What is a trading strategy?

A trading strategy is a coherent idea for deciding what market opportunity to consider, what conditions matter, and how a position might be entered, changed, or exited. It starts with a premise about how prices, participants, information, or market conditions may behave. That premise is expressed through decisions, not just a label or a chart pattern.

For example, a hypothetical strategy might ask whether a sustained move can continue for a time. That premise alone does not specify an instrument, timing rule, or action, nor establish that the behavior will persist. A useful description makes the idea and assumptions understandable.

The premise behind a strategy

Strategy families rest on different premises: trend following looks for directional persistence, mean reversion for movement toward a reference, and event-driven approaches for a relationship between an event and market behavior. These frame questions; they are not laws markets must follow.

Describe what behavior the idea depends on and what could weaken or contradict it. Volatility, liquidity, market structure, and participants can change, so the premise should not be mistaken for a promise.

Strategy, algorithm, bot, and AI model are different things

These terms are related, but they describe different parts of a trading process. A strategy is the decision concept: what opportunity is being considered and what would influence a position. A trading algorithm is a formal set of computational instructions that can apply rules or calculations to data. An algorithm may implement a strategy, but not every strategy is automated or even fully systematic. For the broader role of formal rules, research, testing, and execution, see What Is Algorithmic Trading? and Algorithmic Trading for Beginners.

A trading bot is software that automates one or more workflow tasks, such as monitoring information, preparing orders, or managing order states. It may apply strategy logic, receive a signal from another source, or simply automate an operational task. The software is not the strategy itself; What Is a Trading Bot? explains that distinction from the automation side.

An AI or machine-learning model is a method that estimates, classifies, ranks, or otherwise transforms inputs into an output. That output might inform a strategy, but it does not automatically define the opportunity, position, or risk decision. What Is AI Trading? introduces the broader AI workflow, while Machine Learning in Trading covers model approaches. An existing overview of machine-learning trading strategies considers how a model can contribute to a strategy, rather than treating the model as a complete strategy.

Discretionary and systematic strategies

A discretionary strategy leaves some judgments to a person, such as interpreting context or deciding whether a situation fits the premise. This can accommodate information that is difficult to express as a fixed rule, but may make decisions less consistent unless the reasoning is recorded.

A systematic strategy specifies decisions in rules or procedures. These can be applied manually or implemented in software; “systematic” does not mean fully automated. Explicit rules make assumptions easier to inspect but do not make an idea sound or ensure that real conditions match those assumed.

Many approaches sit between these descriptions. A person might use systematic criteria to identify candidates and then make a discretionary judgment, or use a model estimate as one input while retaining explicit human review. The important question is which decisions are fixed, which remain judgment-based, and what each part is responsible for.

Dimensions that help describe a strategy

A family name such as “momentum” is not enough to explain a particular strategy. To compare ideas clearly, describe several dimensions and be explicit about what is known versus what remains an assumption.

  • Market or asset. Identify the market or instruments the idea concerns and whether its premise depends on particular liquidity, trading hours, or market structure.
  • Time horizon. Describe whether the concept concerns short-lived movements, longer trends, or events unfolding over another stated period. The horizon affects which observations are relevant.
  • Entry concept. Explain what type of condition would make the idea relevant, without turning the description into a buy or sell instruction.
  • Exit concept. State what could invalidate the premise, end the period of interest, or otherwise change the intended position.
  • Position and risk concept. Describe how exposure is considered in principle and what risks could arise if the idea is wrong. A signal alone does not determine appropriate position size.
  • Failure conditions. Name market environments or assumptions that could undermine the premise, such as a persistent move against a reversion idea or a lack of follow-through after a breakout.

This framework is for understanding and comparing ideas, not a recipe for building a strategy. Detailed systematic research and historical evaluation are covered in Algorithmic Trading for Beginners and the dedicated algorithmic trading materials.

Common trading strategy families

Strategy families group ideas by their central premise or decision concept. Their boundaries are useful for learning, but actual approaches can combine elements from more than one family.

Trend following

Trend-following strategies are organized around the possibility that a directional move may persist. Their central question is whether observed behavior is consistent with an ongoing move, and how the idea could be considered while that premise remains relevant. One challenge is that a move may stall or reverse; signals based on past movement can also respond only after some change has already occurred. See the focused guide to trend-following trading strategies.

Momentum

Momentum strategies focus on relative or recent strength and weakness over a defined comparison or observation period. The concept can apply to individual instruments or comparisons across a group. Momentum is related to trend following, but the terms are not identical: a momentum description often emphasizes measured relative movement, while trend following emphasizes participation in a directional move. Either premise can weaken or reverse.

Mean reversion

Mean-reversion strategies are based on the possibility that a measure or price relationship may move back toward a chosen reference after a deviation. The reference might be defined in different ways, and its meaning depends on the context. A deviation can also reflect a lasting change rather than a temporary imbalance, so assuming reversion without considering that possibility is a central failure risk. See the focused guide to mean-reversion trading strategies.

Breakout strategies

A breakout concept treats movement beyond a defined range or level as a possible change in behavior. A breakout can serve as a trigger within a broader trend-following approach; it is not necessarily a separate or competing premise. A move beyond a reference can fail to continue, and the choice of range, timeframe, or confirmation concept affects what an observer calls a breakout. See the focused guide to breakout trading strategies.

Event-driven approaches

Event-driven approaches focus on situations connected with a defined event, such as a scheduled announcement, a corporate development, or a change in market conditions. The strategy premise concerns how the event and its surrounding information might relate to prices or uncertainty. Events can be anticipated, interpreted differently by participants, or already reflected in available prices; the event label alone does not establish a direction.

Machine-learning-assisted strategies

A machine-learning-assisted strategy uses a model output as one component of a broader decision process. The model might estimate a quantity, assign a category, or rank observations; separate strategy logic determines how that output is interpreted and whether it matters to a position. Model type, feature design, training, and validation belong to AI Trading. The strategy-focused question is how an estimate relates to the premise and decision, not how to train the model.

Strategy families can overlap

Families can overlap. A breakout may trigger a trend-following strategy, momentum may describe the observed condition, and an event may explain why a move is being studied. A machine-learning estimate could inform any of them.

Compare the roles: identify the core premise, the trigger, and whether a model or event is an input or the strategy’s central concept. Labels alone do not describe the whole decision process.

A conceptual framework for comparing strategies

Compare a strategy’s premise, market, horizon, relevant observations, potential invalidation conditions, and reliance on judgment, rules, or model outputs. Ask how it might behave in different conditions and what remains uncertain.

For example, a hypothetical trend-following idea depends on possible persistence, while a mean-reversion idea depends on movement toward a reference. Instead of asking which is “better,” compare their assumptions, failure conditions, horizons, and risk concepts.

Historical research is separate from defining a strategy family. The algorithmic trading cluster covers systematic research; its backtesting guide explains historical evaluation.

No strategy is universally suitable

No premise applies equally across markets and conditions. Liquidity, costs, participants, data limitations, and unexpected events can affect behavior; historical descriptions do not establish future results.

Use strategy concepts to ask questions, not to infer that a family is profitable or suitable. An idea, evidence about it, and a decision to act are distinct.

Where to explore next

Visit the Trading Strategies hub for the section’s coverage. For systematic research, continue to Algorithmic Trading for Beginners; for model methods and strategy decisions, see Machine Learning in Trading and the Machine Learning Trading Strategies overview.