Trading Session Filter Case Study: The 180% P/L Math

Trading fewer hours can improve a result when the removed hours contain a repeatable negative slice—but only if the session math reconciles and the filter survives later evidence. This page preserves the original 180% P&L hook as a transparent constructed example. It no longer presents “Alex,” the first-person title, or unattached logs as a real customer case.

The example starts with 450 closed trades across six months, accounts for every trade and dollar, freezes a six-hour retained window, and then compares it with a separate three-month follow-up. The numbers are designed to teach the audit; they are not a TSB-wide benchmark, a causal experiment, or a promised outcome.

Disclosure: Trader's Second Brain is our product. TSB can group evidence by stored session and run an explicit session-filter hypothesis, but the session definition and final decision remain editorial and trader-owned. Current product facts render from the server catalog.

The Background: More Hours Do Not Create More Edge

A strategy has conditions under which it was designed to act. Time of day can be one useful label for those conditions, but “London,” “Asia,” or “New York” is not a cause by itself. Session can proxy for instrument mix, setup mix, scheduled events, spread and fee conditions, volatility, weekday, or the trader's availability.

The decision question is therefore not “Which session is best?” It is: for this frozen strategy and account, did a predeclared session slice differ enough—and with enough coverage—to justify a reversible filter? The session-performance guide covers the general framework; this page focuses on one fully reconciled worked example.

Freeze the Session Contract Before Looking at P/L

• Account and strategy: one comparable account scope and one strategy version.
• Period and cutoff: exact dates, timezone, and export/query timestamp.
• Unit: one logical closed trade; do not mix fills with grouped positions.
• Session assignment: the entry timestamp, exit timestamp, or another declared event—never whichever makes the bucket look better.
• Boundaries: exact clock ranges in the account timezone, including daylight-saving treatment.
• Money basis: net P&L after the costs actually represented in the data, plus conversion coverage when currencies differ.
• Missing policy: unspecified sessions remain excluded and visible; they do not become the control.

Session names are labels, not universal exchange facts. A custom “London core + first NY hour” bucket is legitimate if it is defined before calculation and applied consistently. The journal-fields guide explains the timestamp, setup, account, and cost evidence needed to reproduce it.

Worked Baseline: 450 Trades That Reconcile

Constructed example—not a customer or aggregate result. Window: six complete months. Starting population: 450 logical closed trades. Screen-availability window: 14 hours per trading day. Every trade has one session label, one stable ID, and net P&L in the same display currency.

The counts close: 252 + 72 + 54 + 72 = 450. Wins close: 151 + 30 + 15 + 20 = 216, or 48.0% of 450. Money closes: +$10,800 + $600 − $1,800 − $4,800 = +$4,800 over six months, or +$800 per month. Baseline expectancy is +$4,800 ÷ 450 = +$10.67 per trade.

The three removed buckets contain 198 trades and net -$6,000; the retained six-hour bucket contains 252 trades and net +$10,800. This is a strong exploratory signal, but it still does not prove the clock caused the difference. The retained bucket also has a different mix of trades.

Why the Old “64% of Screen Time” Line Was Corrected

The previous version called nine of 14 hours actively losing, even though one included overlap bucket was positive. That mixed “outside London” with “losing.” The reconstructed rule keeps six hours and removes eight: 8 ÷ 14 = 57.1% fewer scheduled screen hours. The removed trade buckets are collectively negative, and the positive overlap hour stays inside the retained definition.

Visible correction protects the useful insight without preserving broken logic. “Trading less” now refers to the exact 14-to-6-hour rule, not an inflated screen-time claim.

Why the Retained Window Might Differ

The example strategy is a defined-range breakout on major FX pairs. A session label may coincide with the range being complete, the trigger becoming eligible, a different event schedule, or a different execution-cost environment. Those are hypotheses to inspect in the records.

The 2025 BIS Triennial Survey reports that the United Kingdom remained the leading global FX trading location, with about 38% of turnover in April 2025. That establishes market-centre scale. It does not establish that London is the best session for this trader, that spreads were always lower in the sample, or that a particular breakout should work.

Open the retained and removed trades and compare instrument, setup, weekday, direction, event context, size, costs, and missing labels. If the “London effect” disappears after separating strategies or instruments, the original session verdict was confounded.

The Change: One Fixed Six-Hour Filter

The example freezes one prospective rule: new trades from this strategy are eligible only inside the retained six-hour window. The old 450-trade baseline stays immutable. No trades are deleted, re-labeled, or moved between sessions after seeing their outcomes.

The eight removed hours are not “bad forever.” They are outside this strategy test. A different strategy, market, instrument, or trader may have a different session profile. The filter is versioned so it can be reversed when later evidence no longer supports it.

Three-Month Follow-Up: How the 180% Math Works

The follow-up is a separate constructed window: 126 eligible closed trades over three months, 42 per month. It contains 73 wins, net +$6,720, and +$53.33 per trade. The complete-period average is +$2,240 per month.

The headline now reconciles: ($2,240 − $800) ÷ $800 = 180%. The trade counts, win counts, totals, averages, and time change all use their displayed denominators. No result is described as “of 600K+ trades.”

The comparison is still observational. The follow-up occurred later, so market regime, execution, costs, setup mix, and trader behavior may differ. The 180% change belongs to this constructed path; it does not isolate the filter's causal effect.

The Hidden Deal-Breaker: Filter-Selection Bias

A session filter discovered on the same data used to celebrate it is an in-sample finding. The worst bucket will often look unusually bad partly because it was selected as the worst. Deleting it and recalculating the same history exaggerates confidence.

Use the baseline to generate the hypothesis, not validate it. Freeze the session rule, then evaluate later or held-out trades. Keep every eligible and excluded record. The before-and-after filtering guide shows how to separate discovery, baseline, and follow-up windows.

There Is No Universal “30-Trade Session Rule”

The previous guide treated fixed trade counts and calendar durations as if they made a session verdict statistically meaningful. They do not. Thirty trades with one concentrated outlier, incomplete costs, or mixed strategies can be weaker than a smaller clean exploratory set; a much larger dependent sample can still be misleading.

Before acting, inspect:

• coverage of session, account, setup, instrument, timestamp, and costs;
• distribution and concentration—not only mean expectancy;
• number of independent trading days and whether one day dominates;
• overlap with strategy, instrument, side, weekday, and market regime;
• stability across non-overlapping windows;
• whether the prospective rule has later evidence.

If evidence is thin, report the counts and keep collecting. “Not enough to cut” is a valid decision.

How TSB Runs a Session-Filter Test

TSB has processed 600K+ imported trades cumulatively. That scale shows the system handles substantial history; it is not a benchmark for any personal session result. Each analysis still uses its own account, period, filters, eligible n, exclusions, unit, and denominator.

Observed analytics can retain session as an evidence dimension. Retrospective Backtester accepts a declared core or custom session filter, requires an explicit hypothesis, sends trades with unavailable session evidence to an unavailable group, and preserves exact record references. Its composition audit also compares instrument, session, weekday, and direction so a session label cannot silently stand in for a different trade mix.

TSB does not invent the reason a bucket differs and does not convert 600K+ processed trades into a universal London score. The trader defines the session hypothesis; the system keeps subject, control, unavailable records, coverage, costs, and evidence membership inspectable.

Four Mistakes That Create a Fake Session Edge

1. Moving bucket boundaries after seeing outcomes

If every losing trade is moved just outside “London core,” the label explains nothing. Version boundaries before calculation.

2. Mixing strategies or instruments

A session can appear strong because it contains one profitable setup or symbol. Audit composition and repeat the comparison within a stable strategy scope.

3. Treating missing labels as the control

Unknown session evidence is unavailable. Keep its count visible; do not assign it to “other sessions.”

4. Equating observed change with causality

A later window differs in more than the filter. Describe the result, preserve the rule, and collect repeated comparable evidence instead of declaring that the clock created the gain.

Session Filter Workflow

1. Freeze account, strategy, period, timezone, unit, money basis, and evidence cutoff.
2. Define mutually exclusive session buckets from one timestamp rule.
3. Reconcile bucket counts and missing evidence to the starting population.
4. Calculate net P&L, average result, outcome mix, costs, drawdown, and day concentration.
5. Audit composition by instrument, setup, side, weekday, and available regime evidence.
6. Open exact trades in the strongest and weakest buckets.
7. Freeze one reversible filter and one later comparison window.
8. Run the follow-up without deleting skipped opportunities or changing definitions.
9. Keep, revise, or reverse the filter from the later evidence.

If the decision is mostly about time spent rather than P&L, the profit-per-hour guide explains how to define active decision time without pretending all scheduled screen hours were productive work.

When Session Filtering Is the Wrong Analysis

• One session only: there is no within-strategy session contrast.
• Multi-day positions: entry time may be a weak proxy for the exposure that drove the result.
• Unstable labels or timezone: bucket membership cannot be reproduced.
• Mixed strategies: strategy composition dominates the session comparison.
• Thin or concentrated evidence: one trade or one day controls the result.
• No held-out or later evidence: the filter remains exploratory and should be tested reversibly.

Methodology Note

• Preserved hook: 450 trades, six-month baseline, 14-to-6-hour filter, $800-to-$2,240 monthly path, 180% change, 44% fewer trades, and 400% expectancy change remain in a mathematically closed example.
• Correction: the old 64% losing-screen-time line is corrected to the exact 57.1% scheduled-hour reduction because the old numerator included a positive bucket.
• Evidence class: all case numbers are constructed P&L examples, not real anonymized logs, a TSB cohort, or a customer testimonial.
• Causal boundary: baseline and follow-up are observational windows; no percentage is attributed solely to the filter.
• Product boundary: TSB's session dimensions, explicit hypothesis, unavailable-evidence handling, composition audit, and exact references are code-backed; the interpretation remains editorial.

For sourcing, correction, and evidence rules, see our editorial methodology.

Final Verdict: Let Your Data Choose a Test, Not a Myth

The 180% example is useful because the arithmetic now opens all the way down to 450 baseline trades and 126 follow-up trades. It demonstrates how a negative session slice can coexist with an overall profitable strategy and why fewer eligible hours may improve a later result.

Its conclusion is deliberately narrower than “only trade London.” Find the session hypothesis in your own data, reconcile every record, inspect composition, freeze one filter, and test it on later evidence. If the effect survives, keep the rule. If it does not, reverse it without rewriting the baseline.