A Signal Isn’t a Strategy
A signal can point to a pattern. A strategy has to explain what to do with it.
The signal is only the starting point
A signal can identify a pattern: price strength, a valuation spread, a volatility measure, a breakout, a reversal, or any other condition that appears to contain information. But a signal is not yet a strategy. It does not say exactly what to trade, when to trade it, how much capital to commit, what to do when data is missing, how to account for the portfolio, or what comparison would make the result meaningful.
That gap matters. A signal can sound persuasive while leaving the hardest choices unresolved. A strategy has to make those choices explicit enough that the test becomes deterministic and the reader can see what assumptions the result depends on.
A signal tells you where to look; a strategy tells you what has to survive.
This is why signal research and strategy testing should not be treated as the same thing. The signal is the starting claim. The strategy is the full testable object.
A rule set makes the signal testable
The next layer is the declared rule set. This is where a signal becomes specific enough to test. The rule set defines the universe, eligibility timing, signal formula or measurement window, data frequency, entry rule, exit rule, ranking direction, thresholds, rebalancing schedule, holding period, and any exceptions.
It also has to handle the less glamorous decisions that often determine the result: tie-breakers, priority when rules conflict, tradability filters, corporate-action and survivorship treatment, and warm-up requirements. These details can feel mechanical, but they are part of the strategy claim. Without them, the same signal can become many different tests.
A vague signal can be interpreted many ways. A strategy has to choose one path and live with the consequences.
Timing decides what could have been known
A strategy also needs timing rules. It has to say when the signal is observed, when the trade can occur, what price family is used, how next-tradable-bar handling works, and how missing or invalid bars are handled.
This is not a minor implementation detail. A signal that is observed after the close cannot be allowed to trade at a price that came before the signal was knowable. A test that is loose about timing can give the strategy information or execution quality it would not have had.
The goal is simple: the strategy should only act on information that would have been available at the time, under rules that can be applied consistently.
Portfolio construction turns a signal into capital allocation
A signal is not a portfolio. A strategy needs to say how much capital each position receives, how positions are weighted, how residual cash is handled, and how target weights differ from actual weights after trading constraints are applied.
It also needs portfolio-accounting assumptions: whole-share or fractional-share handling, dividend and distribution treatment, cash yield or financing treatment where relevant, rebalancing order, gross and net exposure, leverage, margin, short availability, borrow assumptions, and corporate-action or delisting handling.
The signal can be elegant while the portfolio it creates is crowded, constrained, expensive to maintain, or economically weak. The portfolio is where the idea becomes measurable.
Costs, liquidity, and capacity assumptions test the edge
Trading friction is part of the strategy result, not a footnote after the result. Commissions, spreads, slippage assumptions, entry and exit constraints, missing execution bars, volume limits, low-priced securities, turnover, and borrow costs can change the strategy’s economic meaning.
A signal that looks interesting before friction may look different after the test applies trading assumptions. The same is true when a modeled trade violates the test’s volume or liquidity constraint. In that case, the issue is not just that returns might be lower. The strategy may be asking for a trade path that the test assumptions do not support.
These assumptions should be read carefully. Standardized cost, liquidity, volume, spread, slippage, and borrow assumptions are test assumptions or constraints. They are not a live capacity estimate and they are not an assurance that a real order could be filled at the modeled size or price.
The benchmark defines the question
A strategy also needs a benchmark matched to the question being asked. The benchmark is not decoration. It tells the reader what comparison gives the result context.
Some strategies are mainly asking whether an active rule improves on a familiar market exposure. Others are asking whether a signal adds value inside a specific opportunity set, or whether a more complex implementation does enough to justify its burden. Different questions may require different comparison sets.
Without the benchmark matched to the question, the reader may not know whether the strategy’s result reflects the active decision, the background exposure, the cash policy, or the opportunity set itself.
Failure modes are part of the strategy
A complete strategy has to be evaluated by more than its endpoint value. The path matters. Drawdowns, time underwater, turnover, liquidity strain, concentration, benchmark sensitivity, short-side burden, implementation ambiguity, parameter sensitivity, data availability risk, implementation complexity, regime dependence, and path dependence can all change how the result should be understood.
These are not just risk disclosures at the end of the report. They help reveal what kind of strategy the signal became. A signal may survive a historical test while still leaving behind a process that is too fragile, too costly, too ambiguous, or too dependent on hard-to-repeat assumptions to deserve much confidence.
What BTS does with the gap
At Backtested Strategies, the goal is not to make every signal look like a complete strategy. The goal is to show what happens when a signal is translated into a fully specified implementation.
That starts with the declared rule set. When implementation choices are missing, the test should document them and apply disclosed methodology defaults needed to make the test deterministic where the declared rules are silent. Those defaults are implementation assumptions, not hidden changes to the strategy idea.
If a required input is not available to the test, BTS either defines a disclosed proxy, flags the limitation, or excludes the strategy rather than hiding the assumption.
The result is still uncertain. It does not guarantee live-trading outcomes. It does not prove that a strategy can be traded at unlimited size. It does not remove the need for judgment. But it does make the assumptions more visible and the claim easier to evaluate.
How to read a strategy claim
Before trusting a strategy claim, readers should be able to answer a few basic questions:
- What is the signal?
- What are the rules?
- When is the signal known, and when can the trade occur?
- How is capital allocated?
- How are cash, dividends, financing, shorts, and trading constraints handled?
- What costs, liquidity limits, and capacity-related assumptions are included?
- What benchmark is matched to the question?
- What failure modes would change the interpretation?
A strategy claim is not complete until the reader can identify the rule set, the capital path, the benchmark, the frictions, the liquidity and capacity-related assumptions, and the failure modes.
A signal is the spark, not the claim
Signals matter. They are often where the research begins. But the signal is only the spark. The strategy is the full claim: the rules, timing, sizing, accounting, costs, benchmark, and failure-mode review that make the idea testable.
That distinction protects the reader. It keeps an attractive pattern from being mistaken for a complete process. It also gives useful signals a fairer test, because the strategy has to show what remains after the missing layers are made explicit.
A signal tells you where to look. A strategy tells you what has to survive.