OptionaldatasetId of a dataset you uploaded, in place of instrument. Exactly one of
instrument/datasetId is required. Pairs with exchangeId: 'user'.
OptionaldatasetOptional specific version of datasetId; omit to use its current version. Requires datasetId.
OptionalequityWhich trial curves to retain and how their points are transformed.
Exchange id, e.g. binance, or the reserved value user when sweeping against datasetId.
Range start (ISO-8601, ISO DATE, or BASIC ISO DATE).
OptionalinstrumentInstrument symbol, e.g. BTC/USDT. Exactly one of instrument/datasetId is required.
OptionalobjectiveThe metric to optimize and rank by; omit to keep the platform default
('sharpe'). It is also what Sweep.sensitivity aggregates unless
told otherwise.
The grid: one ParamAxis per strategy property to vary. At least one.
OptionalsamplerHow the grid becomes the list of vectors actually run. Omit to keep the platform default, the full cross product.
OptionalsamplesHow many vectors to draw for 'random' and 'lhs'; ignored by 'grid'.
OptionalseedReproducibility seed. Omit to let the platform generate one and report it back on Sweep.accepted, so a randomly sampled sweep can be replayed exactly by submitting the same seed again.
Strategy source code (Java), compiled once and reused by every trial.
Range end (same formats as from; must be later than from).
OptionalwalkOpt into walk-forward validation, which changes both what runs and the shape of the answer. Omit to run an ordinary sweep.
A parameter sweep over one instrument and one window: the same strategy run once per parameter vector, scored and ranked against a single objective.
Sweep against a dataset you uploaded instead of an exchange instrument by replacing
instrumentwithdatasetId(andexchangeId: 'user'):