> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tqx.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Write factor code

> Write valid formula or Python factor code without introducing future data.

The factor editor supports **Formula** and **Python** modes. Qube validates the code before it submits an analysis.

## Formula mode

Write one complete expression per non-empty line. Use lowercase field names and uppercase operator names.

```text theme={null}
RANK(close / DELAY(close, 20) - 1)
```

Formula mode does not allow assignments, imports, Python attributes, or control-flow keywords. Do not use `FUTURE_RETURNS`; it reads future data and is rejected by the factor quality gate.

## Python mode

Define one class that inherits from `Factor`. Return a `Series` from `calculate(self, factors)` and preserve the `symbol`/`date` MultiIndex.

```python theme={null}
class MomentumFactor(Factor):
    def calculate(self, factors):
        close = factors["close"]
        return close.groupby(level="symbol").transform(
            lambda series: series / series.shift(20) - 1
        )
```

Prefer `transform`, `pct_change`, and rolling operations that preserve the input index. `groupby(...).apply(...)` causes a validation error when it adds another grouping level to the result.

## Check the market contract

Use the [Operator reference](/research/factor/operator-reference) for the fields, supported expressions, and market restrictions available to your factor.

<Warning>
  An analysis can only be reproduced when the code, market, data fields, and analysis parameters are all compatible. Validate the signal on a different period before treating it as robust.
</Warning>