Stop Trading the Wrong Session: A 71% Worked Scenario

A 123-trade session ledger can show +$898 across all sessions and +$1,536 from London Open alone. Retaining that one window changes the same historical net result by +71.0%. The arithmetic is real inside the worked ledger: ($1,536 − $898) ÷ $898 = 71.0%. The lesson is not that every trader should choose London. It is that an aggregate result can hide one personally useful window and one net-negative block.

This rebuilt guide preserves the original 71% hook, closes every trade, win, and dollar, and turns it into a personal session-fit diagnostic. The ledger is constructed—not a customer record, a TSB cohort, or a promised return—and the first historical comparison generates a test rather than proving that the clock caused the result.

Disclosure: Trader's Second Brain is our product. TSB can preserve session evidence and run a declared retrospective filter; the session definition, causal interpretation, and decision remain trader-owned. Current product facts render from the server catalog.

The Question Is Personal Fit, Not the “Best” Global Session

London, New York, Asia, and overlap labels describe time windows. They do not automatically describe a strategy, instrument, setup, spread, news regime, or trader state. A profitable London bucket may contain a setup that never appears later in the day; a weak afternoon bucket may contain a different instrument or a cluster of oversized trades.

That is why the decision is narrower: for one frozen account and strategy, does a reproducible time window contain a material difference worth testing later? The session-performance guide covers the general measurement framework. This page shows how one strong 71% historical result survives scrutiny without becoming a universal claim.

Freeze the Session Contract Before Reading the Result

• Account and strategy: one account scope and one stable strategy version.
• Period and cutoff: exact start, end, and export or query timestamp.
• Timezone: the account timezone plus daylight-saving treatment.
• Assignment event: entry time, exit time, or another declared event—never whichever produces the better bucket.
• Unit: one logical closed trade; fills must be grouped consistently.
• Money basis: the same display currency and a clear statement of represented costs.
• Missing policy: unlabeled rows remain unavailable and visible; they are not silently assigned to “other.”

The journal-fields checklist covers the timestamps, account, setup, cost, and evidence fields needed to reproduce the buckets. If those fields cannot be recovered, report the limitation instead of manufacturing precision.

The Reconciled 123-Trade Ledger

Constructed example—not a customer or aggregate result. Scope: one forex strategy, one account, four complete calendar months, 123 logical closed trades, account timezone UTC, entry-time session assignment, one display currency, and the represented trading costs already included in net P&L.

The counts close: 48 + 35 + 28 + 12 = 123. Wins close: 28 + 18 + 12 + 4 = 62, or 50.4%. Money closes: $1,536 + $490 − $672 − $456 = $898. Nothing from the starting population disappears.

The non-London block contains 75 trades, 34 wins, and net -$638. Retaining London Open alone therefore leaves 48 trades and +$1,536. The historical change is +$638, and $638 ÷ $898 = 71.0%.

What the 71% Result Does—and Does Not—Mean

The 71% figure is deliberately quotable because its denominator is visible. It describes a London-only replay against the complete historical total. It does not mean every non-London bucket lost: the overlap row made +$490. It means the three non-London rows combined to -$638.

Other retrospective choices produce other flattering numbers. Removing only the worst single row, NY afternoon, changes +$898 to +$1,570, or +74.8%. Removing both negative rows changes the result to +$2,026, or +125.6%. Those are valid arithmetic descriptions of the frozen ledger, but choosing among them after seeing the outcomes is model selection. The largest number is not automatically the best operating rule.

Why the Largest Retrospective Number Is Not the Headline

The same ledger can support several questions, and they should not be blended. “What if only London had been eligible?” produces +$1,536 and the preserved 71.0% change. “What if only the worst row had been removed?” produces +$1,570 and +74.8%. “What if every negative row had been removed?” produces +$2,026 and +125.6%.

The last option is numerically largest because it uses the completed outcomes to choose two exclusions. It also keeps the positive overlap bucket that the London-only schedule removes. That makes it a useful sensitivity calculation, not automatically a cleaner rule. A trader could keep searching instrument × weekday × hour combinations until one historical curve looks extraordinary; without a record of every search, the final percentage hides the selection process.

The 71% result remains the article's anchor because it answers one reproducible schedule question and retains the values already associated with this page. The alternative calculations stay visible so a reader can see that the editorial choice was not made by suppressing a larger result.

Keep Four Evidence States Separate

1. Eligible and executed: trades that match the frozen account, strategy, period, and timestamp contract.
2. Excluded by the proposed rule: historical trades outside the declared time window, including both wins and losses.
3. Unavailable: rows whose timestamp, timezone, strategy, cost, or grouping cannot support reliable membership.
4. Later skipped opportunities: signals observed after the rule was frozen but not executed; these require their own coverage and simulated-cost caveat.

These states prevent the most common denominator failure. The product-wide corpus may contain 600K+ processed trades, the teaching ledger contains 123, the London subject contains 48, and a later test will have another exact eligible count. None of those populations can silently stand in for another.

Audit What the Session Label Might Be Hiding

Before acting, compare the retained and excluded rows by instrument, setup version, direction, weekday, scheduled-event context, position size, cost coverage, holding time, and day concentration. A session result can disappear when a profitable setup or one outlier day is separated.

The session label can also hide access rather than edge. A trader may be alert during London and interrupted during New York; another may trade a strategy designed for a US cash open. Neither result makes the clock universally profitable or unprofitable.

Federal Reserve research using high-frequency interdealer FX data documents that trading volume and volatility vary materially through the day and around scheduled US releases. That supports treating time as a potentially informative market condition. It does not establish the return of this strategy or the best session for a retail trader; that conclusion must come from the trader's own comparable records.

Primary source: Federal Reserve International Finance Discussion Paper 823, intraday volume and volatility profile.

Turn the Historical Result Into One Later Test

1. Discovery: freeze the 123-trade ledger and record every session filter inspected.
2. Hypothesis: for this exact strategy, new entries are eligible only from 07:00 through 11:00 UTC.
3. Shadow log: keep later signals outside the window as skipped opportunities with their original timestamps; do not erase them.
4. Comparable period: predeclare a later date window and the minimum evidence required to make any decision.
5. Evaluation: compare eligible, skipped, and unavailable records using net result, expectancy, distribution, drawdown, costs, and composition.
6. Decision: keep, revise, or reverse the rule without rewriting the discovery window.

The before-and-after filtering workflow separates discovery from later evidence. A fixed calendar duration or trade count is not automatically sufficient; coverage, concentration, dependency, and stability matter too.

What About the Trades You Will Miss?

You will miss winners and losers. The honest test preserves both. Calling every excluded winner a “missed opportunity” while ignoring excluded losses is selection bias; calling every excluded loss “saved” while deleting excluded winners is the same error in reverse.

Use a shadow log when the setup signal can be observed without execution. Report how many eligible opportunities were seen, how many were skipped, how many could not be reconstructed, and whether simulated costs are comparable to executed-trade costs. If skipped-opportunity coverage is poor, the later result cannot answer the counterfactual cleanly.

When “Stop Trading That Session” Is the Wrong Advice

• One available window: if a job limits the trader to one period, compare setups or shorter subwindows instead of recommending an impossible schedule.
• Multi-day positions: entry session may be a weak proxy for the exposure that drove the result.
• Mixed strategies: decompose strategy first when each session contains different entry logic.
• Unstable timezone or labels: no reproducible bucket means no reproducible rule.
• One day or outlier dominates: the mean can change while the underlying repeatability stays unknown.
• No later evidence: keep the rule as an exploratory hypothesis, not a permanent ban.

For traders whose main constraint is time rather than raw P&L, the profit-per-hour analysis separates scheduled screen time, active decision time, and trade outcome instead of assuming that more hours are productive.

How TSB Supports a Personal Session-Fit Test

TSB has processed 600K+ imported trades cumulatively. That is product scale, not the denominator of this 123-trade example or evidence that one session wins across the user base. Every real session result still needs its own account, period, strategy, timezone, filters, eligible n, exclusions, unit, cost basis, and cutoff.

Observed analytics can retain session as one evidence dimension. Retrospective Backtester accepts an explicit session-filter hypothesis, preserves matching, control, and unavailable groups, and keeps exact trade references inspectable. Composition checks help reveal when the apparent “session effect” is actually instrument, setup, weekday, or direction mix.

TSB does not infer boredom, fatigue, discipline, or causality from P&L, and it cannot enforce a broker's trading hours. Its advantage is traceability: the rule, membership, excluded records, missing evidence, and later comparison can stay attached to the decision.

The Personal Session-Fit Workflow

1. Freeze account, strategy, period, timezone, assignment event, unit, costs, and evidence cutoff.
2. Define mutually exclusive session buckets and one unavailable state.
3. Reconcile bucket counts, wins, and net result to the complete eligible population.
4. Calculate per-bucket distribution, expectancy, drawdown, and day concentration.
5. Audit instrument, setup, direction, weekday, size, and event composition.
6. Record every filter tested so the chosen window is not presented as independent evidence.
7. Freeze one reversible rule in the trader's written trading rules.
8. Preserve skipped and unavailable opportunities in the later window.
9. Keep, revise, or reverse the rule at the predeclared checkpoint.

Methodology Note

• Preserved hooks: 123 trades, +$898 all-session net, +$1,536 London-only net, 75 non-London trades at -$638, and the exact 71.0% change remain visible.
• Evidence class: the ledger is a constructed, fully reconciled teaching example—not a customer outcome, TSB cohort, or population estimate.
• Correction: rounded win rates now resolve to visible integer wins, and the article states that the positive overlap row is excluded by the London-only rule.
• Selection boundary: the discovery comparison is in-sample; later or held-out evidence owns the operating decision.
• Product boundary: 600K+ is cumulative product scale and never substitutes for the 123-trade scenario denominator.

Final Verdict: Keep the 71% Result—and Test the Rule

The 71% result is worth keeping because readers can now audit every row behind it. The baseline contains 123 trades and closes at +$898; the London-only replay contains 48 of those trades and closes at +$1,536. The difference is +$638, or 71.0% of the baseline.

The correct next step is not “London is best.” It is a personal, reversible test: freeze the exact window, keep skipped and unavailable opportunities visible, compare later evidence, and inspect whether the result survives strategy and instrument composition. Strong arithmetic earns a test. Repeated comparable evidence earns a rule.