Which inputs does Strategy Hopping use?

Direct answer

Strategy Hopping uses two groups of inputs: (1) inputs that define what “strategy switching” means (selection logic, thresholds, and rules), and (2) inputs that describe the execution and evaluation environment (timing, transaction costs, and how performance is measured). Because markets and implementations vary, you should treat any example as dependent on explicit assumptions rather than as a universal recipe.

Mechanism or definition

Strategy Hopping is a process where you use more than one trading approach and change (“hop”) between them according to some condition. In practice, the relevant inputs are usually:

1. Strategy library inputs

• A set of candidate strategies (for example, different rule sets or parameterizations).
• A definition of what makes two strategies “distinct” (different entry rules, different filters, different holding logic).

2. Switching inputs (the “when to hop” logic)

• Signals or measurements that trigger a switch (commonly model outputs, regime labels, or derived statistics). Even if the source is simple, you need to define how it is computed.
• The decision rule: whether switching is based on thresholds, ranking, voting, or a deterministic mapping.
• Hysteresis or cooldown assumptions: inputs that prevent rapid back-and-forth switching when the measured value hovers near a boundary.

3. State and bookkeeping inputs

• Current active strategy state, last switch time, and whether you allow immediate re-entry after a switch.
• Position constraints: whether you can maintain exposure while switching or must flatten first.

4. Execution and cost inputs (often overlooked)

• Assumptions about execution timing (end-of-bar versus intrabar) and order behavior.
• Transaction cost model inputs: spread/commission assumptions, slippage assumptions, and any fee schedule.

5. Evaluation inputs

• The performance metric: what you measure (for example, returns, drawdown, or risk-adjusted measures) and over what sample window.
• Out-of-sample or walk-forward assumptions: whether strategy selection is tested using data not used to choose the strategies.

Evidence or example (with explicit assumptions)

Consider a simplified model where you hop between two strategies, A and B, using one measured input X.

Assumptions (state them upfront):

• X is calculated from past data only (no future information).
• Switching is allowed only once per time interval.
• Transaction costs are represented as a fixed cost per switch and a fixed per-trade cost.

Example setup:

• Input X: a single derived statistic with a decision threshold T.
• Switching rule: if X ≥ T, use strategy A; otherwise use strategy B.
• Switching limitation: impose a cooldown of N intervals so the system cannot switch again until N intervals have passed.
• Execution assumption: orders fill at the next interval’s open price, with costs applied.

How to interpret results:

• If strategy hopping outperforms a baseline, it is conditional on the assumed execution timing and cost model.
• If it underperforms, that can indicate either the switching rule is poorly matched to the conditions, or that costs and frequent switches reduce net performance.

This example illustrates the key dependency: strategy hopping is not just about the indicator X; it is also about switching frequency, cost assumptions, and the evaluation method.

Limitations and risks

A material failure mode is switching noise: when the switching input X changes due to minor fluctuations, the system may alternate strategies too often. That increases costs (spreads, commissions, slippage) and can cause performance degradation even if each individual strategy performs reasonably on its own.

Other important limitations:

• Non-stationarity: the relationship between X and future outcomes may change over time.
• Data leakage risk: if X is computed using information that would not have been available at the time of switching, results can be overstated.
• Evaluation dependence: historical relationships do not establish future results; performance estimates depend on sample selection, regime mix, and the cost model.
• Implementation mismatch: different platforms, execution models, and operational constraints can alter realized switching behavior.

Verification or next question

To independently verify what inputs Strategy Hopping “uses,” define and document the following before you test anything: (1) the strategy library definitions, (2) the switching inputs and exact calculation method, (3) the state rules that govern switching frequency and constraints, (4) the execution and transaction cost assumptions, and (5) the evaluation protocol (including out-of-sample or walk-forward logic).