A trading system should have first-class support for placing orders via broker API. With Zorro, orders are second-class to trades in Zorro. This is a major flaw. Most brokers supported by Zorro are nicknamed "NFA brokers", but this really means they take orders and not trades.
I have a three-part proposal. First is the broker plugin API itself:
1) Revamp the Broker API for emphasis on order control. At minimum, a dedicated function for new orders, another to cancel/replace orders, and another for execution reports. Keep it completely separate from "Trade" management.
2) On Live instance launch, Zorro will check the status of all open orders on the API and add them to the list of open orders.
3) Zorro will track order status from birth to death. That is, Zorro will never assume that an order is dead unless it confirms it is in fact dead via execution report.
Even if Zorro issues an order cancellation, Zorro should not assume it is cancelled until the order is confirmed dead.
4) The broker plugin can push execution reports to Zorro.
Here's an example execution report from the FIX protocol:
https://www.onixs.biz/fix-dictionary/4.4/msgType_8_8.htmlSecond, orders should be clearly distinct from ordinary zorro Trades, which have fundamentally different meanings. If necessary, make an ORDER struct. Or at minimum, give TRADE structs an Order flag.
Third would be a new hedge mode (7?) that allows order placement to yield Zorro trades for accounting purposes only. You can call it "reverse pool trades". Example:
Script directly places order: BUY 10 MSFT. Zorro internally opens a long trade for 10. The execution reports will update the fill information.
Script directly places order: SELL 20 MSFT. Zorro internally closes the long trade and opens a new short trade for 10. The execution reports will update the fill information for both the long and short trades on a FIFO basis.
