Get cached value if it exists and hasn't expired

Verify that the given agent can be run by the game. Args: agent: The agent to verify Returns: Boolean indica

Factory fixture for creating mock players.

Determine the winner of the game based on the result output. Modifies the stats object in place. Args: agents: List of ag

Given the observation / recap, update the codebase

Parse a single simulation log to determine the winner. Args: log_path: Path to the log file sim_meta: Simulation meta

Helper to create a JSON simulation result file.

Get logger. Use this instead of `logging.getLogger` to ensure that the logger is set up with the correct handlers.

Get value from nested dictionary using dot notation path Args: path: Dot-separated path like "config.tournament.rounds"

Helper to create a simulation result file.

Create a simulation log file. Args: results: List of (player_num, name, rank) tuples

Create a simulation log file. Args: winner_key: "A" or "B" to indicate which player won is_coin_flip: If Tru

Convert Bradley-Terry strength to Elo rating. Formula: R_i = R_0 + (β/ln(10)) * s_i where β = 400 (ELO_SLOPE), R_0 = 1200 (ELO_BASE)

Helper to create a simulation log file.

Returns True if running in AWS Batch, False otherwise.

Helper to create a simulation log file.

Create a file with given content on a Docker container. Uses a temporary file on the local filesystem for the transfer.

Return a git diff with any file sections mentioning binary content removed.

Check if a directory contains metadata.json and is therefore a game folder

Decode raw sim bytes into a :class:`ReplayData`. Input format is opaque.

Main execution function that runs all rounds.

Helper to create a TXT simulation result file.

Write text to file atomically to prevent corruption on interruption.

Subclasses implement their game-specific logic here. This is the low level implementation, you probably want to use run_round instead, which

Calculate expected score for player A against player B.

Fit Bradley-Terry model.

Get docker container ID with the game code installed.

Build mapping from AWS Batch job ID to log folder path and round info

Create a player log file with connection info.

Create a results file. Args: results: List of (rank, bot_name, total_score) tuples

Get git tag name for the version of the codebase at the given round.

Copy a file or directory from the local filesystem to a Docker container. The copy operation is recursive for directories. Be extremely

Find all game folders by locating metadata.json files Uses parallel loading to speed up metadata.json reading for many game folders. Results

Clean all uncommitted changes. If base_commit is provided, reset to that commit. Then apply the patch to the codebase.

Pre-process YAML content to resolve all !include directives. This will take the yaml content as a string and return the same string with all !

Run a single round of the game with the given agents. Returns the log output, result output, and winner name. All bookkeeping should

Update Elo rating based on game outcome. actual: 1.0 for win, 0.5 for draw, 0.0 for loss k: K-factor, typically 32 for regular players,

Add percentage labels inside stacked bars. Args: ax: The matplotlib axis stacked_values: List of lists, where each in

Add total value labels at the end of bars. Args: ax: The matplotlib axis totals: List of total values for each bar

Derive the arena from a tournament folder name. e.g. ``PvpTournament.RobotRumble.r1.s1.p2.flail.luisa.260702175500`` -> ``RobotRumble``.

Helper to create round log directory.

Helper to create round log directory.

Create simulation log files (one per agent). Args: round_dir: Directory to create log files in scores: List

Get cached metadata or load it if not cached

Read diff data from changes_r{round}.json file, fallback to metadata.

Best-effort sim index from a filename: the last integer in its stem (``sim_18`` -> 18, ``round_1`` -> 1), else 0. ``discover_games`` de-duplicates

Add a file handler to the logger with standard formatting. Returns: The FileHandler that was added (for later cleanup).

Evaluate multiple instances with optional parallel processing. Args: instances: List of instances to evaluate n_worke

Fit AR(p) on logit(scores) by conditional Gaussian MLE (CSS). scores: (T, R) p: AR order (default 1) Returns: (gammas_hat: (p,), mus