What's Changed

• docs: simplify navigation and book-guide wording by @stefan-jansen in #277
• ci: repair portable artifact qualification and audit cache security by @stefan-jansen in #278
• fix: release Windows PDF pipes without changing inherited handle modes by @stefan-jansen in #279
• fix: patch PDF dependencies and retain native cleanup failure evidence by @stefan-jansen in #280
• ci: bump astral-sh/setup-uv from 10.3.2 to 10.3.3 in the github-actions group by @dependabot[bot] in #281
• fix: harden Windows PDF fixture and prepare 0.2.1 by @stefan-jansen in #282
• Explain the first divergence between order traces by @stefan-jansen in #286

Full Changelog: v0.2.0...v0.2.1

What's Changed

• fix: propagate declared annualization through backtest reports by @stefan-jansen in #115
• fix: preserve and validate dated portfolio input snapshots by @stefan-jansen in #117
• fix: include initial capital in portfolio drawdown workflows by @stefan-jansen in #119
• feat: analyze owned prepared signal panels without fabricated prices by @stefan-jansen in #121
• test: bound SHAP model fixture parallelism without weakening checks by @stefan-jansen in #127
• fix: accept equivalent UTC adapters without weakening date agreement by @stefan-jansen in #129
• fix: compound rolling returns exactly through total loss by @stefan-jansen in #130
• fix: honor each portfolio drawdown query selection and threshold by @stefan-jansen in #123
• test: retain financial reference cases in standard collection by @stefan-jansen in #131
• fix: retain observed cash sessions in uncertainty analysis by @stefan-jansen in #133
• feat: expose exact runtime qcut signal buckets with version provenance by @stefan-jansen in #136
• fix: preserve observation identity and declared time axes in signal workflows by @stefan-jansen in #137
• fix: preserve signal datetime key identity across storage units by @stefan-jansen in #140
• fix: report actual paired samples at each IC horizon by @stefan-jansen in #144
• fix: preserve common dated samples in uncertainty comparisons by @stefan-jansen in #142
• feat: prepare owned common-sample signal panels with loss provenance by @stefan-jansen in #145
• fix: validate direct IC observations and selected values by @stefan-jansen in #147
• fix: retain defined IC point summaries and numeric JSON missingness by @stefan-jansen in #149
• feat: expose grouped IC observations and adjusted returns by @stefan-jansen in #151
• feat: expose daily quantile uncertainty and equal-date statistics by @stefan-jansen in #153
• feat: expose factor portfolio weights and supplied-return contributions by @stefan-jansen in #155
• fix: preserve finite quantile and pooled spread statistics by @stefan-jansen in #157
• feat: expose arbitrary quantile spreads and propagated errors by @stefan-jansen in #159
• fix: preserve exact signal persistence observation keys by @stefan-jansen in #161
• feat: expose dated quantile turnover on explicit observation clocks by @stefan-jansen in #163
• feat: expose dated full-cross-section rank autocorrelation by @stefan-jansen in #165
• feat: retain dated persistence through signal results and reports by @stefan-jansen in #167
• fix: retain empirical quantile observations through reports by @stefan-jansen in #169
• feat: add selected-factor cumulative wealth with shared kernels by @stefan-jansen in #171
• fix: render actual dated spreads and honest propagated-error bands by @stefan-jansen in #173
• fix: report undefined quantile monotonicity as unavailable by @stefan-jansen in #175
• fix: align IC report statistics and availability with retained observations by @stefan-jansen in #177
• test: verify actual viewport widths in existing report workflows by @stefan-jansen in #179
• fix: retain actual monthly IC coordinates and missing cells by @stefan-jansen in #181
• fix: render recorded dated persistence in signal plots and reports by @stefan-jansen in #183
• fix: keep exposure normalization unavailable at nonpositive equity by @stefan-jansen in #185
• fix: preserve unavailable turnover and render the declared active mean by @stefan-jansen in #187
• fix(report): preserve holding-bar units across report views by @stefan-jansen in #189
• feat(portfolio): analyze complete prepared holdings on observed keys by @stefan-jansen in #191
• fix(report): render native disclosure triangle correctly by @stefan-jansen in #193
• feat(portfolio): analyze static and dated sector exposures by @stefan-jansen in #195
• feat(portfolio): expose top holdings and signed concentration summaries by @stefan-jansen in #197
• feat(report): render prepared holdings with explicit availability by @stefan-jansen in #199
• fix(report): retain every observed monthly return year by @stefan-jansen in #201
• fix(report): explain unavailable sample normal fits by @stefan-jansen in #203
• fix(pdf): batch pages with a timestamp-safe renderer by @stefan-jansen in #204
• fix(report): render signed and unavailable PnL attribution safely by @stefan-jansen in #208
• fix(integration): preserve declared contract units in fill analytics by @stefan-jansen in #209
• fix(portfolio): keep flat-period profit factor unavailable by @stefan-jansen in #211
• test(viz): retain PDF lifecycle evidence through interpreter exit by @stefan-jansen in #214
• fix(integration): preserve unavailable trade profit factor through reports by @stefan-jansen in #213
• fix(report): honor disabled statistical inference across templates by @stefan-jansen in #217
• test(viz): wait for PDF process exit without inherited stdout EOF by @stefan-jansen in #218
• fix(ci): retry exact published-package install during index propagation by @stefan-jansen in #215
• fix(deps): require a Polars version that supports existing signal workflows by @stefan-jansen in #222
• fix(viz): own renderer processes across PDF close failures by @stefan-jansen in #223
• feat(signal): expose dated factor alpha beta and compounded annualization by @stefan-jansen in #220
• fix(cv): adopt declared calendar selection and preserve observation ownership by @stefan-jansen in #205
• Add owned signal event paths and complete event-returns reporting by @stefan-jansen in #225
• ci: qualify timeout guards and resolve compatible security updates by @stefan-jansen in #233
• fix(pdf): preserve rendering failures through owned browser cleanup by @stefan-jansen in #253
• feat(compat): add tested native-backed Empyrical migration workflows by @stefan-jansen in #231
• feat(metrics): add inspectable native-backed benchmark fragility by @stefan-jansen in #254
• feat(metrics): add inspectable GPD tail-risk workflow and migration by @stefan-jansen in #255
• Complete owned account inputs, actual execution costs and producer report workflows by @stefan-jansen in #251
• feat: complete supplied-holdings attribution workflows by @stefan-jansen in #256
• feat(portfolio): add FIFO round-trip reports and explicit intraday workflows by @stefan-jansen in #252
• feat(portfolio): complete supplied-market capacity analysis and reports by @stefan-jansen in #257
• Complete auditable factor preparation and Alphalens migration adapters by @stefan-jansen in #249
• fix: preserve explicit MathJax disable in isolated PDF exports by @stefan-jansen in #259
• Add explicit-expiry synthetic portfolios and Alphalens-to-Pyfolio adapters by @stefan-jansen in #250
• feat(portfolio): complete stress live and wealth-cone workflows by @stefan-jansen in #258
• feat(reports): reconcile complete Alphalens and Pyfolio workflows by @stefan-jansen in #260
• fix(pdf): preserve actual resource closure and worker shutdown evidence by @stefan-jansen in #261
• fix(reports): preserve literal titles and owned browser cleanup by @stefan-jansen in #262
• ci: bump virtualenv from 21.14.12 to 21....

What's Changed

• feat(metrics): add calibration diagnostics for probability forecasts by @stefan-jansen in #92
• ci: bump gitpython from 3.1.61 to 3.1.62 by @dependabot[bot] in #93
• fix: support factor timestamps and empty signal frames on Polars 2 by @stefan-jansen in #99
• fix(docs): retry transient link failures with bounded checks by @stefan-jansen in #105
• fix(stats): use consistent population moments for DSR by @stefan-jansen in #100
• fix(factor): align rolling attribution after missing rows by @stefan-jansen in #101
• fix(distribution): estimate Hill loss tails without gains by @stefan-jansen in #102
• fix(config): reject non-mapping documents with ValueError by @stefan-jansen in #103
• docs: verify legacy migration paths on published packages by @stefan-jansen in #95
• ci: bump the github-actions group across 1 directory with 3 updates by @dependabot[bot] in #96
• docs: adopt shared focused-contribution policy by @stefan-jansen in #98
• chore: prepare 0.1.9 maintenance candidate by @stefan-jansen in #106
• fix: respect signal turnover horizons and clarify migration boundaries by @stefan-jansen in #108
• fix: retain overnight exchange sessions in calendar-first CV by @stefan-jansen in #110

Full Changelog: v0.1.8...v0.1.9

What's Changed

• docs: verify Diagnostic task and book guidance by @stefan-jansen in #86
• chore: prepare Diagnostic 0.1.8 release by @stefan-jansen in #87

Full Changelog: v0.1.7...v0.1.8

What's Changed

• fix: replace placeholder-backed diagnostic outputs by @stefan-jansen in #70
• release: set Diagnostic version to 0.1.6.post1 by @stefan-jansen in #71
• docs: remove legacy evaluation branding by @stefan-jansen in #72
• release: prepare 0.1.6.post2 by @stefan-jansen in #73
• docs: remove obsolete evaluation package name by @stefan-jansen in #75
• docs: complete Diagnostic agent-guide audit by @stefan-jansen in #77
• ci: bump codecov/codecov-action from 7.0.0 to 7.1.1 in the github-actions group by @dependabot[bot] in #78
• Report HAC kernel and finite-sample correction by @stefan-jansen in #80
• chore: release ml4t-diagnostic 0.1.7 by @stefan-jansen in #81

Full Changelog: v0.1.6...v0.1.7

What's Changed

• docs: remove legacy evaluation branding by @stefan-jansen in #72
• release: prepare 0.1.6.post2 by @stefan-jansen in #73
• docs: remove obsolete evaluation package name by @stefan-jansen in #75

Full Changelog: v0.1.6.post1...v0.1.6.post2

What's Changed

• fix: replace placeholder-backed diagnostic outputs by @stefan-jansen in #70
• release: set Diagnostic version to 0.1.6.post1 by @stefan-jansen in #71

Full Changelog: v0.1.6...v0.1.6.post1

What's Changed

• Use valid repeated-measures inference for conditional IC by @stefan-jansen in #62
• Accept calendar-month label horizons by @stefan-jansen in #63
• Standardize diagnostic release and public surfaces by @stefan-jansen in #61
• Allow release workflow qualification by @stefan-jansen in #65
• Real-data integration tests have been skipping since the fixture repo was deleted by @stefan-jansen in #67
• ci: bump astral-sh/setup-uv from 10.0.1 to 10.1.0 in the github-actions group by @dependabot[bot] in #66
• fix: bind Diagnostic metadata and docs to releases by @stefan-jansen in #68

Full Changelog: v0.1.5...v0.1.6

What's Changed

• Establish public agent orientation policy by @stefan-jansen in #55
• maximum_drawdown returned an absolute distance, not a drawdown by @stefan-jansen in #56
• release: prepare ml4t-diagnostic 0.1.5 by @stefan-jansen in #57

Full Changelog: v0.1.4...v0.1.5

What's Changed

• fix: emit backward CV folds chronologically by @stefan-jansen in #53
• release: prepare ml4t-diagnostic 0.1.4 by @stefan-jansen in #54

Full Changelog: v0.1.3...v0.1.4