LLM trading - data sources comparison

[kaidatavis]

All the notes and discussions about the data sources.

[kaidatavis]

Please fill the table with 1) information and 2) a link to the source if available.

Please add new columns, i.e., data sources, and new row, i.e., new type of data wherever necessary.

Data Source	Alpaca	Refinitive	Yahoo	Source
Stock - price - interval	shortest bar is 1 minute.	'minute', 'hour', 'daily', 'weekly', 'monthly', 'quarterly', 'yearly'	browser: daily, weekly, monthly. api: no official api. free api yfinance: shortesr bar is 1 minute. 1m only for the last 7 days, 1d only for the last 60 days.
Stock - price - history	Historical prices usually don't go back earlier than 2016.	1980 >= 'daily', last 1 year <= 'hour'	Historical prices usually don't go back earlier than 1970. For example coke: from 1973.
Stock - volume - interval	same as the stock price	'minute', 'hour', 'daily', 'weekly', 'monthly', 'quarterly', 'yearly'	same as the stock price
Stock - volume - history	same as the stock price	1980 >= 'daily', last 1 year <= 'hour'	same as the stock price
Stock - any other attributes	Free plan: data feed only includes quotes and trades occurring on the order books of IEX.	count, fundamental data: Revenue and Gross Profit	yfinance: By scraping yahoo finance web page. Also provides: 1, finance summaries like earnings, balance sheet. 2, stock actions (including splits and dividends).
Stock - news - history	News api provides historical news data dating back to 2015 provided directly by Benzinga.	App: last 22 years, API: last 15 months	yfinance api: no news data. Rapidapi: provide news data, but not free, I could't find api documents.
Crypto - price - interval	shortest bar is 1 minute.	'minute', 'hour', 'daily', 'weekly', 'monthly', 'quarterly', 'yearly'	browser: daily, weekly, monthly. api: no official api. free api yfinance: shortesr bar is 1 minute. 1m only for the last 7 days, 1d only for the last 60 days.
Crypto - price - history	Historical prices usually don't go back earlier than 2016.	BTC only. 2014 >= 'daily', last 1 year <= 'hour'	BTC from 2014
Crypto - volume - interval	same as the crypto price.	N/A	same as the crypto price.
Crypto - volume - history	same as the crypto price.	N/A	same as the crypto price.
Crypto - any other attributes	Free plan: data feed only includes quotes and trades occurring on the order books of IEX.	Count
Crypto - news - history	News api provides historical news data dating back to 2015 provided directly by Benzinga.	App: Sept 2016, API: last 15 months	yfinance api: no news data. Rapidapi: provide news data, but not free, I could't find api documents.
Other asset: currency - Interval	Only provide data for U.S. listed equities. No additional asset classes.	'minute', 'hour', 'daily', 'weekly', 'monthly', 'quarterly', 'yearly'	same as above
Other asset: currency - History	N/A	1971 >= 'daily', last 1 year <= 'hour'	same as above
Other asset: currency any other attributes	N/A	count	same as above
Other asset: bond	N/A		Yes
Other asset: commodity (gold, wheat, etc.)	N/A	same data characteristic data as stock	Yes
Other asset available	N/A	Oil, UK Natural Gas, Coal, Copper, Alum, Zin, Nickel, Steel, Silver, Gold, Platinum, Palladium, Corn, Rice, Wheat, soybean, canola, live Cattle, Cotton, Cocoa, Coffee, Sugar, etc.	many available
Limitation	Limited to 30 symbols. Limited to 1000 bars per call. 200 API calls/min. 15 min delayed.		yfinance api: 2,000 requests per hour. 15 min delayed.

Notes on Refinitiv data
Based on COKE data (one of the NASDAQ oldest stock):

• Max Historical price in 'daily', 'weekly', 'monthly', 'quarterly', 'yearly': 17 March 1980
  API has limitation 3,000 records per execution (roughly almost 12 years on daily data)
• Max Historical price in 'minute', 'hour': last 1 year
  API has limitation 50,000 records per execution (roughly almost 6 months historical (in minutes))

Based on BTC Crypto (the only available crypto on Refinitiv):

• Max Historical price in 'daily', 'weekly', 'monthly', 'quarterly', 'yearly': 17 July 2014
• Max Historical price in 'minute', 'hour': last 1 year
• API has limitation 50,000 records per execution (roughly 1 months historical (in minutes))