Local contest-log screening for automation-like operating patterns.
Rows are review signals, not final adjudication. Methodology and version notes are available below.
Peer reference corpus
Historical Single-Op contest-log reference base
CW · RTTY · SSB
LOGS
—
processed reference logs
QSOs
—
processed reference QSOs
GROUPS
—
contest/category peer groups
CONTESTS
14
contests covered
Embedded Single-Op reference corpus covering major and regional contest profiles. BotScan is one unified local tool: it detects contest, mode and category from each uploaded log, then applies the matching CW, RTTY or SSB review logic. The log/QSO top-card footprint is kept conservative to avoid double-counting overlapping calibration families.
Peer comparison is category-aware: contest, year, mode, power, band and assistance are used where available. Automation Evidence is independent from claimed rank. Logger/CREATED-BY clustering is context only and never adds score by itself.
1Drop a Cabrillo log to analyze
📄
Drop a Cabrillo log here, or click to select
Accepts .log, .cbr, .txt — single-op logs only
Public Results Library
Load processed BotScan batch results from logs.lu9esd.com/results. Original public Cabrillo logs stay in R2; this page reads the generated JSON results.
Waiting for public library…
1. Choose contest
Choose a contest family to select mode and year.
Waiting for public library…
Batch scan results
Contest / year / category
Waiting for logs…
Waiting for logs…
Sorted by Automation Evidence first, then Review Priority, claimed rank and score.
Click a row to open the full review. Default view shows review-relevant cases. Rank is claimed/pre-adjudication and is computed from CLAIMED-SCORE inside each loaded category group. JSON is kept as the technical/dev evidence export; CSV/PDF are for easier human review.
2File integrity
3Automation-like pattern markers
4Full peer-group comparison
5Detailed operating statistics
Methodology
Cabrillo contest-log review methodology.
BotScan is a first-stage analytical detector. It highlights logs that deserve deeper review; it does not replace committee judgment.
1. Input and eligibility
BotScan reads Cabrillo logs locally in the browser. It parses the Cabrillo header and QSO lines, removes exact repeated QSO-line artifacts from behavioral metrics, separates non-scored records, rejects Multi-Op categories for Single-Op review, and analyzes only logs with enough cleanly parsed QSOs.
2. Metrics
The analysis combines file integrity, rest breaks, continuous operating blocks, frequency-change rate, run/S&P texture, minute-level frequency spread, band movement, rate context, duplicate/reworked-station context and frequency-detail quality. Each metric is interpreted by mode and contest profile, not as a single global rule.
3. Peer comparison
Each contest and category has its own historical peer profile. BotScan compares the uploaded log against matching Single-Op groups by contest, year, mode, power, band and assistance where available, so EA-RTTY, CQ-160-CW, CQ-WW, WPX, WAE, ARRL and IARU are not judged with one generic threshold set.
4. Evidence logic
BotScan avoids relying on one weak signal. RED-level review requires hard timing, scanner/no-run structure, hard minute-level spread, or multiple independent evidence families. In RTTY Single-Op review, 5+ distinct logged frequencies in one minute remains a high-priority hard spread signal. Frequency-spread display now starts at 4+ and dynamically reveals higher 7+/8+ levels only when they are present. Three-frequency minutes are retained only as internal/informational context. In CW, isolated single-band 5-frequency minutes are treated as aggressive S&P context; repeated 5+ or any 6+ minute remains high-priority review. In SSB, stable dual-run/2BSIQ and aggressive S&P are handled separately, and SSB scatter is compared against the matching historical peer-group peer reference/max before it affects review level. Fast duplicates and same-band reworked stations remain context only and do not create RED by themselves.
5. Batch ranking context
In batch mode, claimed score is used only to show pre-adjudication rank inside the loaded category group. Automation Evidence is calculated separately from claimed rank.
6. Limitations
Cabrillo timestamps are usually minute-level, so the tool is a screening layer. Strong flags should be followed with higher-resolution evidence when available: ADIF seconds, raw RBN timing, audio, station layout, operator declarations and written explanation.
Version history
User-visible changes and review workflow notes.
v0.13.5
2026-06-15
Allowed the Public Results Library to read the BotScan R2 mirror at logs.lu9esd.com while keeping uploads and unauthorized cross-origin calls blocked.
Keeps public log reads as GET/HEAD only; private logs still stay local in the browser.
No analysis thresholds, evidence families or verdict logic were changed.
v0.12.13
2026-06-15
Pure search-and-pounce operation below the high-volume floor is treated as operating context rather than a standalone review family.
Standalone zero-run / high-FCR S&P markers require the higher 1,500-QSO floor, while spread-backed cases remain reviewable.
Validated against the 4,883-log corpus and reference set with no tier changes to known RED/YELLOW/reference logs.
v0.12.x
2026-06-12 to 2026-06-15
Refined SSB, mixed-mode and S&P context so normal operating style is displayed as context instead of automation evidence.
Improved bilingual metric labels, dynamic peer-comparison rows and frequency-spread display starting at 4+ frequencies.
Kept fast duplicates, reworked-station counts, rest/endurance and run-anchoring rows as context unless paired with independent hard evidence.
v0.11.x
2026-06-09
Integrated the CW peer-aware historical calibration layer into the main runtime.
Kept CW, RTTY and SSB calibration separated by mode instead of mixing all modes into one threshold set.
Expanded the visible reference footprint and made top-card log/QSO counts reflect the processed local calibration base.
v0.10.x
2026-06-04 to 2026-06-06
Moved the browser build into a modular source package with separate core, rules, UI, styles and build output.
Cleaned public batch/report output so software metadata remains raw per-log context and is not aggregated or ranked.
Improved batch finalization, version alignment and visible methodology notes.
v0.9.x
2026
Expanded batch scanning, claimed-rank context, export options and review-priority sorting.
Refined evidence families for endurance, scanner texture, run anchoring, frequency spread and duplicate/rework context.
v0.8.x
2026
Introduced the self-contained browser analyzer with local Cabrillo parsing and peer-percentile comparison.
Added evidence export, PDF/CSV reports, file-integrity checks and initial dark UI review panels.
v0.7.x
2026
Established the first detection metrics for rest behavior, operating blocks, frequency-change rate, run texture and same-minute frequency spread.
Started historical calibration work for Single-Op CW/RTTY contest logs and early review thresholds.
v0.6.x
2026
Expanded the calibrated metric set with run structure, rest/block patterns, frequency-change rate, minute-level spread and same-band rework metrics.
Introduced objective signal families for concurrent multi-frequency texture and no-run/autoresponder-like S&P structure, with mode-aware interpretation.
Moved software/logger information into supporting context only, never as an independent signal.
v0.5.x
2026
Added consecutive same-frequency run length, runs of 3+ and 5+ QSOs, hourly rate variation and 3+/4+/5+/6+ frequency-minute counters.
Added contest-profile rule handling and clearer natural-language marker output.
v0.1–0.4
2026
Started as a local Python Cabrillo analyzer, then moved to a browser-based HTML tool with drag-and-drop loading.
Added contest-header detection, category-aware comparison and the first RED / YELLOW / GREEN review levels.