Description

The up-to-date combined file is fear-greed.csv at the repo root.

| Column | Notes | |------------|-------------------------------------------------------| | Date | ISO YYYY-MM-DD | | Fear Greed | Float score 0-100 (CNN era) / integer (pre-2021 era) | | Rating | CNN label (extreme fear, fear, neutral, greed, extreme greed). CNN supplies this directly for 2021-02-01+; for older rows it is computed from the score using CNN's own band cutoffs. |

Coverage:

• 2011-01-03 through 2021-01-29 is frozen, sourced from datasets/archive/fear-greed-pre-2021.csv (derived from the hackingthemarkets + openstockalert combined file; see history sections below).
• 2021-02-01 onward is regenerated from the live CNN endpoint on every run.

Automation

A GitHub Actions workflow (.github/workflows/update-data.yml) rebuilds fear-greed.csv Monday through Friday at 23:00 UTC (after US market close) and also supports manual runs via workflow_dispatch. It commits the diff directly to main as github-actions[bot].

Scripts:

• scripts/fetch_cnn.py - fetches CNN endpoint, writes datasets/cnn_fear_greed.csv and json/cnn_output.json.
• scripts/build_combined.py - merges the frozen archive with the CNN tail into fear-greed.csv.
• scripts/convert_macromicro.py - one-shot helper (not in CI) that regenerates datasets/macromicro_fear_greed_data.csv from a manually captured json/macromicro.json.

uv sync
uv run python scripts/fetch_cnn.py        # -> datasets/cnn_fear_greed.csv + json/cnn_output.json
uv run python scripts/build_combined.py   # -> fear-greed.csv

Dependencies are declared in pyproject.toml and resolved in uv.lock. build_combined.py aborts rather than writing if the combined output would be shorter than the archive, contains duplicate dates, or is more than 7 days stale.

The third-party comparison datasets under datasets/ (openstockalert, hackingthemarkets, alexey-formalmethods, macromicro, archive.org) are historical snapshots and are not refreshed by the workflow.

Datasets

I have pulled multiple datasets and json files from different websites that had historical Fear and Greed data. I've included them in the datasets and json folder.

Openstockalert.com

• Data is from 2015-2022
• https://openstockalert.com/FearGreedChart.php
• Note: https://openstockalert.com/GetFnGMktIndex.php - Pulled json from here in response
• File: datasets/openstockalert_fear_greed_data.csv (also datasets/openstockalert_cnn_fear_greed_data.csv)

Hackingthemarkets

• Data is from 2011-2020
• https://raw.githubusercontent.com/hackingthemarkets/sentiment-fear-and-greed/master/datasets/fear-greed.csv
• File: datasets/hackingthemarkets_fear_greed_data.csv

alexey-formalmethods

• Data is from 2011-2022 and is expanded from hackingthemarkets
• https://raw.githubusercontent.com/alexey-formalmethods/datasets/main/money.cnn%20fear-and-greed/data.csv
• File: datasets/alexey-formalmethods_fear_greed_data.csv

macromicro.me

• Data is from 2021-current
• https://en.macromicro.me/collections/34/us-stock-relative/50108/cnn-fear-and-greed
• Note: In order to see the data you need an account (free) and in network tab of inspector you have to view: https://en.macromicro.me/charts/data/50108 The json data saved as macromicro.json from the response
• File: datasets/macromicro_fear_greed_data.csv (regenerated via scripts/convert_macromicro.py from json/macromicro.json)

cnn.com

• Data from 2021-02-01 is accurate, however before that date data is not accurate.
• See https://izy.codes/backtest-strategy-fear-and-greed-index-en/
• See https://github.com/gman4774/fear_and_greed_index github issues
• Data scraped from https://production.dataviz.cnn.io/index/fearandgreed/graphdata/2021-03-01 (now refreshed automatically; see Automation above)
• Files: datasets/cnn_fear_greed.csv, json/cnn_output.json

archive.org scrape

• Scraped available data from 2012-2022 from waybackmachine
• Not everyday is covered
• File: datasets/archive_org_fear_greed_data.csv

Correlation with archive.org data

Higher the coefficent the more correlated with archive.org data:

| DATASET | CORRELATION | |---------------------------------------|--------| | openstockalert_fear_greed_data: | 0.4078 | | alexey-formalmethods_fear_greed_data: | 0.4029 | | macromicro_fear_greed_data: | 0.3437 |

Combining all the Data

The original combined file used hackingthemarkets (2011-2016), openstockalert (2016-2022), and CNN scrape (2022-2023). That file has been superseded by fear-greed.csv (see Canonical dataset above), which uses the same 2011 - 2021-01-29 archive but pulls 2021-02-01 onward directly from the live CNN endpoint.

• File: fear-greed.csv

About the data

CNN's score bands (derived empirically from its own labels in datasets/cnn_fear_greed.csv):

| EMOTION | RANGE | |---------------|--------------| | Extreme Fear | 0 - < 25 | | Fear | 25 - < 45 | | Neutral | 45 - < 55 | | Greed | 55 - < 75 | | Extreme Greed | >= 75 |

Number of total days in each range from 2011-2023

| Range | Days | |---------------|-----------| | Extreme Fear | 503 days | | Fear | 1072 days | | Neutral | 1283 days | | Greed | 240 days | | Extreme Greed | 154 days |

Average number of days in each range from 2011-2023

| Range | Average Duration (Days) | |----------------|-------------------------| | Extreme Fear | 41.6 | | Fear | 56.8 | | Neutral | 16.0 | | Greed | 11.8 | | Extreme Greed | 7.7 |