Overview

A research model can identify a desirable position without identifying a trade that could have created it. The last recorded transaction may already be stale, the best displayed offer may cover only a fraction of the desired quantity, and a passive order can remain unfilled while the forecast comes true. This topic establishes the market state and order semantics needed to turn a forecast into an executable instruction.

The three lessons separate three questions. What did the system know? What liquidity did that information describe? What did the instruction authorise? Keeping those questions separate is essential when explaining a difference between a backtest and live trading. A correct signal can coexist with a wrong timestamp join, a correct quote can coexist with insufficient depth, and a correct order can coexist with no execution.

By the end of this note you should be able to:

• reconstruct the best bid and offer available at a decision time;
• distinguish an observed transaction from a current trading opportunity;
• walk a small order book and explain the assumptions behind the calculation;
• distinguish market, limit, stop and time-in-force instructions;
• explain why a cancellation request does not prove that an order is cancelled;
• document an execution simulation without granting it information or fills it could not have had.

1. Contents
2. The big picture