A trading signal is the output of a decision rule indicating a trading intention, such as entering, maintaining, reducing or closing an exposure. It connects analysis to action. A buy signal can identify a candidate trade without specifying how many shares to buy, whether an order will fill or whether the trade will make money.
The word signal has several uses in quantitative finance. An alpha signal concerns a prospective return opportunity; signal generation concerns the process that constructs the output from available data. Here the focus is the output's operational meaning: what decision does it express, and when does that decision change? A signal can also come from a risk rule or a hedge requirement, without forecasting a positive standalone return.
Specify a state or an event
“Buy = +1” is incomplete documentation. It could mean a persistent preference to be long, an instruction to add exposure every time the rule runs, or a one-off entry event. Those interpretations can produce very different portfolios from identical observations.
For a deliberately simple long/flat state convention, let be a fictional score known after the close of session . Choose
Here means the rule favours a long exposure and means it favours no exposure. Equality at the threshold belongs to the long state by choice. Neither value is a share count, portfolio weight or predicted percentage return. A flat state is an exit preference for an existing long, rather than an instruction to open a short.
Define an entry event only when the state changes from to , and an exit event when it changes from to . A repeat of leaves the rule's state unchanged. An actual position might still differ because of risk limits, an unfilled order or an execution failure.
Score 3 produces a buy signal, not three shares
Suppose successive observed closes A–D have scores 1, 3, 3 and 1. Initialise the rule in the flat state at A. At B, score 3 meets the buy threshold 2, so the rule produces a long state and a new buy-entry candidate. At C the state remains long: there is one entry event across two long-state observations. At D it returns to flat and produces an exit candidate.
Original synthetic example. Lines join observed scores and do not specify an intraday path. Scores have arbitrary units; sizing, order submission and possible execution are reviewed in the following session.
This convention prevents a persistent signal from being mistaken for repeated permission to buy more. It does not eliminate order management. If B's purchase remains partly unfilled at C, a separate policy must decide what happens to the remainder, accounting for fills and working orders before submitting anything else.
The score is synthetic and has arbitrary score units. Its sign convention is chosen for this illustration: a higher score favours buying. Other rules reverse that meaning. For example, the mean-reversion rule in Avellaneda and Lee's author preprint, section 4.1 buys on a sufficiently negative normalised residual and distinguishes opening from closing trades. A number's sign acquires trading meaning through the rule, not by itself.
Put the signal on the right clock
In this example, the close must be observed before the score and signal are available. Any resulting sizing and order decision is reviewed in the next trading session. This is an example convention, not a universal execution rule. A backtest that calculates the signal from a completed close and grants an earlier fill would use information before it was available: look-ahead bias.
The next-session convention supplies no guaranteed execution price. A market order leaves the fill price uncertain; a limit order constrains price but can remain unfilled. Investor.gov's order guide explains these distinctions for U.S. securities orders.
To evaluate a signal, retain its instrument, observation and decision times, rule version, score units, intended horizon, state and transition meaning. Then specify sizing, risk constraints, transaction costs and execution separately. Realised profit requires the resulting position and cash flows; the signal alone cannot supply it.
For that next step, From Signal to Executed Position Capstone follows an intention through an authorised quantity, partial fills and cancellation to a reconciled position. The full lesson requires membership; the linked page offers a public preview.
Sources
- Rishi K. Narang, Inside the Black Box: A Simple Guide to Quantitative and High-Frequency Trading, 2nd edition (Wiley, 2013), Chapter 2, pp. 17–18 — the distinct functions of forecasts, risk, costs, portfolio construction and execution, allowing for variation in system architecture.
- Marco Avellaneda and Jeong-Hyun Lee, Statistical Arbitrage in the U.S. Equities Market, author preprint dated 10 July 2008, section 4.1, printed pp. 19–20 — a concrete rule with separate opening and closing conditions; checked 26 October 2026. The synthetic threshold example above is independent of that strategy.
- SEC Investor.gov: Types of Orders — market and limit orders, execution prices and price restrictions; checked 26 October 2026.