Indicator Sets

Indicator Sets are reusable collections of configured Indicator Instances. They let you prepare a feature configuration once and apply the same ordered choices to future Jobs.

Indicator type versus Indicator Instance

An indicator type defines the reviewed calculation, inputs, parameters, and outputs. An Indicator Instance is one configured use of that type.

For example, an RSI with one period and an RSI with another period are separate instances. DataCat allows multiple instances of the same indicator type in one set when their parameter choices serve different research purposes.

Creating a set

1. Create an Indicator Set and give it a recognizable name.
2. Use Add Indicator to browse the Registry-backed catalog.
3. Select one or more indicator types.
4. Review the new instances.
5. Adjust parameters in the Inspector.
6. Save the Indicator Set.

The Add Indicator flow can show types already used by the set because adding another configured instance is valid.

Using a set in a Job

The Job field is named Indicator Set. It opens a picker that previews the selected set’s indicators. No Indicator Set is a valid choice and produces an OHLCV-only dataset.

When the Job starts, DataCat freezes the selected instances and resolved parameters into the Job configuration. Editing or deleting the reusable Indicator Set later does not alter that started Job or its dataset.

Plan limits

The commercial limit applies to Indicator Instances per Job, not to the existence of an Indicator Set as a standalone object. DataCat checks the effective total while composing a set so a future Job does not silently exceed the active plan.

If a set is above the active limit, reduce its instance count or use a plan that supports the intended Job configuration. Dismissing an upgrade screen does not save an over-limit addition.

Designing useful sets

Prefer a set that represents one understandable research purpose instead of accumulating every available indicator.

Good patterns include:

1. Trend context with a few moving-average horizons and a trend-strength measure.
2. Momentum context with one or two oscillator families rather than many equivalent variants.
3. Volatility context with range-based and dispersion-based measures.
4. Feature-engineering sets that intentionally repeat one indicator across several periods.

Name sets after the intended workflow, market horizon, or experiment. A name such as BTC Trend 1h is easier to compare than Set 7.

Parameters and comparability

Parameters are part of Indicator Instance identity. Changing a period, smoothing type, or multiplier changes the resulting feature and can change its warmup behavior.

When comparing datasets, inspect indicator_definitions rather than relying only on a human-readable set name. The dataset persists resolved parameters and deterministic output-column mappings for the frozen Job snapshot.

Warmup and missing inputs

Many indicators need earlier bars before they can produce a value. DataCat writes NULL during warmup rather than inventing an output.

Missing or pending provider input can also affect indicator availability. The indicator output-state metadata distinguishes warmup, awaiting input, provisional values, derived values, and unavailable values. A numeric value alone is not the complete quality signal.

Choosing indicators

Use the Indicators section to understand interpretation and common mistakes. Parameter definitions, accepted bounds, generated output descriptions, and implementation facts come from the reviewed Indicator Registry on each indicator page.