First Officer · Strategy Testing and Uncertainty · About 30–40 minutes
Safety rule: A forward test must use the rule version and information available when each decision occurs.
Your mission today
By the end you can run a forward demo test with frozen rules, realistic fills and a full opportunity ledger.
Before you start
Complete C07.1 and bring its strategy card, cost assumptions and backtest ledger. Use paper or demo only, never real funds for this exercise.
1. Forward testing means the outcome has not happened yet
In a manual backtest, you hide later historical candles. In a forward paper test, you apply rules as new information actually arrives. Think of testing a new bus timetable on future mornings rather than selecting only old days when buses were punctual. The next outcome is unknown at decision time, which helps expose unrealistic rules and execution assumptions. A paper trade is recorded without risking real money. A demo order runs in a simulated broker environment. Both can differ from live execution. Write a version number on the strategy card and freeze it for the planned observation period. Choose the exact instrument, timeframes, sessions and eligibility conditions. Decide in advance what constitutes a rejected setup, late fill, outage or missing quote. Your goal is to discover whether the process can be followed consistently, not to manufacture a profitable month.
2. Capture the downside before each imaginary order
Suppose EURUSD is quoted Bid 1.08500 and Ask 1.08520. An imaginary long enters at Ask 1.08520 and uses a planned Bid stop at 1.08420. Price distance is 10 pips. With an invented position value of $1 per pip, the planned price loss is −$10 before any other charges or worse fill. If the account's hypothetical limit is $8, the current position size is not allowed. The correct response is to reduce size within real broker steps or skip. In another possible path, a later Bid at 1.08670 would produce +15 pips before separate charges. Do not enter that favourable outcome into the record until the actual demo observation occurs. If the order fills at a different Ask, recompute the risk from the fill, not from a prettier earlier quote.
3. Build a timestamped offered-opportunity ledger
A forward journal should show each time the strategy could have been considered. Save the date, timezone, market, most recent completed signal candle and exact quote used. Add market condition, spread, event window, intended structural stop, size and permitted action. Eligibility means all written conditions have been met; a good-looking chart alone is not enough. Log an opportunity that becomes WAIT because the M15 candle is unfinished. Log SKIP when the spread exceeds its maximum. Log TRADE only when the full rule allows the order and the risk fits. Keep rejected and missed candidates. A strategy that offers only two permitted trades over four weeks must not be marketed as producing reliable daily income. Frequency is part of the observed evidence. Capture screenshots at the decision time, not when a later candle makes the story easier to explain.
4. Record the gap between an intended and simulated fill
Slippage means the difference between a requested or expected price and the actual executed price. It can arise during rapid movement or thin liquidity. Suppose the intended buy Ask is 1.08520 but the demo fills at 1.08535. The entry is 1.5 pips worse. Against the same Bid stop at 1.08420, the planned price loss becomes 11.5 pips, or −$11.50 at the fictional $1 per pip before fees. That may breach an $11 total-loss cap. Do not backfill the original 1.08520 just to keep the trade within budget. Some demo platforms simulate fills unrealistically smoothly, so note that limitation too. If the broker rejects a stop or a pending order never fills, record the actual platform status. A chart touching the requested price does not establish an executable fill without the correct quote side and order rules.
5. Do not change the experiment midstream
Perhaps ten early observations all lose money. You may be tempted to add a new indicator or move your stop. That creates a new strategy version. It does not repair the first sample. Finish the current evaluation under its frozen rules unless a safety issue requires stopping. If the method is dangerous or its product terms are misunderstood, stop immediately and record why; student safety is more important than completing an experiment. Later you can design a new version and begin a fresh study. Even a strong demo result is not proof of future live performance. Demo latency, historical gaps, spreads and emotional conditions can differ. Your conclusion should name the sample dates, number of offered setups, accepted and skipped counts, net result under stated costs, worst observed decline and any compliance errors.
6. Your four-week field assignment
Use four consecutive teaching weeks and record every opportunity, including no-trade days. A week with zero valid setups is a legitimate observation. After each day, compare expected and actual quote-side fills, risk, net cash and net R. Once each week, audit a random selection of screenshots against the earlier timestamps and frozen strategy card. Consider this fictional summary: 20 offered setups, 12 accepted, 8 skipped and four rule breaches. That is not a clean sample of 12 perfect decisions; the eight skips and four violations are central to the lesson. Write how you would improve execution discipline without changing the test's trade rules midstream. The final report must say which conclusions are supported and which require more data. Continue to Captain only when you can explain both the numbers and their limitations.
Common mistake
A student changes the stop after a losing day but keeps all results under the original strategy name. The rules are no longer the same test. Fix the log by freezing version 1 and starting version 2 as a new experiment.
Plain-English glossary
- Forward test: applying frozen rules as new observations arrive.
- Paper trade: simulated decision with no live funds at risk.
- Slippage: difference between expected and actual execution.
- Eligibility: satisfaction of all predeclared entry and risk gates.
- Strategy version: dated, unchanged rules for one study.
Try it yourself
Run a four-week forward demo or paper log. Capture all offered setups, quote sides, planned loss, fills, fees, skips, violations, net R and weekly summaries. Hand a second person three random decisions and ask them to reproduce the original eligibility judgement.
What you can now do
- Explain the experiment and its key limitations.
- Recalculate loss or drawdown with the correct unit.
- Preserve a complete, dated and independently reviewable record.
Next lesson
C07.3 · Monte Carlo: How Bad Could a Bad Streak Get? Explore different outcome orders without inventing a profitable edge.
