Define the hypothesis
Write what the rule is supposed to catch, what it should ignore, and when it should stop acting.
Learn
Forward testing in trading
Forward testing means letting rules run against future market conditions before live capital is involved. In Trading Boy, the safest version is a paper-trading agent with a reviewable decision trail.
Run the workflow
Let the paper agent scan, record decisions, and send workflow alerts without treating those alerts as signals.
Review the evidence
Look for sample size, repeated mistakes, drawdown, missed context, and behavior that does not match the rule.
Useful output
A good forward test produces better rules, clearer limits, or a decision to keep collecting paper evidence. It should not produce overconfidence from a tiny sample.
How forward testing works in a paper-agent workflow
Forward testing is valuable because it studies rules after the rule has been written, not after the outcome is known. Instead of asking how an idea looked in old data, it lets the paper workflow face new market conditions and records what the agent actually did.
Start with a fixed hypothesis. Define the setup, the market universe, the allowed frequency, the risk limit, and the review period. If those inputs keep changing during the test, the sample becomes hard to interpret. Trading Boy is most useful when the agent runs with a stable rule set and every paper decision is recorded in the journal.
During the test, avoid treating alerts as live signals. A forward test is a review environment. The point is to see whether the paper agent follows instructions, whether the rule creates too many trades, whether drawdown stays inside the intended limit, and whether the journal entries are detailed enough for a later human review.
After the test period, summarize the sample before changing the workflow. Count how many decisions matched the rule, how many were skipped, where the largest paper drawdown occurred, and which mistake tags repeated. Then use the feedback loop to choose one adjustment rather than rewriting the entire strategy.
Example forward test
Hypothesis: A paper agent will test whether a trend-following rule can avoid low-volume breakouts. The trader sets a two-week review window and writes the skip conditions before the test begins.
Paper sample: The agent records twelve simulated decisions. Eight match the rule, two are skipped correctly, and two enter before the volume condition is complete. The paper results are mixed, but the repeated early-entry behavior is clear.
Review decision: The trader does not judge the rule only by profit and loss. They tighten the volume confirmation instruction and start a new forward-test sample so the next review can compare behavior before and after the change.
What to freeze during the test
Freeze setup definition, risk limit, market universe, alert behavior, and review window. Stable inputs make the paper evidence easier to interpret.
What to review after the test
Review sample size, skipped trades, rule-fit failures, maximum paper drawdown, frequency, journal quality, and whether the agent followed the written hypothesis.
When to restart a forward-test sample
Restart the sample when the rule changes in a way that affects behavior. A new confirmation filter, risk cap, watchlist, exit rule, or alert threshold means the next paper trades should be compared as a new version rather than mixed with the old evidence.
Do not restart only because the first few trades are uncomfortable. Forward testing needs enough evidence to expose repeated behavior. If the written rules have not changed, keep the sample intact and use tags to describe what happened. If the rules have changed, start a new sample and explain why.
This versioning discipline is useful for AI paper trading agents because prompts can drift quietly. By recording the rule version, the trader can connect future paper results to the exact instructions that produced them.
A forward test can also end with no change. If the paper agent followed the rule, stayed inside risk limits, and produced a reviewable sample, the best next action may be to continue collecting evidence instead of optimizing too early.
Forward testing FAQ
What is forward testing in trading?
It means running rules against future market conditions, often in paper mode, so the workflow can be reviewed before live capital is involved.
How long should a forward test run?
Long enough to create a meaningful paper sample for the rule being studied. The right length depends on trade frequency, market type, and review goal.
Is forward testing the same as live trading?
No. Paper forward testing can use current conditions, but it does not reproduce every live fill, fee, slippage, liquidity, or emotional pressure.