Monthly Universe Selections and Rebalancing in Framework Algorithms

[DerekMelchin]

Expected Behavior

We document a reliable way to perform universe selection and rebalancing on a monthly basis in framework algorithms.

The docs explain that the universe selection model's constructor should have

self.hours = None

and the selector function should have

if not self.hours or algorithm.LiveMode:
    self.hours = algorithm.MarketHoursDatabase.GetEntry(Market.USA, "SPY", SecurityType.Equity).ExchangeHours
self.next_open = self.hours.GetNextMarketOpen(self.algorithm.Time, False)

if self.month == self.next_open.month:
    return Universe.Unchanged

as demonstrated in this algorithm.

This content is explained in https://www.quantconnect.com/docs/v2/writing-algorithms/securities/exchange and the universe selection docs.

Actual Behavior

Undocumented. The standard way to perform universe selection and rebalancing is to add

if self.month == self.algorithm.Time.month:
    return Universe.Unchanged
self.month = self.algorithm.Time.month

to the universe selection function and

if data.Time.month == self.month:
    return []
self.month = data.Time.month

to the Alpha model Update method.

However, if the first trading day of the month is a Monday, then it leads to issues.

For example, in this backtest, the logs show

Universe selection at 2023-04-29 04:00:00 Universe selection at 2023-05-02 04:00:00

2023-05-01 is a Monday, so the Alpha model emits a bunch of insights at 2023-05-01 9:31 AM. Universe selection runs at 2023-05-02 4 AM. It removes some of the securities that we emit insights for, canceling their insights. When the warm-up period ends at 2023-05-18, only some of the insights are still active, so the portfolio rebalances for the first time and has 70% exposure.

Checklist

• [x] I have completely filled out this template
• [x] I have confirmed that this issue exists on the current master branch
• [x] I have confirmed that this is not a duplicate issue by searching issues