboard only during meetings	Rhythm collapse from scoreboard addiction
Handoff rule	No progress for 2 hours → swap assignee or swap problem; rotating sleep at dawn	The whole team burning out on one problem

What the four mechanisms share is turning judgment into rules. A tired brain at competition hour 14 can’t make good judgments. What still works at that hour is a rule decided in advance. When "no progress for 2 hours → hand off" is written in a document, the rule extracts the pride that wants to keep holding on.

2-4. Rotating Dawn Sleep — The Hidden Battleground of a 24-Hour Competition

A team that stays awake all 24 hours of a 24-hour competition sees its efficiency drop below half for the final 8 hours. Yet qualifier medium problems tend to fall exactly in that final stretch — they open to the person who kept the problem close the longest. And that "person who kept it close" needs a clear head.

So assign staggered sleep blocks, one per team member. For a three-person team, schedule each person a different 4–5 hours of the night, and while someone sleeps, they leave a handoff document (attempts so far and next candidate attempts) in the thread for their problems. Sleep is not desertion — it’s part of shift work. This instinct will decide life and death at 48-hour international competitions (Step 328); the domestic qualifier is the practice stage for it.

3. Follow Along

3-1. Confirming Competition Info — Gathering the Strategy’s Input Values

Once your participation is confirmed, fill the first page of the strategy document with competition info. Step 298’s qualifier-prep checklist already exists; today you add three qualifier-specific items.

Qualifier info check (screen example):
[ ] Duration — exact start/end times (24 hours? longer?)
[ ] Scoring — static points, or dynamic scoring where points drop as more teams solve?
[ ] Finals slots — how many top teams advance (10? 20?)
[ ] Historical cutline — estimate of past qualifying scores (retrospectives, public scoreboards)

Why the scoring method matters: with dynamic scoring, solving fast is itself worth points, raising the value of the initial enumeration and quick strikes. With static scoring, order is purely a matter of efficiency. The same problem list has a different optimal engagement order depending on the scoring method.

3-2. The Day-Of Strategy Document — What Gets Fixed Before the Competition

Before the competition starts, hold a team meeting and fix a one-page strategy document. This is where the rules verified in Step 300’s rehearsal get written not in "future tense" but as "decided."

Qualifier-day strategy document (screen example):
- First hour after start: everyone enumerates all problems — record category/points/first-impression difficulty/primary assignee in the table
- Engagement order: score-efficiency order (paste in the tool output from 3-3)
- Handoff rule: no progress for 2 hours → post "handoff request" in the thread — swap assignee or swap problem
- Meetings: 15 minutes each at hours 4, 10, 16, and 22; scoreboard checks only during meetings
- Sleep rotation: A 02:00-06:30 / B 00:00-04:30 / C 04:30-09:00 (handoff documents required)
- 1 hour before deadline: final check for unsubmitted flags, tidy the solving records

The document’s value lies in its existence more than its content. Mid-competition, when everyone is exhausted and one problem won’t crack, the moment you start debating "what should we do," the debate itself burns points. With the document, the answer is already decided.

3-3. Computing Priorities — The Tool Produces the Engagement Order

Once the initial enumeration is done, feed the problem list into the script to compute the order. Save it as step324_qual_strategy.py — each row of the list is (problem name, category, points, estimated minutes, solving confidence).

# step324_qual_strategy.py — qualifier-day problem priority & cutline-feel tool
# Replace PROBLEMS with your team's data

# (problem name, category, points, estimated minutes, solving confidence 0~1)
PROBLEMS = [
    ("welcome",     "misc",      50,  10, 1.00),
    ("web-login",   "web",      100,  40, 0.90),
    ("rev-basic",   "rev",      100,  50, 0.70),
    ("crypto-rsa",  "crypto",   150,  60, 0.90),
    ("foren-pcap",  "forensics",150,  70, 0.60),
    ("web-ssti",    "web",      200,  90, 0.80),
    ("pwn-stack",   "pwn",      200,  90, 0.85),
    ("misc-osint",  "misc",     150,  45, 0.85),
    ("rev-packed",  "rev",      250, 150, 0.30),
    ("crypto-aes",  "crypto",   250, 120, 0.40),
    ("pwn-heap",    "pwn",      350, 240, 0.20),
    ("web-chain",   "web",      300, 180, 0.30),
]

print("=== Qualifier Problem Priorities (score efficiency = points x confidence / estimated time) ===n")
rows = []
for name, cat, pts, mins, conf in PROBLEMS:
    eff = pts * conf / mins          # expected points per minute
    exp = pts * conf                 # expected score
    rows.append((eff, name, cat, pts, mins, conf, exp))
rows.sort(reverse=True)

total_exp = 0
cum_time = 0
print(f"{'rank':<5}{'problem':<13}{'category':<11}{'pts':>5}{'est.':>6}{'conf.':>6}{'exp.pts':>8}{'cum.time':>10}")
for i, (eff, name, cat, pts, mins, conf, exp) in enumerate(rows, 1):
    total_exp += exp
    cum_time += mins
    h, m = divmod(cum_time, 60)
    print(f"{i:<5}{name:<13}{cat:<11}{pts:>5}{mins:>5}m{conf:>6.0%}{exp:>8.0f}{h:>6}h {m:>2}m")

print(f"nExpected total score: {total_exp:.0f} pts / max {sum(p[2] for p in PROBLEMS)} pts")
print(f"Expected score solving only the top 8: {sum(r[6] for r in rows[:8]):.0f} pts,"
      f" taking {sum(r[4] for r in rows[:8])//60}h {sum(r[4] for r in rows[:8])%60}m")

# Handoff-rule check: problems below 0.5 confidence are 'hand off after 2 hours without progress' candidates
print("n=== Handoff-Rule Candidates (confidence under 50% — swap assignee after 2h without progress) ===")
for _, name, cat, pts, mins, conf, _ in rows:
    if conf < 0.5:
        print(f"  {name} ({cat}, {pts} pts) — est. {mins} min, confidence {conf:.0%}")

Here’s the measured output from running it with the example data:

=== Qualifier Problem Priorities (score efficiency = points x confidence / estimated time) ===

rank problem       category      pts   est.  conf. exp.pts  cum.time
1    welcome       misc           50   10m  100%      50     0h 10m
2    misc-osint    misc          150   45m   85%     128     0h 55m
3    web-login     web           100   40m   90%      90     1h 35m
4    crypto-rsa    crypto        150   60m   90%     135     2h 35m
5    pwn-stack     pwn           200   90m   85%     170     4h  5m
6    web-ssti      web           200   90m   80%     160     5h 35m
7    rev-basic     rev           100   50m   70%      70     6h 25m
8    foren-pcap    forensics     150   70m   60%      90     7h 35m
9    crypto-aes    crypto        250  120m   40%     100     9h 35m
10   web-chain     web           300  180m   30%      90    12h 35m
11   rev-packed    rev           250  150m   30%      75    15h  5m
12   pwn-heap      pwn           350  240m   20%      70    19h  5m

Expected total score: 1228 pts / max 2250 pts
Expected score solving only the top 8: 892 pts, taking 7h 35m

=== Handoff-Rule Candidates (confidence under 50% — swap assignee after 2h without progress) ===
  crypto-aes (crypto, 250 pts) — est. 120 min, confidence 40%
  web-chain (web, 300 pts) — est. 180 min, confidence 30%
  rev-packed (rev, 250 pts) — est. 150 min, confidence 30%
  pwn-heap (pwn, 350 pts) — est. 240 min, confidence 20%

How to read it: look at three things. ① The order — pwn-heap, the highest-value problem (350 pts), is dead last. The table confirms it’s efficiency, not points, that sets the order. ② The top 8 problems’ expected score of 892 points in 7h 35m — if the historical cutline is around 800, the qualifying score lives inside the first 8 hours. That one line is the executive summary of qualifier operations: the golden time is operating the first 8 hours perfectly, and the remaining 16 hours are for stacking a safety margin above the cutline from problems ranked 9–12. ③ The four handoff-rule candidates — problems under 50% confidence get marked from the start as "problems you enter with a 2-hour timer set."

3-4. Running the Qualifier — 24 Hours as a Screen Example

Here’s a screen example reconstructing a real qualifier day by time slot. Your competition will flow differently, but the texture of the mechanisms working is the same.

[10:00] Competition starts. Everyone in enumeration mode — open the problems in your category, record in the table
[10:45] Enumeration complete (12 problems). Run the script, pin the priority table in the team channel
[11:00] Engage in order: welcome, misc-osint, web-login — one per assignee, three people
[13:20] welcome, web-login, crypto-rsa solved (cumulative 300 pts). No scoreboard peeking
[14:00] First meeting, 15 min — pwn-stack is slower than expected; decision: keep going
[16:05] pwn-stack solved. Engage foren-pcap — confidence 60%, 2-hour timer set
[18:00] Second meeting — no progress on foren-pcap. Handoff rule fires; handed to B
[20:40] B spots an odd session in the pcap — flag. The handoff rule's first harvest
[00:00] B starts sleeping. C on night duty — attempts crypto-aes (2-hour timer)
[02:10] No progress on crypto-aes — lets go per the rule, switches to rev-basic
[04:30] rev-basic solved. A wakes for shift, reads the handoff docs, engages web-ssti
[08:00] Third meeting — current expected rank is on the qualification boundary. Re-examine web-chain among the 4 remaining
[09:00] One hour to deadline — check unsubmitted flags, tidy records
[10:00] End. Results recording begins

How to read it: notice the two points where the handoff rule fired (18:00, 02:10). Both were surely "a little longer and it’ll crack" moments, but the rule made the call instead. foren-pcap got solved through the handoff; crypto-aes didn’t get solved, but those 2 hours went to harvesting another problem. The handoff rule is not a rule for abandoning problems — it’s a rule for rescuing time.

3-5. Record Immediately After the End — Capital for the Next Competition

Right after the end, before fatigue crashes in, set aside 30 minutes to record the results. These records are the raw material for Step 325 (finals prep) and Step 327 (gap analysis).

Qualifier results record template (screen example):
- Final score / expected rank (including pre-announcement estimate):
- Solved problems: name, assignee, engaged→solved times, where the decisive progress happened
- Unsolved problems: name, what was tried, where it stalled, the last hypothesis
- Handoff-rule firings: time, reason, outcome
- Operations review: 1 rule that worked well, 1 rule that broke, 1 thing to change next competition

Why "immediately"? Memories of where you stalled distort within a day. If it gets crushed into "I just didn’t know it," that problem stays forever unsolvable; if it’s written as "mistook a CBC bit-flipping type for ECB," that problem becomes a line on the special-training list. The resolution of your records is the resolution of your growth.

4. Missions & Exercises

Mission — Complete the Qualifier and Record the Results

Fill the competition-info checklist (3-1) with your competition’s real information, and estimate the historical cutline in advance.
Before the competition starts, fix the day-of strategy document (3-2) in a team meeting — engagement-order rule, handoff rule, sleep rotation, all in decided form.
After the initial enumeration, replace PROBLEMS in step324_qual_strategy.py with the real problem list, run it, and pin the priority table in the team channel.
Operate the 24 hours per the strategy document — leave records of handoff-rule firings and sleep rotation.
Complete the results record template (3-5) within 30 minutes of the end, and save every unsolved problem’s files and attempt history.

Exercises

Exercise 1. In a qualifier’s difficulty structure (the placement of easy/medium/hard problems), which tier actually decides the cutline — and what does that demand of team operations?

Exercise 2. Explain why the score-efficiency formula multiplies by "confidence," using the comparison in 3-3’s measured output between pwn-heap (350 pts, last place) and misc-osint (150 pts, 2nd place).

Exercise 3. Imagine the moment the "hand off after 2 hours without progress" rule collides with pride. Explain why this rule must substitute for judgment, connecting it to your cognitive state mid-competition.

Exercise 4. Explain what the 3-3 measured output — "top 8 problems: expected 892 pts / 7h 35m" — implies for time-allocation strategy across the qualifier’s 24 hours.

5. Model Answers & Completion Criteria

Mission Model Answer

Check against these verification criteria.

Information completeness: are the four fields — duration, scoring method, finals slots, cutline estimate — filled? In particular, did the static/dynamic scoring distinction feed into the engagement-order strategy?
Strategy document’s definiteness: is every rule written as "we do X," not "we’ll try to do X"? Does the sleep rotation have actual clock times?
Priority table in live use: was the script output pinned in the team channel, and is there a trace (the log’s time order) of the actual engagement order following that table?
Handoff records: are handoff-rule firings recorded with time, reason, and outcome? If it never fired once, double-check whether every problem really went smoothly.
Results-record resolution: does every unsolved problem have "where it stalled" and "the last hypothesis" written in at least one sentence?

Exercise Answers

Answer 1. The cutline is decided by the medium tier. Easy problems are solved by every team in contention, and hard problems go unsolved by most anyway, so neither differentiates. So team operations demand ① full-category coverage that engages every medium problem without omission (no assignee gaps) and ② shortening the time spent on each medium problem (score efficiency). The handoff rule that keeps hard problems from incinerating team firepower is the safety device for these demands.

Answer 2. Points are not a fixed value but a "when solved" value, so they must be converted to expected value by multiplying the solve probability before comparison is meaningful. In the measurement, pwn-heap is worth 350 points, but at 20% confidence its expected score is 70 and it needs an estimated 240 minutes. misc-osint is worth 150 points, but at 85% confidence its expected score is 128 and it takes 45 minutes. Converted to expected points per minute, the gap widens further — 2.3x the points means nothing in the face of an efficiency reversal. Since the early-qualifier goal is the swift harvest of sure points rather than the honor of high values, confidence is the formula’s core variable.

Answer 3. The feeling "a little longer and it’ll crack" is often emotion rather than grounded judgment — especially past competition hour 10, when fatigue and attachment to sunk time (sunk cost) contaminate judgment. The handoff rule is a device for executing, at a tired moment, a judgment made at a clear-headed moment. And a handoff is not abandonment — it’s a swap of eyes: the new assignee re-examines the premises you took for granted and skipped. The proof is foren-pcap in 3-4’s screen example, solved 2 hours 40 minutes after the handoff.

Answer 4. It means the qualifying score is made inside the first third of the total time (the first 8 hours), so the strategy’s center of gravity must shift to the opening. Concretely: ① arrive with peak condition (no all-nighter the night before), ② minimize non-productive activity (scoreboard, chatter) during the first 8 hours, and ③ populate this stretch with verified-confidence types, since these problems must succeed. Conversely, the last 16 hours are the safety-margin zone — problems ranked 9–12 and rotating dawn sleep go here, and only with this slack does attempting hard problems become rational.

Completion Criteria Checklist

[ ] I confirmed and documented the competition info (duration, scoring method, finals slots, cutline estimate)
[ ] I fixed the day-of strategy document in a team meeting before the competition
[ ] I marked every problem’s category, difficulty, and assignee within the 1-hour initial enumeration
[ ] I ran the priority script on the real problem list and applied it to the engagement order
[ ] I actually operated the handoff rule (2 hours) and sleep rotation, and recorded them
[ ] I finished the qualifier and confirmed finals advancement (or the team’s best record)
[ ] I completed the results record within 30 minutes of the end
[ ] I saved every unsolved problem’s files, attempt history, and stall points

6. Common Pitfalls & Fixes

Wall 1. The script’s output shows garbled characters

Symptom: you ran it on Windows and the Korean (or non-ASCII) text in the output is garbled.

=== 삁꼫잌 訝쀍큐  슬꼫뀡꼮  ===

Cause: the file is saved as UTF-8 but the terminal code page is CP949 (the Korean Windows default) (measured).

Fix: before running, switch the code page to UTF-8 with chcp 65001, or in PowerShell set $OutputEncoding = [Text.Encoding]::UTF8. More fundamentally, redirecting output to a file (python step324_qual_strategy.py > result.txt) and opening it in an editor is the reliable method. If it’s text you’ll paste into the team channel, this also prevents the accident of sharing it with broken encoding.

Wall 2. I don’t know how to set the confidence numbers — they’re just feelings

Symptom: the confidence column in PROBLEMS becomes numbers without basis.

Cause: that’s normal — confidence starts as a subjective estimate. The problem isn’t estimating; it’s the absence of a basis.

Fix: turn confidence’s basis into records. Step 298’s past-exam analysis and Steps 295–297’s competition logs hold "for this type, how many of the last N attempts did we solve, in how many minutes on average" — those numbers are your confidence. Solved 4 of the last 5 of a type? That’s 0.8. For types with no data, be conservative at 0.3 or below — overestimated confidence contaminates the whole priority table. After the competition, compare estimates against reality to calibrate the next estimates.

Wall 3. The initial enumeration runs past one hour

Symptom: "solving each problem a little bit" stretches the enumeration to two, three hours.

Cause: the boundary between enumerating and solving collapsed. Enumeration’s purpose is reading problems, not solving them.

Fix: define the actions allowed during enumeration — read the problem statement carefully, check the attached-file list, record category and points, one line of first-impression difficulty, assign a primary owner. A problem whose entrance is visible gets marked as an engagement-order candidate and you move on. Five minutes per problem is the ceiling. If you still run over, it’s a many-problem competition — split the enumeration itself by category. Not everyone needs to see every problem; each category owner enumerates and shares the table.

Wall 4. Nobody is awake at dawn

Symptom: the sleep rotation existed, but around 4 AM the whole team was asleep and progress froze for those hours.

Cause: the rotation defined only "sleeping hours," never "waking responsibility."

Fix: rewrite the rotation around duty, not sleep — "who is responsible for each time slot" is the primary column, and sleep is the margin around it. And make handoff documents (problem in progress, next candidate attempts) mandatory at shift changes — sleep without a document is desertion; sleep with a document is a relocation of duty. For teams smaller than three, give up the dawn stretch entirely and synchronize everyone to "sleep at 22:00, wake at 06:00" — that beats a forced rotation.

Wall 5. I got eliminated in the qualifier — this chapter’s meaning feels broken

Symptom: failed to reach the finals. Team morale sinks.

Cause: advancement is a matter of probability — the same skill can go either way near the cutline.

Fix: re-read the completion criteria — "reach the finals (or the team’s best record)." And look at the last field of the results record (3-5), the "operations review." If the rules worked and records remain, this competition is not a failure but data — you hold in your hands where points leaked and how far estimates diverged from reality. Step 325’s reproduction training proceeds regardless of whether you advanced. The qualifier’s unsolved-problem list is not an asset of advancing teams only — it belongs to every team that recorded it.

7. Summary

Today’s Concepts

Concept	One-line explanation
The qualifier’s structure	Three tiers: easy/medium/hard — the cutline is decided by the medium tier’s harvest rate
Score efficiency	Expected score (points x confidence) ÷ estimated time — the sole criterion for engagement order
Initial enumeration	Mark category, difficulty, and assignee for every problem within 1 hour — the leak-prevention device
Handoff rule	No progress for 2 hours → swap assignee or problem — a rule that rescues time
Sleep rotation	A schedule that assigns duty, not sleep — handoff documents are the core
Cutline feel	Comparing historical trends against expected total score — "the qualifying score lives in the first 8 hours"

Today’s Tools & Commands

Tool/command	What it does
`python step324_qual_strategy.py`	Compute problem priorities, expected total score, and handoff candidates
Day-of strategy document	One page fixing engagement order, handoff, meeting, and sleep rules
Per-problem threads	Immediate sharing of findings; prevents duplicate attempts
Results record template	Solved/unsolved problems, stall points, operations review — raw material for the next Steps
Handoff document	The device that turns sleep from desertion into a relocation of duty

The Core Instinct

A qualifier is not a test of skill — it’s a test of conversion rate. If Level 3’s twelve competitions built the skill, today’s operational mechanisms plug the holes where that skill leaks away before it becomes total score. Eyes that see efficiency instead of point values, hands that follow rules instead of pride, a schedule that divides 24 hours instead of enduring them — these three are the skill of a qualifier.

And whatever the result, the qualifier isn’t over until the record 30 minutes after the end. If you advanced, that record is the starting point of finals prep (Step 325); if you fell short, it’s the blueprint of next season. Scores vanish when the competition ends, but records stay with the team.

Once every box is checked, Step 324 is complete.