Ask one narrow question
Name the pair, timeframe, market condition and decision being studied. “Does this rule improve the baseline?” is testable; “What is the best strategy?” is not.
FOREX EVIDENCE LAB
Turn an idea into a fair test.
A strategy is more than a chart that worked. The WickAtlas research workflow makes the rules, sample, costs, uncertainty and losing outcomes visible before any conclusion is published.
Freeze the rules
Write the signal, entry timing, stop, exit, position size and no-trade filters before revealing outcomes. Every discretionary choice needs a recorded definition.
Build the eligible sample
Count every event that met the frozen rules, including losses, ambiguous cases and long periods with no signal. Attractive screenshots are illustrations, not a sample.
Model trading friction
State the spread, commission, financing, slippage and bar-execution assumption. If reliable cost data is unavailable, test a range and label the uncertainty.
Protect unseen data
Use one period to design the idea and a separate period to challenge it. Parameters are not adjusted after seeing the validation result without starting a new version.
Report the distribution
Show trades, expectancy, drawdown, losing streaks, sensitivity and failure regimes—not only return or win rate. A result can be interesting without being trade-ready.
THE RESEARCH RECORD
If a rule is vague, the test is not ready.
These fields form the minimum research record. Together they make the study repeatable and expose places where hindsight or hidden discretion could change the outcome.
01QUESTION
One decision the study is trying to improve
02UNIVERSE
Pairs, session, timeframe and date range
03REGIME
A measurable condition for when the rule is eligible
04SIGNAL
Inputs, parameters and information available at that moment
05EXECUTION
Entry timing, order assumption and duplicate-signal handling
06RISK
Stop, size, portfolio exposure and maximum concurrent positions
07EXIT
Target, trailing rule, time exit and invalidation
08FRICTION
Spread, commission, slippage and overnight financing
09VALIDATION
Design sample, untouched sample and any demo forward test
10DECISION
Reject, revise, observe further or consider for demo use
EVIDENCE LABELS
Say what is known—and what is not.
Every WickAtlas study should carry one of these labels. Moving up a level requires new evidence, not stronger language.
Worth testing
The logic and rules are clear enough to investigate. No claim about performance has been established.
Observed in a stated sample
A result is reported with its data, costs and limitations. It may still be noise, overfit or sensitive to execution.
Survived independent checks
The frozen method held up on unseen data and was reproduced. This still does not predict future returns or make it suitable for every trader.
Replace the impressive sentence with observable rules
“A moving-average crossover catches trends” leaves the pair, timeframe, crossover lengths, entry bar, exit and market condition undefined. Almost any attractive chart can be made to fit it after the event.
A research protocol names those choices in advance, compares the rule with a simple baseline and records every eligible event. Parameter alternatives belong in a sensitivity test, not in a search for the one setting that made the past look best. This example describes process only; it is not evidence for a crossover strategy.
Stop weak ideas before they reach a trading account
- The rule cannot be applied consistently by another reader.
- The apparent advantage disappears under plausible costs.
- A small parameter change reverses the conclusion.
- Most results depend on one pair, month or exceptional trade.
- The untouched sample does not support the original finding.
- The drawdown or losing streak is incompatible with the risk plan.
A backtest is a model of execution, not execution itself
Historical prices may contain gaps, different broker feeds, revised data or incomplete spread information. Bar data cannot always show the order in which an entry, stop and target were touched. A simulated fill may not reflect latency, rejection or widening spreads during live conditions.
For that reason, research findings are educational evidence, not personalised advice or a promise of profitability. Demo forward observation can reveal implementation problems, but it cannot remove future market uncertainty.