Free Quant Data with QMT: XtQuant Setup Guide

Introduction to QMT and XtQuant

QMT is one of the most accessible interfaces for live trading in quantitative trading. It ships as a locally deployed quant platform that supports both backtest and live trading, plus an SDK that runs independently of the platform — XtQuant.

XtQuant only exposes market data and trading APIs — there is no backtest engine. Standard market data is currently free through XtQuant, though rate-limited. In testing, one request per second is well within the limit.

Fetching Market Data with XtQuant

In XtQuant, data access is a two-step process: cache first, then read.

Cache-stage APIs generally start with download_.

So to get historical bars, you first populate the local cache with:

123456def download_history_data(stock_code: str='', 
                      period: str='', 
                      start_time: str='', 
                      end_time: str='', 
                      incrementally: Optional[bool]=None
                      )

To download bars for multiple securities in one go, use download_history_data2.

Once the start_time to end_time range is cached, you can read it with get_market_data:

123456789def get_market_data(field_list = [], 
                    stock_list = [], 
                    period = '1d',
                    start_time = '', 
                    end_time = '', 
                    count = -1,
                    dividend_type = 'none', 
                    fill_data = True
)

This method supports '1m', '5m', '15m', '30m', '1h', '1d', and tick data. Available fields vary by period. Except for ticks, you get timestamp, OHLC, volume (in lots) and amount (in currency value).

Note that this method cannot return adjustment factors. If you plan to archive the data into another database rather than just using the returned bars directly, also call `get_divid_factors` and store the unadjusted bars together with the factors. Persisting already-adjusted data is meaningless — even backward-adjusted data can introduce errors.

Example

1234567891011121314151617181920from xtquant import xtdata

stocks = ['000001.SZ', '600000.SH']
xtdata.download_history_data(stocks[0], '1d')
xtdata.download_history_data(stocks[1], '1d')

# 或者
# XTDATA.DOWNLOAD_HISTORY_DATA2(STOCKS, '1D')

end = "20231220"
bars = xtdata.get_market_data(stock_list=stocks, 
                              period='1d', 
                              end_time=end, 
                              count=-1, 
                              dividend_type="front_ratio")

end = "20231220"
bars = xtdata.get_market_data(stock_list=stocks, period='1d', end_time=end, count=-1, dividend_type="front_ratio")

display(bars['close'].T.tail())

get_market_data returns a dict. Each key is a market-data field such as time, open, or close, and each value is a frame indexed by ticker with timestamps as columns. In the example above we transpose it, which is how you will normally work with it.

The final output looks like this:

Note the date format used by xtquant: "YYYYMMDD" for the 1d period, and "YYYYMMDDHHmmss" for intraday periods. If you start from a datetime.datetime, format it like this:

1234567import datetime

now = datetime.datetime.now()
print(now.strftime("%Y%m%d%H%M%S"))

import arrow
arrow.get(now).format("YYYYMMDDHHmmss")

strftime format strings are hard to remember. The arrow library improves on this with a progressive, largest-to-smallest pattern that is much easier to recall. When performance is not critical, convert to an Arrow object first and then format.