Experiment

Defined in fynance.research

class Experiment(name, spec=<factory>, code=None, seed=0, metrics=<factory>, series=None, created_at=<factory>, fynance_version=<factory>)

Bases: object

Serializable record of a single strategy experiment.

Parameters:

namestr

Short slug identifying the experiment (also the output sub-directory).

specdict, optional

Everything needed to reproduce the run: strategy description, params, walk-forward config, cost config and the data spec (e.g. {"kind": "synthetic-gbm", "seed": 7, "n": 1000}).

codestr, optional

The generated strategy source, stored verbatim for audit/reproduction.

seedint

Master seed driving every stochastic step.

metricsdict of str to float, optional

Summary metrics (sharpe/sortino/…); empty until a run fills it.

seriesdict of str to list of float, optional

JSON-friendly curves (e.g. equity/returns) for the report.

created_atstr

ISO-8601 UTC creation time (set automatically).

fynance_versionstr

Version of fynance that produced the experiment (set automatically).

Examples

>>> from fynance.research import Experiment
>>> e = Experiment(name="demo", seed=7, metrics={"sharpe": 1.5})
>>> e2 = Experiment.from_dict(e.to_dict())
>>> e2.name, e2.seed, e2.metrics["sharpe"]
('demo', 7, 1.5)

classmethod from_dict(data)

Rebuild an Experiment from a to_dict mapping.

classmethod load(path)

Load an Experiment from an experiment.json file.

save(output_dir, *, name=None)

Write <output_dir>/<name>/experiment.json and return its path.

Always writes under the caller-provided output_dir — never inside the package. Parent directories are created as needed.

Parameters:

output_dirstr or pathlib.Path

Base directory the artifact is written under.

namestr, optional

Sub-directory name; defaults to name.

Returns:

pathlib.Path

Path to the written experiment.json.

to_dict()

Return a JSON-serializable dict of the experiment.