---
language:
- en
license: other
task_categories:
  - tabular-regression
features:
- name: symbol
  dtype: string
- name: datetime
  dtype: string
- name: probability_light
  dtype: float64
- name: probability_convolution
  dtype: float64
- name: probability_rocket
  dtype: float64
- name: probability_encoder
  dtype: float64
- name: probability_fundamental
  dtype: float64
- name: probability
  dtype: float64
- name: sans_market
  dtype: float64
- name: volatility
  dtype: float64
- name: multiplier
  dtype: float64
- name: version
  dtype: int64
extra_gated_prompt: "To get access to this dataset, you must subscribe to Papers With Backtest. To subscribe, go to https://paperswithbacktest.com/ > Login > Choose Your Plan > Subscribe."
---
# Dataset Information

Monthly bankruptcy probability estimates for US-listed equities. Scores are sourced from SOV.AI's ensemble models and refreshed each month.

## Instruments Included

- 4,700+ US Stocks

## Dataset Columns

- `symbol`: Stock ticker symbol for each company.
- `datetime`: Month-end date of the bankruptcy prediction snapshot (YYYY-MM-DD).
- `probability_light`: Bankruptcy probability predicted by the LightGBM model.
- `probability_convolution`: Bankruptcy probability predicted by the convolutional model.
- `probability_rocket`: Bankruptcy probability predicted by the ROCKET time-series model.
- `probability_encoder`: Bankruptcy probability predicted by the encoder-only transformer model.
- `probability_fundamental`: Bankruptcy probability predicted by the fundamentals-driven model.
- `probability`: Ensemble bankruptcy probability averaged across contributing models.
- `sans_market`: Market-neutral bankruptcy probability adjustment supplied by SOV.AI.
- `volatility`: Monthly equity volatility metric produced by SOV.AI.
- `multiplier`: Scaling coefficient associated with the probability ensemble.
- `version`: Upstream SOV.AI model bundle version number.

## Data Splits

The data is provided as a single `train` split.

## Dataset Maintenance

The dataset is updated monthly by [Papers With Backtest](https://paperswithbacktest.com).