Intraday Futures Calendar Spreads and the Impact of Transaction Costs

Excerpt

The mechanics of trading calendar spreads

To backtest an intraday calendar spread strategy for crude oil futures (symbol CL), I first collect 1-minute bid/ask bars for all CL futures contracts from Interactive Brokers.

After loading the prices into a pandas DataFrame, I use the function get_contract_nums_reindexed_like to obtain a DataFrame showing each contract’s numerical sequence in the contract chain as of any given date:

>>> from quantrocket.master import get_contract_nums_reindexed_like
>>> contract_nums = get_contract_nums_reindexed_like(bids, limit=3)
>>> contract_nums.head()
ConId            CLM9       CLZ9       CLK9       CLJ9       CLN9
Date
2019-03-04        3.0        NaN        2.0        1.0        NaN
2019-03-05        3.0        NaN        2.0        1.0        NaN
2019-03-06        3.0        NaN        2.0        1.0        NaN
2019-03-07        2.0        NaN        1.0        NaN        3.0
2019-03-08        2.0        NaN        1.0        NaN        3.0
2019-03-11        2.0        NaN        1.0        NaN        3.0

I isolate the bids and asks for contract months 1 and 2 by masking the prices with the respective contract nums and taking the mean of each row. In taking the mean, I rely on the fact that the mask leaves only one non-null observation per row, thus the mean simply gives us that observation.

are_month_1_contracts = contacts_nums == 1
month_1_bids = bids.where(are_month_1_contracts).mean(axis=1)
month_1_asks = asks.where(are_month_1_contracts).mean(axis=1)

are_month_2_contracts = contacts_nums == 2
month_2_bids = bids.where(are_month_2_contracts).mean(axis=1)
month_2_asks = asks.where(are_month_2_contracts).mean(axis=1)

I then use the bids and asks to compute the calendar spread. To reflect the fact that I must buy at the ask and sell at the bid, I compute the spread differently for the purpose of identifying long vs short opportunities:

# Buying the spread means buying the month 1 contract at the ask and
# selling the month 2 contract at the bid
spreads_for_buys = month_1_asks - month_2_bids

...

# Selling the spread means selling the month 1 contract at the bid
# and buying the month 2 contract at the ask
spreads_for_sells = month_1_bids - month_2_asks

I use the spreads to construct Bollinger Bands set two standard deviations away from the spread’s 60-minute moving average, and I buy (sell) the spread when it moves below (above) its lower (upper) band.

Any trading symbols displayed are for illustrative purposes only and are not intended to portray recommendations.

Visit QuantRocket website to read the full article and to download the code:
https://www.quantrocket.com/blog/intraday-futures-calendar-spreads