Playtest Notebook
Record matches you actually played to compare deck versions and strategic plans. Results stay in this browser unless you export them. Crack Packs does not receive your match log.
Offline notebook
Only public notebook application files are downloaded. Matches, drafts and templates stay in existing unencrypted device storage, never in the application-file cache. No account sync or background upload is enabled. Browser eviction/private browsing can remove local data; export a backup. Card searches, account tools and the rest of the site still require a connection.
Device-only records; account sync is not enabled. Checking offline file availability.
Run a cleaner strategy test
Write the test set and strategy before you start. Change one meaningful variable at a time, keep the deck format consistent, and log every match in the sample, including losses and draws. Use the same test-set label for variants you intend to compare; split opponents and first/second results before changing the list. These results are observational, not a controlled tournament study.
Detailed entry. Layout settings contain no match data.
Event rounds
Review before saving
No event setup captured. Unsaved rounds stay in this tab only.
Erase local playtest data
Deletes this browser's match history, private match images, entry draft, saved protocols and layout preferences, plus this tab's pending opening reference and unsaved notebook entries and image previews. Saved decks, collections, sign-in and marketplace data stay unchanged. This is not account deletion or secure forensic erasure. Downloaded files, other devices and other tabs' unsaved/session data must be removed separately.
Deletion scope
No deletion reviewed. Clear notebook only removes saved match records; this option also removes the listed drafts and preferences.
Encrypted backup and restore
Encrypts a portable backup in this browser with a passphrase. The passphrase is never saved and nothing is uploaded; move the encrypted file to another device to restore it there. This is not automatic cloud sync.
No matches logged yet.
Observed results
Wilson score range for W/(W+L), with z=1.96 for approximately 95% two-sided coverage under independent binary-sample assumptions. It is not a 95% chance that this deck's future win rate is inside the range. Repeated or selectively logged matches can violate those assumptions.
Categories with the same recorded opponent label and starting position. Adjustment includes only categories with decisive results in both selected variants and applies the same pooled decisive-count weights to each. Missing or unshared context is excluded, not imputed; player skill and other differences remain uncontrolled.
N is the number of recorded completed matches, including draws. n is wins plus losses: the binary rate excludes draws. These are match-level counts, not individual games or unique players. The 20/50-match checkpoints are practical review reminders, not statistical power, significance or proof of a winner.
Calculation method and assumptions
Rates use p = W / n with n = W + L; draws stay visible but are excluded from this binary calculation. The two-sided Wilson score interval uses z = 1.96 (approximately 95%), without continuity correction.
centre = (p + z*z/(2*n)) / (1 + z*z/n)radius = z * sqrt(p*(1-p)/n + z*z/(4*n*n)) / (1 + z*z/n)range = [max(0, centre-radius), min(1, centre+radius)]
No interval is shown when n = 0. Shared-stratum adjustment applies weights (n_A + n_B) / sum_shared(n_A + n_B) to each variant's rate within each shared category; it has no interval or causal interpretation. All figures describe only the current recorded and filtered sample. Missing or unlogged matches cannot be inferred. Self-selected variants, repeated opponents, deck changes and selective logging can bias results; no power calculation, significance test or predictive ranking is provided.
NIST Wilson method reference (online)Each row is descriptive evidence from your own log, not a win-rate forecast or proof that a strategy caused the result. The 95% Wilson interval shows uncertainty in a small sample; opponent selection and play skill can still bias it.
Opening policies by matchup and position
Actual recorded decisions only. Different captured criteria, source lists and rule snapshots stay separate. Missing observations and unknown opponents/positions remain visible; counts do not establish which policy wins more matches.
Compare two tested variants
Alongside the raw results, the comparison checks shared opponent and play/draw categories to reduce matchup-mix bias. Small or non-overlapping samples remain inconclusive.
Choose variants with the same game, ruleset, and test set. Comparisons are observational, not proof of causation.
Seeded resampling of raw match rates
Optional 95% percentile bootstrap under an independent-match approximation, not a forecast, causal estimate or significance test. Draws are counted but excluded. Paired/block assignments require a different resampling design and are not supported here. Opponent repetition, selective logging and deck/player changes can violate independence. Wilson ranges remain the primary view; sparse or all-identical samples can produce misleadingly narrow resampling ranges.
Compare variants before resampling. Nothing is uploaded or saved by resampling.
Resampling calculation
Within each variant, take n = W + L draws with replacement from its binary empirical sample for each resample. The reported ranges are the 2.5th and 97.5th percentiles of W*/n and 100*(rate_B* - rate_A*), using linear type-7 quantiles. The same pinned generator, seed, counts and resample count reproduce the same result; changing the seed does not add new played matches. No adjusted-rate, paired or clustered interval is supplied.
SciPy bootstrap method reference (online)Matchup matrix
Observed win-loss-draw records by tested variant and opponent label. Sparse cells are not predictive.