Direct answer: the minimum data set
To assess Forex signals, collect data that lets you understand (1) what the signal is claiming, (2) which inputs it uses, (3) where those inputs came from, and (4) when they were available. If you cannot verify those items independently, you cannot reliably evaluate the signal.
At a practical minimum, you want: the signal definition (what action and timeframe it refers to), the raw or summarized inputs used to generate it, the provenance of those inputs, the timestamps that show timeliness, and any stated assumptions about costs and execution. Then add quality checks and limitations to understand what could cause the signal to fail.
Mechanism and definition: separate stable logic from variable conditions
“Forex signals” are information artifacts that claim to indicate a potential trading decision. They usually involve a mapping from market data to a recommendation-like output, such as an instrument choice and a timing window. The core assessment task is to separate stable mechanics (how the mapping works) from variable conditions (market regime, liquidity, costs, and execution).
The key data you need for the “mechanics” part:
- Claimed output format: what exactly is specified (instrument, direction, entry timing, exit timing, or holding period).
- Claimed input features: the indicators, price levels, volatility measures, sentiment, or other variables.
- Transformation rules: whether inputs are normalized, filtered, combined, or thresholded.
- Time handling: how the signal aligns with candle close/open times or other sampling rules.
Because market conditions and trading costs change, these mechanics alone do not guarantee results. Historical relationships also do not establish future results.
Evidence and example: what to verify with the right data
Consider a signal that claims it was “generated from” recent price data.
To evaluate it without assuming outcomes, verify these evidence items:
- Provenance of input data: identify the feed or provider type (official exchange data vs. broker quotes vs. aggregated data). At minimum, record where the numbers were obtained.
- Timeliness: capture timestamps for the inputs and the signal generation time. Confirm that the inputs would have been known at signal time. Delayed data is a common reason evaluations become misleading.
- Data quality checks: look for missing periods, outliers, inconsistent time zones, and mismatched sampling (for example, using one timeframe while labeling another).
- Execution assumptions: if performance is shown, determine whether it accounts for spreads, commissions, slippage, and order fill rules. Without realistic assumptions, backtests can be non-comparable.
- Calculation assumptions: if metrics are provided (returns, drawdowns, win rate), confirm the formula basis (time window, compounding, currency conversion, or whether fees are included).
If any of these evidence items are absent, you are left with a statement that cannot be independently verified.
Limitations and risks: material failure modes to expect
Even with good data, multiple limitations can prevent a reliable assessment:
- Non-replicable performance: if the signal depends on proprietary processing or undisclosed parameters, you may not reproduce it.
- Regime change: relationships that worked in one period can break when volatility or liquidity changes.
- Execution mismatch: performance can degrade when live fills differ from assumptions.
- Selection bias: if only favorable periods are shown, quality checks and comparisons become unreliable.
A material limitation to keep in mind is that outcomes vary with market conditions, costs, execution quality, and jurisdiction. Therefore, historical results do not guarantee future performance.
Verification and next question: use a checklist
To assess Forex signals independently, run a checklist focused on data integrity:
- Do you have a precise definition of the signal output and timeframe?
- Can you identify input provenance and confirm timestamps for timeliness?
- Are input and claimed calculations verifiable (consistent sampling, no missing data)?
- Are costs and execution assumptions stated clearly enough to evaluate them?
- Do you understand at least one failure mode (for example, delayed data, regime shift, or execution mismatch)?
If you want, tell me what kind of signal you are looking at (for example: price-only rules, indicator-based rules, or a provider’s discretionary calls). Then I can help translate that into the exact data fields you would need to collect and verify.