# ibnr > Gallery-centric probabilistic loss reserving: Bayesian MCMC and neural network methods with mandatory evaluation, over a long-format triangle data layer. ## Docs ### API Reference #### Triangle > The long-format triangle data layer - a tidy table of origin_period / dev_lag / eval_date cells over an ibis expression (duckdb or polars backend), carrying grain, cumulative/incremental, and units metadata. - [Triangle](https://ekthesage.github.io/ibnr/reference/Triangle.html): A reserving triangle over an ibis table expression - [TriangleMeta](https://ekthesage.github.io/ibnr/reference/TriangleMeta.html): Triangle-level metadata. Carried through every transformation #### Triangle transforms & interop > Cumulative/incremental conversion, grain changes, as_of() backtest slicing, and lossless round-trips with chainladder-python and bermuda. - [Triangle.from_long](https://ekthesage.github.io/ibnr/reference/Triangle.from_long.html) - [Triangle.from_chainladder](https://ekthesage.github.io/ibnr/reference/Triangle.from_chainladder.html) - [Triangle.to_chainladder](https://ekthesage.github.io/ibnr/reference/Triangle.to_chainladder.html) - [Triangle.from_bermuda](https://ekthesage.github.io/ibnr/reference/Triangle.from_bermuda.html) - [Triangle.to_bermuda](https://ekthesage.github.io/ibnr/reference/Triangle.to_bermuda.html) - [Triangle.to_cumulative](https://ekthesage.github.io/ibnr/reference/Triangle.to_cumulative.html) - [Triangle.to_incremental](https://ekthesage.github.io/ibnr/reference/Triangle.to_incremental.html) - [Triangle.with_dev_grain](https://ekthesage.github.io/ibnr/reference/Triangle.with_dev_grain.html) - [Triangle.with_origin_grain](https://ekthesage.github.io/ibnr/reference/Triangle.with_origin_grain.html) - [Triangle.as_of](https://ekthesage.github.io/ibnr/reference/Triangle.as_of.html) - [Triangle.latest_diagonal](https://ekthesage.github.io/ibnr/reference/Triangle.latest_diagonal.html) - [Triangle.filter](https://ekthesage.github.io/ibnr/reference/Triangle.filter.html) - [Triangle.select_fields](https://ekthesage.github.io/ibnr/reference/Triangle.select_fields.html) - [Triangle.to_wide](https://ekthesage.github.io/ibnr/reference/Triangle.to_wide.html) - [Triangle.to_pandas](https://ekthesage.github.io/ibnr/reference/Triangle.to_pandas.html) - [Triangle.to_polars](https://ekthesage.github.io/ibnr/reference/Triangle.to_polars.html) - [Triangle.execute](https://ekthesage.github.io/ibnr/reference/Triangle.execute.html) - [Triangle.count](https://ekthesage.github.io/ibnr/reference/Triangle.count.html) - [Triangle.with_expr](https://ekthesage.github.io/ibnr/reference/Triangle.with_expr.html) - [Triangle.validate](https://ekthesage.github.io/ibnr/reference/Triangle.validate.html) #### Gallery > Discover, retrieve, and fit reserving models through the GalleryEntry contract. Every entry - Bayesian, neural, or statistical - produces a PredictiveDistribution. - [gallery.list](https://ekthesage.github.io/ibnr/reference/gallery.list.html) - [gallery.get](https://ekthesage.github.io/ibnr/reference/gallery.get.html) - [gallery.fit](https://ekthesage.github.io/ibnr/reference/gallery.fit.html) - [gallery.GalleryEntry](https://ekthesage.github.io/ibnr/reference/gallery.GalleryEntry.html) #### Held-out evaluation > The four steps between a fitted entry and a leaderboard row: which cells a fit at as_of is scored on, one model's arrays at those cells, the cross-model intersection, and the board. - [gallery.next_diagonal](https://ekthesage.github.io/ibnr/reference/gallery.next_diagonal.html) - [gallery.CohortForecast](https://ekthesage.github.io/ibnr/reference/gallery.CohortForecast.html) - [gallery.Absence](https://ekthesage.github.io/ibnr/reference/gallery.Absence.html) - [gallery.align_panel](https://ekthesage.github.io/ibnr/reference/gallery.align_panel.html) - [gallery.leaderboard](https://ekthesage.github.io/ibnr/reference/gallery.leaderboard.html) - [gallery.SCORE_DIRECTION](https://ekthesage.github.io/ibnr/reference/gallery.SCORE_DIRECTION.html) - [gallery.stack](https://ekthesage.github.io/ibnr/reference/gallery.stack.html) #### Chain ladder & reserve risk > The distribution-free chain ladder (Mack 1993) and its run-off MSEP, plus the one-year claims development result - the Solvency II reserve-risk view. cdr_methods() is the option surface: which model generates next year's diagonal, how the reserve is re-estimated once it exists, and what each route has actually been validated against. - [kernels.MackFit](https://ekthesage.github.io/ibnr/reference/kernels.MackFit.html) - [kernels.fit_mack](https://ekthesage.github.io/ibnr/reference/kernels.fit_mack.html) - [kernels.simulate_ultimates](https://ekthesage.github.io/ibnr/reference/kernels.simulate_ultimates.html) - [kernels.cdr_methods](https://ekthesage.github.io/ibnr/reference/kernels.cdr_methods.html) - [kernels.get_cdr_method](https://ekthesage.github.io/ibnr/reference/kernels.get_cdr_method.html) - [kernels.one_year_cdr](https://ekthesage.github.io/ibnr/reference/kernels.one_year_cdr.html) - [kernels.simulate_one_year_cdr](https://ekthesage.github.io/ibnr/reference/kernels.simulate_one_year_cdr.html) - [kernels.rereserve](https://ekthesage.github.io/ibnr/reference/kernels.rereserve.html) - [kernels.cdr_risk_measures](https://ekthesage.github.io/ibnr/reference/kernels.cdr_risk_measures.html) - [kernels.CDRResult](https://ekthesage.github.io/ibnr/reference/kernels.CDRResult.html) - [kernels.MackDiagonal](https://ekthesage.github.io/ibnr/reference/kernels.MackDiagonal.html) - [kernels.ODPBootstrapDiagonal](https://ekthesage.github.io/ibnr/reference/kernels.ODPBootstrapDiagonal.html) - [gallery.GalleryDiagonal](https://ekthesage.github.io/ibnr/reference/gallery.GalleryDiagonal.html) #### Evaluation kernels > The unifying predictive-distribution output type and the cross-backend posterior-parity tooling. - [kernels.PredictiveDistribution](https://ekthesage.github.io/ibnr/reference/kernels.PredictiveDistribution.html) - [kernels.ParityReport](https://ekthesage.github.io/ibnr/reference/kernels.ParityReport.html) - [kernels.compare_posteriors](https://ekthesage.github.io/ibnr/reference/kernels.compare_posteriors.html) #### Wire format > The single place an ibnr result becomes bytes - one Arrow IPC stream per artifact, uncompressed by default, plus a one-way JSON summary for callers that cannot take the draws. - [kernels.to_arrow](https://ekthesage.github.io/ibnr/reference/kernels.to_arrow.html) - [kernels.from_arrow](https://ekthesage.github.io/ibnr/reference/kernels.from_arrow.html) - [kernels.peek_kind](https://ekthesage.github.io/ibnr/reference/kernels.peek_kind.html) - [kernels.to_summary](https://ekthesage.github.io/ibnr/reference/kernels.to_summary.html) - [kernels.CODEC_VERSION](https://ekthesage.github.io/ibnr/reference/kernels.CODEC_VERSION.html) - [kernels.CONTENT_TYPE_ARROW](https://ekthesage.github.io/ibnr/reference/kernels.CONTENT_TYPE_ARROW.html) - [kernels.CONTENT_TYPE_JSON](https://ekthesage.github.io/ibnr/reference/kernels.CONTENT_TYPE_JSON.html) - [kernels.DEFAULT_QUANTILES](https://ekthesage.github.io/ibnr/reference/kernels.DEFAULT_QUANTILES.html) #### Retrospective harness > Parallel company x line backtesting with staged sampler escalation - the compute seam behind the study scripts. - [kernels.harness.run_retro](https://ekthesage.github.io/ibnr/reference/kernels.harness.run_retro.html) - [kernels.harness.run_task](https://ekthesage.github.io/ibnr/reference/kernels.harness.run_task.html) - [kernels.harness.RetroTask](https://ekthesage.github.io/ibnr/reference/kernels.harness.RetroTask.html) - [kernels.harness.SamplerSettings](https://ekthesage.github.io/ibnr/reference/kernels.harness.SamplerSettings.html) - [kernels.harness.ConvergenceGates](https://ekthesage.github.io/ibnr/reference/kernels.harness.ConvergenceGates.html) - [kernels.harness.precompile](https://ekthesage.github.io/ibnr/reference/kernels.harness.precompile.html) - [kernels.harness.default_max_workers](https://ekthesage.github.io/ibnr/reference/kernels.harness.default_max_workers.html)