Is there a way to backtest as t+1 ？

[XMethues]

I'm trying to backtest the Chinese market, which is t+1.

[initial-d]

For the Chinese A-share / ETF T+1 rule, I would separate two levels of accuracy.

For a simple daily-bar strategy that is either flat or long, the usual approximation is:

1. generate signals using information available at day t;
2. execute those signals on day t+1 by shifting the executable entries/exits by one bar;
3. make sure an exit cannot close a position that was opened on the same bar.

In other words, do not let a signal computed from today's close trade at today's close. Shift it to the next tradable bar. That already avoids most look-ahead and same-day round-trip problems.

For a stricter T+1 simulation, especially if you allow pyramiding, partial exits, or multiple buys before selling, simple shifting is not enough. You need to track sellable inventory: shares bought today are not sellable until the next trading day, while shares carried from previous days are sellable. In that case I would use a custom order function / simulation path rather than only Portfolio.from_signals, because the sell order size has to be capped by the quantity that is T+1 eligible.

A practical rule of thumb:

• single-position, daily strategy: shift signals by one bar and disallow same-bar entry/exit;
• multi-lot / partial-fill / intraday-like strategy: implement custom order sizing that tracks position_bought_today vs position_available_to_sell.

Also note that T+1 is only one part of the China-market realism layer. If you are trying to model A-shares more faithfully, you may also want to handle price limits, suspended days, non-tradable bars, lot size, commissions/stamp duty, and realistic open/close execution assumptions.