Added bugfixes for release 1.0.0

[Thematiq]

No description provided.

[Copilot]

Pull request overview

This PR updates tie-handling semantics in win-table construction, expands test coverage around those behaviors, and adds supporting release artifacts (NOTICE, example notebook, dev dependency updates).

Changes:

• Adjust paired/local-ROPE win-table construction and tie-solver behavior (spread/add/forget) and add targeted tests.
• Update PyMC model input typing/casting for win counts.
• Add a Critical Difference Diagram example notebook, a NOTICE file for the reference implementation, and add graphviz to dev dependencies (with lockfile updates).

Reviewed changes

Copilot reviewed 8 out of 9 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
bbttest/bbt/alg.py	Changes tie-solver behavior (spread now assigns 0.5 per tie) and adjusts paired local-ROPE table construction.
bbttest/bbt/model.py	Updates win1/win2 typing and casts n/w1/w2 to int32 inside the PyMC model builder.
bbttest/bbt/_utils.py	Updates _validate_params to allow extra kwargs when the function declares **kwargs.
tests/bbt/test_alg.py	Adds tests for paired local-ROPE path and explicit tie-solver semantics for spread/add/forget.
tests/bbt/test__utils.py	Adds a test ensuring _validate_params ignores unexpected kwargs when **kwargs is present.
examples/02_critical_difference_diagram.ipynb	Adds an example notebook demonstrating CDD plotting with PyBBT.
pyproject.toml	Adds graphviz to the dev dependency group.
uv.lock	Lockfile update to include graphviz.
NOTICE.md	Adds NOTICE content for the referenced bbtcomp implementation/paper.

Comments suppressed due to low confidence (1)

bbttest/bbt/alg.py:124

• _solve_ties now produces fractional win counts for tie_solver="spread" (e.g., 0.5 per tie). Downstream, _mcmcbbt_pymc/_build_bbt_model uses a Binomial likelihood that requires integer n and integer observed wins, and it currently truncates win1/win2 to ints. As a result, spread will silently drop the 0.5 contributions (and can also make n inconsistent with the intended tie semantics). Consider either (a) converting fractional tables to an equivalent integer representation before sampling (e.g., scale w1/w2/n by 2 when halves are present), or (b) rejecting spread/fractional tables when using the Binomial model and requiring tie_solver="davidson" for tie-aware sampling.

def _solve_ties(table: np.ndarray, tie_solver: str) -> np.ndarray:
    if tie_solver == "davidson":
        return table
    if tie_solver == "spread":
        tie_val = table[:, TIE_COL] / 2
    elif tie_solver == "add":
        tie_val = table[:, TIE_COL]
    else:
        tie_val = 0
    table[:, ALG1_COL] += tie_val
    table[:, ALG2_COL] += tie_val
    return table

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.