Strategy Audit — When to Fix the System, When to Fix Yourself

A strategy audit checks whether its rules, costs and execution still match the documented assumptions. It should not begin with parameter changes after a few losses. We first examine data quality and compliance with the plan, then the results of individual setups, market conditions and estimation uncertainty. You will see how to distinguish execution problems from a possible deterioration in the market edge and when the evidence supports observation rather than a conclusion. Illustrative weekly, monthly and quarterly reviews lead to a specific decision: retain the rules, research a change, reduce risk or suspend trading. The thresholds are model assumptions to adapt to your own plan. The process helps prevent impulsive intervention and records the reasons for each change.

Record the hypothesis, rules, sample and results in Toolkit N: the strategy testing protocol. The form complements the method in this chapter with local saving, export and print.

1. Audit vs. Panic — The Fundamental Operational Difference

After six consecutive losing trades, an amateur trader opens their playbook and writes: "Setup A+ Long London Breakout — expanding Stop Loss from 1.5 ATR to 2.0 ATR because price keeps getting wicked out." They label this knee-jerk reaction an "audit." A genuine audit is initiated strictly when dictated by the calendar or by formal, pre-defined operational triggers — never because the pain of recent losses has overwhelmed emotional composure. Rather than fabricating stories about the market, you open your broker trade export, import executions into your trade journal, and audit execution integrity (tools and journal architecture: Article 15.9). The sequence is immutable and directly inverts amateur instincts: first verify data and execution integrity, second evaluate friction costs, third check market regime alignment, and audit strategy parameters last. An audit diverges from emotional panic across four measurable operational criteria:

Four Diagnostic Criteria Separating an Audit from Emotional Panic

The Panic Self-Test — Three Questions Before Modifying Any Parameter

Before placing your fingers on the keyboard to modify a stop-loss multiplier or indicator period, answer three diagnostic questions. If you lack an empirical, documented answer to any single one — you do not possess data to modify a strategy; you possess data demonstrating flawed record-keeping. Log the question in your audit journal along with the specific condition required before revisiting it:

1. Is your trading journal for this period 100% reconciled against your official broker statement, and can you identify the exact playbook version active on every trade? Without full statement reconciliation down to the cent, you are evaluating accounting errors rather than strategic edge.
2. Does the confidence interval for the difference between the period's net expectancy and your historical baseline lie entirely on one side of zero? Comparing two separate intervals and asking whether they overlap is too conservative: the intervals can overlap even when the difference is distinguishable. Evaluate the interval for the difference itself. As long as the interval of the difference encompasses zero, "underperformance" is a psychological perception, not a statistically demonstrated fact.
3. Does your proposed modification possess an empirical foundation rooted in market microstructure or execution friction, rather than "because it hasn't worked recently"? The former represents quantitative engineering; the latter represents retroactive emotional retrofitting.

In practice, a trader seeking to alter their stop loss after six consecutive losses cannot answer affirmatively to any of these three questions. Six trades represent a sample size from which a confidence interval proves absolutely nothing against a playbook threshold. They represent six data points from which ego attempts to pronounce a death sentence. Six consecutive losses within a strategy possessing a genuine 55% win rate is a routine statistical event expected across any substantial sample — especially when executing multiple setups and hundreds of trades annually. It is not an engineering failure; it is the natural reality of outcome variance.

The operational boundary between an audit and panic is precise and definitive. Without a rigid procedural firewall, every "adjustment after losses" is merely emotional rationalization. Hang this constitutional rule beside your workstation:

Every single modification to a trading playbook requires three mandatory components: an objective metric, an explicit numerical threshold, and a scheduled verification date. Lacking any of these three, an adjustment is not a system enhancement; it is an emotional reaction.

Panic does not produce a metric, threshold, and review date, because emotions resist waiting two weeks for more evidence. Put a barrier between a loss and a playbook edit: an audit date, a planned sample requirement, one parameter to change, and a log entry containing the metric, threshold, and verification date. The sample requirement must be interpreted through interval precision, not as a universal minimum n. Without that barrier, you can end up trading a different strategy every month even if the filename stays the same.

Decision Mini-Sheet: Am I Permitted to Modify This Setup?

Four conditions must be satisfied simultaneously for a setup modification to qualify as a valid strategy audit rather than an emotional reaction:

1. Sample Adequacy: The confidence interval for setup expectancy is sufficiently narrow to resolve against the playbook threshold — not merely "I have accumulated 30 trades." If the interval spans across the threshold, decision-making is deferred, regardless of nominal trade count.
2. Durability: Performance remains below the demotion threshold (for example, below +0.15R) across two consecutive audit periods, rather than one weak quarter. This is the scenario's persistence rule, not proof that all seasonal effects have been excluded.
3. Pristine Execution: Rule breach rate is verified strictly < 8% across the period, and friction costs are fully reconciled against official broker statements. If you breached rules or broker spreads doubled, realized performance reveals nothing about the setup.
4. Microstructural Rationale: The proposed modification is justified by an identified shift in market structure, broker liquidity, or execution friction ("median stop slippage expanded permanently from 0.3 to 1.4 pips, altering net friction in R"), never by a narrative excuse ("it feels out of sync").

If even one of these four conditions fails — you do not modify the setup; you maintain observation status. A trader considering a parameter change must check all four boxes. Three "yes" checks are insufficient. Four "yes" checks permit entering the calibration protocol in Section 4 — while the change-testing process — in-sample calibration, out-of-sample validation, and forward testing — belongs to Article 15.4.

Minimal Data Package Required for a Strategy Audit

Without the following discrete fields, an audit is merely speculative storytelling:

• trade_id, date, pair, setup_id, tier (A+/A/B/C — definitions in Article 15.2), direction
• entry, SL, TP, exit, r_net
• mae_r, mfe_r (maximum adverse / favorable excursion in R) — essential to evaluate whether stops were placed too tight or targets too close
• slippage_pips — entry and exit slippage, evaluated via medians per setup
• rule_breach (0/1) — boolean flag indicating whether any playbook execution rule was breached
• playbook_ver — exact playbook version active at order entry (e.g., v2026.1), without it, you cannot locate the boundary needed to compare the pre-change and post-change periods
• screenshot_before, screenshot_after — chart captures at trade entry and exit for retrospective setup reconstruction

A spreadsheet logging only date, pair, direction, and profit in PLN is an archive of past transactions, not an auditable performance dataset. Full journal specifications and broker reconciliation mechanics: Article 15.9.

2. When to Audit — Calendar and Event Triggers

A strategy audit has two triggers and only two: calendar (end of each quarter) and event (breached drawdown limit from the trading plan, an execution breakdown, or altered broker trading conditions). Everything below this threshold — daily trade logging and weekly execution reviews — is not an audit and never concludes with a structural strategy decision. The routine of trade logging is documented in Article 15.9, and weekly workflow rhythms are established in Article 15.5. Here, a single operational boundary remains: who observes, and who decides.

Between scheduled audits, your playbook is strictly frozen. This is not an aesthetic guideline: every mid-period adjustment destroys the statistical sample before you have even measured it. By quarter-end, you will no longer know which version of the strategy you are evaluating. There is exactly one exception, and it concerns operational risk, not statistical edge — halting trading following a plan limit violation, execution failure, or infrastructure outage is permitted always and immediately, without waiting for sample size (circuit breakers, STOP limits, and kill switches: Article 15.2).

When Trade Frequency is Low

A full-time working professional trading swing or D1 setups across 3–5 pairs typically logs a dozen or so trades per quarter. This does not mean there is nothing to audit — it means not all conclusions can be extracted from that sample alone. After n = 12 trades, a 95% Wald confidence interval for win rate spans roughly ±28 percentage points (assuming p = 0.5 and independent observations; with autocorrelated trades or clustered market events, the interval widens further). Any assertion of statistical edge drawn from such a sample is pure randomness masquerading as diagnostics. Conversely, auditing data integrity, trading costs, and execution discipline is valid and necessary across any sample size.

• Quarterly, regardless of n — Gates 1–3: reconcile journal entries with broker statements, verify entry completeness, inspect rule_breach rates, compare execution costs and slippage against baseline, and confirm played setups match playbook specifications.
• Gate 4 (Statistical Edge) — evaluate only when the confidence interval for expectancy narrows sufficiently to resolve against your playbook threshold. At a dozen trades per quarter, this typically requires a semi-annual or annual pooled sample rather than a quarterly assessment.
• 90-Minute Format for Working Professionals — split the session: Saturday (30 min) for data preparation and broker reconciliation, Sunday (60 min) for gate evaluations and calibration decisions. Very few traders possess 90 minutes of razor-sharp analytical concentration on a Sunday evening.

Tax accounting is not part of a strategy audit. After-tax results, conversion at an NBP reference rate, and loss offsets in the Polish source belong to Section 14. The original example concerns Poland and PIT-38 (Polish tax return); international readers must use the rules for their own tax jurisdiction — your trade journal provides source audit documentation, but is not a tax calculation engine. In a strategy audit, you work strictly with net performance after transaction costs; after-tax figures factor in only when deciding on capital scaling and profit withdrawals (Article 15.11).

3. The Four Audit Gates: Data, Costs, Regime, Edge

A strategy audit proceeds in a single, unyielding sequence that runs exactly counter to trader instinct. Instinct urges: “Let’s check whether the strategy is still making money.” Professional protocol dictates: first establish whether you possess the evidentiary material that permits asking that question at all. A quarterly return generated while breaking rules or absorbing changed transaction costs tells you nothing about strategy design — neither good nor bad. This is why the edge evaluation gate is positioned last, not first. Skipping this order does not accelerate the audit; it degrades disciplined analysis into guesswork.

Gate 1: Are Data and Execution Reliable?

The first question does not concern market direction; it audits operational logging integrity through four controls:

• Broker Statement Reconciliation — trade count, traded volumes, commission charges, and overnight swaps in your journal must reconcile to the broker’s official monthly/quarterly ledger. Discrepancies signify missed trades, unrecorded partial fills, or duplicate entries. Reconciliation procedures and root causes of platform desynchronization are detailed in Article 15.9.
• Playbook Versioning — verify which trades were executed under which explicit rulebook version (playbook_ver). Without this field, you are comparing quarterly results where fundamentally different strategies shared the same generic name.
• rule_breach Rate — the proportion of trades executed with a documented rule violation. The higher this metric, the less your results reflect system edge and the more they reflect personal execution breakdown.
• Dataset Completeness — ensure the sample is not missing emergency manual exits, rejected market orders, or a sequence after which you stopped recording trades. An audit encompasses every single trade in the calendar window, especially the uncomfortable ones.

If any of these controls fail, your audit terminates immediately at Gate 1: you repair logging practices and execution discipline, deferring strategy rule evaluation to the following period. This is not postponing a decision — it is the decision, targeted at operational reality.

Gate 2: Are Friction Costs Identical to the Baseline Period?

The second gate evaluates transaction friction, because rising costs degrade performance without changing a single word in your playbook. You compare median entry slippage, stop-loss slippage, take-profit slippage, turnover commissions, overnight swaps, and currency conversion markups against your historical baseline. Slippage is signed: at entry, unfavorable fills occur worse than intended price for the entry direction; at exit, worse for the closing direction. Divide the price difference by the instrument’s pip size; spread is already embedded within fill prices — do not deduct it twice. The conversion is direct: cost in R equals cost in pips divided by initial SL distance in pips (e.g., 0.8 pips slippage on a 50-pip SL equates to 0.016R drag). Execution cost definitions and metric formulas: Article 15.9.

Furthermore, one fundamental denominator rule must hold, without which your entire audit unravels: R0 — the planned monetary risk established prior to order transmission — remains strictly fixed regardless of trailing stop adjustments, partial profit takes, or ultimate trade outcome. If you altered your R calculation methodology mid-quarter, you are not comparing performance across two periods; you are comparing two incompatible measurement units.

Gate 3: Does Market Regime Fall Within Setup Card Specifications?

The third gate asks whether the market environment matches the operational domain for which your setup was calibrated. Volatility and trendiness thresholds (ADX, ATR, spread percentiles) represent strategy parameters, not universal market standards. A complete catalog of objective screening filters is provided in Article 15.6, while calibration procedures on empirical data are detailed in Section 6. A verdict of “regime out of bounds” is not an indictment of strategy viability: it confirms that current data reflect environmental conditions outside the setup’s tested boundaries, rendering edge conclusions statistically inapplicable.

Gate 4: What Do the Data Indicate About Edge?

Only upon passing the first three gates do you calculate expectancy in R net of transaction costs — aggregate and per setup — and never as an isolated point estimate. A point estimate without a confidence interval is not an actionable metric. Formulas, bootstrap iterations, and CI mechanics are detailed in Article 15.9 (along with core definitions of expectancy, profit factor, and SQN); interpretation belongs here and crystallizes into a single axiom:

The number of trades alone decides nothing. What decides is the width and position of the confidence interval relative to the decision threshold established in the playbook prior to the audit.

Consequently, Gate 4 yields three operational conclusions, not two:

• Confidence Interval Entirely Above Maintenance Threshold — the necessary statistical condition is satisfied; the setup remains in its current tier without alteration. This does not “prove” perpetual edge; it confirms the absence of statistical grounds for modification.
• Confidence Interval Straddles Decision Threshold — decision deferred to the next audit cycle with documented re-evaluation criteria. The setup is neither promoted nor deactivated.
• Confidence Interval Entirely Below Threshold for Two Consecutive Audits — under verified clean execution and reconciled friction costs, the setup is demoted or retired according to the playbook calibration matrix in Section 4.

What Gate 4 explicitly avoids: declaring that an edge has vanished after a single quarter below threshold. A single subdued quarter with historical expectancy of +0.40R falls well within standard sampling variance — a strategy possessing a true long-term edge of +0.30R will experience quarters of +0.50R alongside quarters of +0.10R. Consistency across multiple periods and CI width dictate strategic decisions, never an isolated datapoint. For instance, a streak of six consecutive losses with a 55% win rate represents an expected tail event over larger samples: assuming independent trials, six consecutive losses in a pre-specified block carries a probability of 0.45⁶ ≈ 0.83%. “Expected in the tail” does not mean a calendar schedule — it describes a probability distribution.

Gate 1 Diagnostics — What to Inspect in Execution

• Median Slippage on Entry, SL, and TP — Evaluated Separately. Comparing SL slippage against TP slippage is critical: if stop losses experience systematically worse execution drag than take profits, asymmetric execution friction is quietly eroding expectancy. If median slippage jumped 50% over baseline, did you switch brokers, alter account types, or trade during illiquid session crossovers?
• Rule Breach Percentage — baseline was 5%, current quarter is 18%? You are skipping confirmation filters, adjusting stops mid-trade, or taking premature profits.
• Median Latency from Signal Trigger to Entry — has reaction time doubled or tripled over the quarter? You are hesitating at order entry due to recent drawdown anxiety.
• Setup Tier Distribution Played — baseline was 60% A+ setups, current quarter is 30%? You are trading lower-probability B and C configurations simply “to stay in the market.”

Execution diagnostics are inherently uncomfortable because trade journals reveal behaviors traders prefer to hide: premature exits, moved stops, tier-B trades rationalized as A+, and boredom-induced trades. Without a detailed journal, you cannot answer this gate — you can only answer “how disciplined do I feel,” which is entirely uninformative.

Mandatory Audit Decision Matrix

The four gates are documented within a unified decision table, logging one row per plausible root cause. This matrix represents the true deliverable of your audit — not an equity curve chart or an aggregate report, but a row that yields an immediate operational decision alongside an explicit re-evaluation trigger. Below is the mandatory matrix completed for the canonical scenario: an identical performance decline stemming from three entirely different root causes.

Shared Input Across All Three Variants Illustrative example using synthetic data, not an account ledger.: Intraday setup with a 20-pip SL, historical baseline expectancy of +0.30R over the prior four quarters. In the current quarter, performance dropped to +0.18R across 60 trades (36 losses); with std(R) ≈ 1.3, the 90% confidence interval calculated via Section 4 is (-0.10R; +0.46R). The headline number (+0.18R) is identical across all three cases — only the underlying diagnostic data differ. This is a separate setup and sample from the Section 4 heatmap.

Visual summary of the decision matrix above: one headline metric (+0.18R), three mutually exclusive root causes, and three entirely different operational decisions. Synthetic figures; independent from Section 4 heatmap.

Three rows, one headline metric. Had the audit begun with an unguided inquiry into edge, all three cases would have received the identical diagnosis — and in two out of three scenarios, that diagnosis would have been completely false. Note the “Immediate Decision” column: in none of the variants does it read “the strategy is dead” or “paralyzed waiting for certainty.” Operational decisions remain fully actionable despite statistical uncertainty because they govern position sizing, friction costs, and execution discipline — factors within your direct control — rather than hasty verdicts on edge that current data cannot support.

4. Playbook Calibration — Demotion, Promotion, Deactivation

A trading playbook is not an immutable stone tablet. Every quarter, you rigorously audit each setup against objective thresholds documented prior to the audit — because without this discipline, an A+ setup quietly degrades while a B-tier pattern might gain statistical traction without receiving capital allocation. After eight quarters without structured calibration, a playbook accumulates a dozen or more setups, many of which are dead weight, while capital remains distributed indiscriminately.

Definitions of setup tiers (A+, B, C), mandatory setup conditions, quality grading rubrics, and baseline position sizing per tier belong to your trading plan — Article 15.2. An audit does not invent these rules; it evaluates whether an existing setup still satisfies the statistical criteria of its assigned tier and executes one of four operational decisions.

The Four Setup Audit Decisions

• Maintenance — the setup performs within historical parameters; zero modifications.
• Observation (Deferral) — the confidence interval spans the decision threshold; setup tier remains unchanged, but a mandatory re-evaluation condition is logged.
• Demotion — the setup is downgraded by one tier (e.g., A+ → B), triggering the corresponding reduction in baseline position sizing.
• Deactivation (Archival) — the setup is removed from the active playbook, the playbook receives an incremented version number, and the setup is placed into cold storage until re-validated via formal backtesting (Article 15.4).

Criteria Table — Based on Confidence Intervals, Not Point Estimates

The following benchmarks are illustrative procedure assumptions, not universal standards. The values below represent operational assumptions calibrated for retail FX models with an average win-to-loss ratio between 1.5 and 2.0. You must calibrate these cutoffs against your own empirical payoff distribution and cost structure, locking them into your playbook before conducting the audit. These thresholds act as operational circuit breakers, not definitive mathematical proofs — their purpose is to halt emotional tampering, not to proclaim that an edge has vanished forever.

Two overarching precedence rules govern this entire table, superseding individual row conditions. First: as long as a confidence interval straddles a decision threshold, the “Observation” rule governs — no other row applies, regardless of how attractive or concerning an isolated average may appear. Second: no tier modification takes effect after a single audit — conditions for demotion, deactivation, and promotion must be sustained across two consecutive quarters under clean execution and verified friction. These two rules prevent overlapping triggers: either the statistical interval resolves conclusively, or the decision waits for data. Conversely, a single stellar quarter from a Tier-B setup (e.g., +0.55R) does not warrant promotion, as the subsequent quarter may mean-revert to +0.15R. Two consecutive cycles do not eliminate randomness entirely, but they reduce the chance of restructuring your playbook based on transient luck.

The four setup audit decisions governed by confidence interval position relative to the decision threshold. Numerical cutoffs correspond to the table above; setup tier definitions and sizing rules belong to Article 15.2.

Playbook Heatmap — Quarterly Diagnostic Matrix

Illustrative example using synthetic data, not an account ledger. The figures below are synthetic and mathematically reconciled: sample size n per setup sums exactly to the 87 total trades logged during the period, expectancy reflects stated win rates and payoff ratios, and confidence intervals are computed via E ± 1.645 × std(R) / √n using the population standard deviation of the illustrated trade vector (division by n), assuming independent observations (under autocorrelated runs or clustered market volatility, empirical intervals are wider). For sample sizes around a dozen trades — as in the News Fade row — normal approximations are purely indicative: because R-distributions possess heavy tails, the standard formula often underestimates interval width at n = 13. Approaches to small-sample uncertainty and bootstrap calculations are detailed in Article 15.9.

Aggregate quarterly performance is derived from the exact same dataset: n = 87, Win Rate 52% (45 wins), aggregate expectancy +0.32R, 90% confidence interval (+0.09R; +0.55R), profit factor 1.66. The rule_breach column comprises the exact trade counts: 2 of 34, 1 of 22, 1 of 18, and 2 of 13 — totaling 6 of 87 trades, or 6.9% logged into the audit ledger in Section 9. Because this sits below the maximum allowable threshold (< 8%), Gate 1 passes and Gate 4 edge metrics may be evaluated. The sum of parts must reconcile to the whole — if your quarterly review does not reconcile, you are missing trades or double-counting fills, requiring an immediate return to Gate 1.

Why Four Rows Yield Four Deferral Decisions. This is not an oversight of the example; it is its central lesson. With quarterly samples ranging between 13 and 34 trades, the half-width of a 90% confidence interval spans from 0.40R (LON Breakout, Asia Range) to 0.56R (News Fade) — larger than the distance separating each setup’s observed expectancy from its tier threshold. None of the four intervals lies entirely on one side of a threshold. Consequently, the precedence rule overrides all other rows: even LON Breakout cannot claim definitive maintenance status purely from this sample, as placing the lower bound of this 90% interval above the threshold would require an observed expectancy around +0.70R. A quarterly heatmap is not an executioner’s block; its purpose is to highlight which setups are drifting toward decision boundaries and establish explicit re-evaluation triggers. A verdict requires an interval narrow enough to resolve against the threshold, together with the persistence and execution conditions for the relevant decision. Repeating an inconclusive result does not override the Observation rule.

Why News Fade is Not Immediately Deactivated. The emotional urge to scrap it is powerful: a second consecutive negative quarterly expectancy, and pooled with the prior audit, n = 36 with E = -0.03R. However, the pooled 90% confidence interval for those 36 trades spans (-0.36R; +0.31R), covering both zero and the maintenance threshold for Tier B. The data cannot distinguish a broken setup from a viable +0.30R setup experiencing expected variance. Scrapping it immediately would represent an ungrounded extrapolation. Disciplined operational response looks different: scale down position sizing per Article 15.11, treat the 15% rule_breach rate as an urgent execution defect, log an explicit conditional trigger — “deactivate if the upper CI bound falls below +0.15R at the next audit, with the table's execution and persistence conditions also satisfied” — and set a firm review date. This marks the boundary between an actionable operational decision and an unverified statistical judgment.

Color-coding in your spreadsheet (green / yellow / red) must map to the audit criteria table, never to isolated win rates or raw averages. A quarterly diagnostic heatmap reveals within seconds where capital is underperforming; running this heatmap weekly produces mostly noise and tempts you into making structural changes on three trades.

Parameter Adjustments Following an Audit

When an audit legitimately justifies modifying strategy rules, the calibration process itself does not occur during the audit session: you adjust a single parameter, validate it on out-of-sample data, and deploy it live under reduced position sizing. The complete workflow, sample partitioning rules, and safeguards against data mining (overfitting, multiple testing adjustments, trial logs) are detailed in Article 15.4. During the audit, a single iron rule governs: never optimize rules on the identical sample where the problem was discovered, and never present that historical backtest as independent proof of edge.

A playbook that only grows dilutes trading capital across setups of decaying quality. Retiring unviable setups is just as crucial as adding new ones, requiring the identical rigor: confidence intervals, two consecutive audit cycles, and an explicit re-evaluation trigger.

5. Event-Driven Audit — Drawdown and Winning Streaks

The second audit trigger is event-driven: a breached drawdown threshold from your trading plan, an execution breakdown, or an unexpected change in broker trading conditions. An off-calendar audit navigates the identical four gates in the exact same sequence — only the data window narrows to the event and its immediate perimeter. It never replaces a scheduled quarterly audit, nor does it alter the quarterly calendar date.

Deep Drawdown — What an Audit Can Legitimately Establish

Two strategies sharing the exact same win rate and identical historical expectancy can exhibit radically different drawdown risk profiles if their underlying R-distributions differ in loss tails or win clustering. Therefore, the central question of an event-driven audit is not “is this drawdown painful,” but “does this drawdown fall within the expected probability distribution for this specific payoff vector and position sizing model?” You extract this distribution via Monte Carlo simulation on your empirical trade records — the complete procedure, reproducible code with random seeds, capital modeling, and block bootstrap adjustments for trade autocorrelation are established in Article 15.4. During the audit, you reference these simulated percentiles rather than recalculating them on the fly; remember that drawdown percentiles expressed as percentages of capital presuppose a specific position sizing model, meaning they reflect risk architecture rather than pure setup edge.

The magnitude of position sizing reductions and the exact criteria for restoring standard risk are governed by capital scaling protocols — unified across this curriculum in Article 15.11. The audit issues the operational verdict (“reduce,” “maintain,” “restore”) and logs it alongside the re-evaluation trigger; the exact sizing tiers are drawn directly from Article 15.11 to avoid conflicting protocols within your playbook.

One mathematical reality must be calculated once and remembered permanently: drawdowns do not recover linearly. Returning to the starting high-water mark requires a percentage gain of d / (1 - d) on remaining capital: recovering from a −10% drawdown requires +11.1%, −20% requires +25.0%, −30% requires +42.9%, and −50% requires +100.0%. This is a hyperbolic curve, not a straight line: every additional 10 percentage points of capital erosion requires a larger recovery gain than the preceding ten-point decline. The operational takeaway is immediate: intervene early with modest capital reductions rather than waiting for a severe drawdown zone from which recovery demands returns beyond the strategy’s mathematical capacity.

What Never to Do Following a Deep Drawdown

• Never increase position sizing to “recover faster” — it increases both the pace of a possible recovery and the pace of a further, deeper drawdown.
• Never alter strategy parameters mid-drawdown — first enforce execution stabilization and pass Gates 1–3; only then evaluate strategy rules.
• Never hunt for an “alternative strategy that is currently working” — selecting a new system based on recent short-term performance commits the classic data-fitting fallacy at the portfolio level.
• Never trade on borrowed capital or injected debt — injecting borrowed funds following a drawdown converts market risk into legal financial liabilities.

The inverse extreme is equally destructive: slashing position sizing by 80% following an ordinary −18% drawdown and trading a single setup for six months. Sizing reductions are by definition temporary and must have an explicit reinstatement condition documented before deployment. If a “temporary” reduction persists past a full quarter, it has ceased to be a risk intervention and has become an unrecorded, ad-hoc revision of your trading plan.

Winning Streaks — The Identical Audit in Reverse

A protracted winning streak triggers an event-driven audit for the exact same underlying reason as a losing streak: it alters trader psychology and behavior before it shows up in negative results. The universal behavioral pattern involves unearned sizing increases, excessive trade frequency, and loosening entry filters below Tier-A+ standards. This mechanism aligns with empirical findings across retail trading accounts (Barber & Odean^[1]): excessive turnover and hyperactive trading correlate systematically with degraded net performance — the cited study concerns equities, not CFDs, so this is broader contextual evidence, not a measured finding for the FX market. The cognitive underpinnings are documented across the behavioral finance literature on overconfidence and recency bias (Kahneman^[2]; Steenbarger^[3]).

The operational rule is brief and non-negotiable: position sizing percentages never increase in response to recent winning runs. Sizing has exactly two states: baseline risk defined in your trading plan, and a reduced risk state triggered during drawdown protocols with an explicit calendar return date. Under this framework, nominal monetary risk expands exclusively through organic account equity growth: risking 1% on a larger balance allocates more capital automatically without altering the risk percentage. Conditions, sequence, and rollback of a scale change: Article 15.11.

Early warning indicators to inspect in your trade journal before a winning streak collapses into overtrading: five or more consecutive wins, weekly returns several times higher than typical, a declining proportion of Tier-A+ setups among executed trades, and shortening latency between trade executions. All four indicators can be extracted from the standard journal columns documented in Article 15.9 — they require no specialized software, merely the discipline to inspect them when emotional euphoria is at its peak.

6. Market Regime — When a Threshold is a Filter vs. Decoration

Market regime represents the underlying statistical environment: prevailing volatility, directional trend persistence, cross-asset correlations, and liquidity depth. Regimes shift — sometimes over weeks, sometimes gradually across multiple quarters. A trading strategy engineered for one regime will not necessarily thrive in another — and this is not a failure of strategy logic, but an environmental mismatch against the setup card’s domain. A comprehensive taxonomy of regime metrics and mechanical screening filters is provided in Article 15.6; an audit addresses a distinct question: verifying whether your regime filter genuinely separates statistical performance or is merely decorative.

Calibrating Regime Filters on Empirical Data

An entry filter such as “do not trade if ATR exceeds 1.5× average” remains pure decoration until you establish whether it correlates with an actual decline in the expectancy of your specific strategy. The mandatory calibration protocol prior to codifying any regime filter into your playbook proceeds across six steps:

1. Extract all historical trade records from your journal — for identical outcome variance, a larger sample produces a narrower confidence interval in Step 4 (interval width scales as 1/√n).
2. For every trade, log the exact numerical reading of the regime indicator at the moment of order transmission (e.g., D1 ATR(14) divided by its 90-day moving average). This must be an objective journal column, not a retrospective guess.
3. Partition the sample into two subsets according to your proposed threshold: “favorable regime” and “adverse regime.”
4. Compute the expectancy of each subset independently, and then compute the 90% confidence interval for the difference between the two expectancies — this difference interval is what resolves the question, not simply comparing two separate intervals.
5. If the confidence interval for the difference includes zero — your proposed filter is decorative, not operational, regardless of how intuitive it sounds. The mere fact that two independent confidence intervals overlap does not prove this conclusively; such an overlapping test is overly conservative and can conceal true divergence.
6. If the confidence interval for the difference lies entirely on one side of zero, the threshold genuinely separates performance — yet this remains an in-sample hypothesis extracted from historical data. Before locking it into your playbook, track its performance forward across new trades.

A trader who uncritically copies external indicators often imports a rule like “never trade when the VIX exceeds 25” into a currency strategy where equity volatility bears zero correlation to setup expectancy — and then wonders why the filter fails to improve returns. A technical metric cited in trading literature represents a testable hypothesis, not an established fact about your specific system.

Crucial Caution for Mean-Reversion Strategies: verify whether elevated volatility correlates with superior rather than degraded expectancy — for many mean-reversion models, the empirical relationship is the exact opposite of trend-following systems. High ATR generates wider deviations from moving averages, providing richer reversion opportunities. In contrast, a trend-following system in identical choppy conditions suffers constant whipsaws. The exact same volatility reading represents a risk reduction signal for one setup and a full planned-allocation signal for another. Without empirical calibration on your own trade logs, you risk slashing position sizing during conditions most favorable to your own strategy.

Distinguishing Structural Regime Shifts from Transient Noise

• Persistence — the indicator signal persists for at least two consecutive weeks, not two trading sessions. Transitory volatility spikes following economic releases do not constitute a structural regime change.
• Multi-Indicator Confluence — confirmed across at least two independent indicators from the framework in Article 15.6 that have passed the calibration protocol above. A single indicator reading is insufficient.
• Identifiable Fundamental or Structural Catalyst — monetary policy divergence, geopolitical realignment, or liquidity regime shifts. A technical indicator divergence lacking an identifiable macroeconomic anchor is frequently a false alarm.

An audit does not attempt to forecast how long a market regime will last — you inspect regime indicators at the next audit and adapt capital scaling to prevailing conditions. Operational adjustments in an unfavorable regime govern position sizing and concurrent trade limits (Article 15.11), never core strategy rules: altering entry logic in the middle of an adverse regime creates an unverified new system at the worst possible moment.

7. When to Fix the System vs. When to Fix Yourself

The most consequential judgment in a strategy audit. Returns are deteriorating — yet the question is never “what parameters can we tweak,” but “where does the failure originate: in the statistical design of the system, or in the human executing the rules?” These two root causes demand fundamentally opposed responses. Conflating them exacerbates performance decay: you dismantle rules that are mathematically sound while leaving catastrophic behavioral habits untouched.

Diagnostic Signals of a System Problem

A technical design defect — residing within the rules themselves — manifests through distinct quantitative signatures:

• Structural Elevation in Median Slippage (e.g., jumping from 0.3 to 1.1 pips) — your broker has altered liquidity routing, spread profiles have widened, or your execution window has drifted into illiquid session transitions.
• Stable Win Rate Accompanied by Collapsing Average Win R-Multiple — e.g., win rate remains steady at 55%, but average winner drops from +1.8R to +1.2R. This may indicate profit targets are set too tight for expanding volatility, or a trailing stop logic is suffocating trades prematurely.
• Setup Frequency Drops Substantially Below Historical Norms — the mechanical regime filter may be overly restrictive for current baseline market conditions.
• Signals Fire During Sessions Without Historical Precedent — e.g., the source's setup historically appeared at 09:00–11:00 Polish local time and now also appears after 14:00 in that same timezone (Europe/Warsaw, CET/CEST; retain the dates and timezone when comparing sessions) — signaling a shift in underlying session microstructure.
• Rule Specifications Denominated in Units Incapable of Adapting to Volatility — e.g., a stop specified as a fixed number of pips when volatility rises beyond the setup card's range. Caution against a common misconception: a stop defined as a multiple of current ATR scales dynamically, so if ATR doubles, a 1.5×ATR stop expands automatically. A rising ATR alone does not prove a stop is too tight — empirical MAE distribution relative to stop distance resolves the question, not raw ATR expansion.

Diagnosis: System design. Operational response: calibrate parameters, but strictly one parameter at a time, followed by rigorous out-of-sample testing before deploying to a live account (Article 15.4).

Diagnostic Signals of an Execution / Behavioral Problem

A psychological and operational defect — residing within the trader’s discipline — reveals itself through unmistakable journal footprints:

• Sharp Escalation in rule_breach Percentage — baseline was 5%, current quarter sits at 22%. The system rules remain identical, but you have stopped obeying them.
• Modifying Stop Losses Mid-Trade — widening a stop “because the market is just testing support.” This is not risk management; it is loss aversion masking an unwillingness to accept a statistical loss.
• Prematurely Exiting Positions Before TP1 — driven by the fear of surrendering unrealized gains, systematically amputating the right tail of your payoff distribution.
• Executing Sub-Par Setups Rationalized as “Almost A+” — fear of missing out (FOMO) masquerading as discretionary chart reading.
• Trade Frequency Surges Without More Opportunities — entering Tier-B and Tier-C configurations you previously skipped.
• Decision Latency Lengthens Dramatically — hesitating at entry, second-guessing signals, and changing your mind.
• Revenge Trading Footprints in the Journal — immediate follow-up orders transmitted within minutes of a stopped-out position “to win back the loss.”

Diagnosis: Execution breakdown. Operational response: do not touch strategy parameters. Address execution: scale down position sizing per Article 15.11, recite playbook rules aloud prior to every session, consult a mentor or audit partner, and sometimes take a one-week pause.

Diagnostic Comparison Matrix — System vs. You

Formal Quarterly Audit vs. Quick Sanity-Check

Both diagnostic processes follow the identical gate sequence — they differ in analytical scope and operational endpoints:

• Formal Quarterly Audit — complete traversal of all four gates (Section 3), population of the decision matrix, setup tier calibration, and a documented log entry establishing operational decisions and re-evaluation triggers.
• Quick Sanity-Check — deployed immediately following an emotional impulse (“the system is broken,” “I need to overhaul everything” mid-quarter). It halts abruptly at Gate 1: have you executed the system cleanly across the last n trades, and is the journal complete? If the rule_breach rate exceeds a few percent in the source's quick-check example, all downstream questions are declared premature. A sanity-check never concludes with a playbook modification — it concludes either with psychological stabilization or a decision to trigger a formal event-driven audit.

A sanity-check resolves one fundamental question: does your dissatisfaction stem from poor execution or broken rules? Until you separate these two domains, modifying strategy parameters is pure gambling — and the answer resides in your trade journal, not in your memory.

Operational Decision Table by Root Cause

To investigate “system vs. me,” the single most potent diagnostic tool is comparing the Maximum Adverse Excursion (MAE) and Maximum Favorable Excursion (MFE) across every trade — keeping in mind that both metrics measure price path during the trade’s lifespan, providing hypotheses rather than absolute causal proofs. If MFE systematically approaches target levels while realized exits cluster far lower, premature exit is a hypothesis to investigate (trader execution). If MAE regularly approaches the stop level without breaching it before the trade reaches its target, stop distance is a hypothesis to investigate (system design) — however, price movements after a trade is stopped out do not belong to that trade’s MAE or MFE. Evaluating whether “the stop was just a whisker too tight” requires an alternative exit test via backtesting (Article 15.4), not post-hoc chart gazing. Metric definitions, 4-cluster scatter plot interpretations, and diagnostic rubrics are detailed in Article 15.9. Without MAE and MFE columns, Gate 1 degenerates into subjective introspection (“I think I executed well”), which carries zero evidential weight.

Three Intervention Tiers for Execution Breakdowns

When an audit conclusively diagnoses an execution failure, you deploy one of three structured intervention levels based on the severity of the breakdown:

Returning to a demo account after trading real capital is not meant to wound your ego. Its sole purpose is to verify whether you can execute 30 consecutive trades without tampering with stops, targets, or setup criteria. If you cannot execute with discipline on demo, continuing on a live account merely forces you to pay an exorbitant tuition fee for the identical lesson. A trader who steps back to demo for 4 weeks of flawless execution and returns with restored discipline is in an incomparably superior position to one who spends 6 months hemorrhaging capital while battling personal psychology.

What NEVER to Change When the Problem is You

The most ubiquitous error in this scenario is attempting to “engineer around” personal discipline flaws by altering the trading system. A trader who consistently cuts winning trades prematurely decides to move the take-profit target closer “so it feels less tempting to exit early.” This severely compounds the problem: tightening a profit target changes two variables at once — it shrinks the average win while boosting the win rate — meaning its net impact on expectancy is uncertain and requires formal alternative exit testing (15.4). What is entirely certain is that moving the target fails to address the underlying psychological trigger: anxiety around giving back unrealized gains. Operational rule: if the diagnosis is execution, never modify the strategy. In the scenario where the system is sound and execution is the problem, rule changes do not remedy the failure. A breach diagnosis alone does not establish that the system has an edge. Tweaking rules to compensate for emotional weakness merely masks the symptom at the expense of your mathematical edge.

8. Top 10 Strategy Audit Errors and How to Avoid Them

Error 1: Tweaking Parameters on a Non-Informative Sample

The hallmark error of developing traders. After five or ten trades, the trader concludes that “the setup is broken” and modifies entry indicators. The mathematics of small samples expose the fallacy: across 10 trades with 5 wins, the 95% Wilson score confidence interval for win rate spans from 24% to 76% (assuming independent trials; with serial correlation, it is even wider). An observed 50% win rate sits inside an interval that stretches all the way from “virtually useless” to “world-class edge.” Furthermore, with 10 trades, only increments of 10 percentage points are mathematically observable — a 55% win rate cannot physically exist on a sample of ten.

The operational rule is not “wait for 30 trades.” The rule is: compute the confidence interval and verify whether it resolves against your playbook decision threshold. As long as it spans the threshold, all strategy modifications are deferred — whereas halting trading due to operational risk (risk limit breached, execution error, platform crash) is permitted immediately, without waiting for sample size.

Error 2: Changing Multiple Parameters at Once

The trader audits performance, identifies four perceived shortcomings, and simultaneously modifies four separate parameters. The following quarter, performance changes, but it is mathematically impossible to isolate which modification caused the effect. Unsystematic changes eliminate diagnostic clarity.

Operational rule: maximum of 1 parameter modification per setup per quarter. Progress may feel deliberate, but you maintain rigorous attribution over what drives results.

Error 3: Retrofit — Retrospective Overfitting

A trader inspects the last 10 losing trades, spots a shared attribute (e.g., “all of them occurred on Wednesdays”), and inserts a rule: “never trade on Wednesdays.” An in-sample backtest over historical data shows stellar returns — because the filter was reverse-engineered around the exact data it was designed to eliminate. In live forward trading, the filter provides zero predictive value and often damages performance.

Retrofitting assumes many guises: filters engineered around specific historical losses; selecting currency pairs after inspecting past quarterly gains; indicators utilizing look-ahead information; or running a dozen or more filter combinations and reporting only the single variant that “worked.” The mechanics of data-snooping biases, sample partitioning, and multiple testing corrections are covered in Article 15.4.

An audit extracts a single, non-negotiable principle from backtesting science: never optimize rules on the identical sample where the flaw was detected, and never present that sample as independent validation. A new filter requires an economic rationale rooted in market microstructure or execution friction — “it would have saved money historically” is not a justification; it is merely a description of past noise.

Error 4: Overcorrection After a Single Drawdown

Following a −12% drawdown, a trader slashes position sizing by 70% and introduces a restrictive filter that eliminates 60% of all signals. Total market exposure plummets to 12% of baseline (0.3 × 0.4), compressing both future profits and losses — which the trader misinterprets as “the strategy is now stable.” Sizing is restored to normal three months later, only for the very next large trade to trigger a −10% drawdown. The cycle repeats because overcorrection never addressed the root cause; it merely masked it through capital starvation.

Operational rule: risk interventions must be mathematically proportionate. A −12% drawdown that falls within the expected simulated distribution for your trade vector is not an operational breakdown; it is an ordinary tail event — and comparing an isolated drawdown against an average drawdown is meaningless: individual drawdowns are evaluated against the distribution, not the mean.

Error 5: Auditing from Memory Instead of Journal Data

“I recall that setup XYZ worked brilliantly back in September.” Human memory is notoriously distorted by emotional bias — you vividly remember outsized winners and conveniently forget grinding strings of losses. Without hard numbers logged in a structured trade journal, an audit is a projection of wishful thinking rather than an analysis of data. Consistently: without a reconciled journal, no audit exists — only subjective narrative.

Error 6: Inverted Gate Sequence

Amateur audits invariably begin and end with the question: “Is the strategy making money?” Yet data integrity, execution compliance, and friction costs determine whether that question has any validity — and explain a large part of the problems traders attribute to rules. Operational rule: the sequence of the four gates in Section 3 is invariant — Data and Execution, Friction Costs, Market Regime, and only then Statistical Edge. An audit that starts at the finish line is not faster; it is simply invalid.

Error 7: Modifying Strategy Rules Mid-Quarter

A trader conducts an ad-hoc review in February, changes parameters, experiences worse results in March, reviews again, and tweaks a second time. After six revisions in a single quarter, zero reliable sample data exist — neither for the original strategy nor for any of the modifications. Operational rule: audits adhere strictly to a fixed calendar. Between audits, strategy rules remain locked — you do not modify parameters, even when external commentary suggests a “superior setting.”

Error 8: Confirmation Bias in Dataset Selection

A trader audits performance but selectively carves out “that chaotic week where central bank news wrecked the market” or “those three trades where I technically broke the rules.” Every post-hoc exclusion artificially inflates metrics. Operational rule: an audit encompasses every single transaction executed during the period. You may tag specific trades (e.g., rule_breach = 1), but they remain fully embedded within sample statistics.

Error 9: Ignoring the Psychological Cost of System Churning

Altering a trading system carries an emotional cost alongside statistical risks. Every parameter revision resets psychological trust: over the subsequent 30–60 trades, you are trading an unproven hypothesis rather than an internalized system. A trader who overhauls rules quarterly exists in permanent psychological doubt — which rapidly deteriorates execution discipline. A stable, average strategy executed with iron discipline can outperform a theoretically superior version that you second-guess under pressure. Operational rule: unless a change is justified by ≥ 2 consecutive quarters of signals and clear diagnostic logic, do not change it.

Error 10: In-Sample Threshold Calibration in Real Time

A trader sets the setup demotion threshold at +0.12R after observing that their setup delivered +0.14R — deliberately reverse-engineering the cutoff so the setup survives. Conversely, a trader sets the promotion threshold at +0.45R upon seeing that a Tier-B setup hit +0.47R, granting a promotion based on one favorable quarter. These numbers intentionally differ from the Section 4 thresholds: the example illustrates the mechanism, not those particular cutoffs.

This is real-time in-sample bias — a close relative of overfitting. Operational rule: calibration cutoffs must be established in the playbook prior to the audit session and held immutable throughout. Modifying thresholds requires an independent audit log entry justified by structural shifts (for example, a regime change or a new cost category), never by a desire to spare an underperforming setup.

9. Step-by-Step 90-Minute Strategy Audit Checklist

A structured, time-boxed operational checklist. Executed on the final Sunday of each quarter in a dedicated 90-minute block without interruptions.

Minimum Viable Audit — Execution Without Paid Software

A full professional audit leverages a dedicated journaling database and quantitative analytics scripts. Most retail traders, however, begin with a spreadsheet — and that is entirely sufficient. An actionable audit that protects against emotional panic requires only 6 core columns:

• setup_id — specific setup classification (e.g., “LON Breakout”, “NY Reversal”).
• r_net — realized trade payoff in R-multiples, net of all transaction friction.
• mae_r — maximum adverse excursion in R (how deeply price moved against you before turning).
• mfe_r — maximum favorable excursion in R (how far price moved in your favor before exit).
• rule_breach — boolean flag, 0 or 1 (whether any playbook execution rule was breached).
• playbook_ver — explicit rulebook version active at order entry; without it, you cannot distinguish performance prior to a parameter tweak from performance afterward.

From these six columns, you calculate three essential diagnostic metrics:

1. Expectancy per Setup — the arithmetic mean of r_net grouped by setup_id. In Excel or Google Sheets, use AVERAGEIF conditioned on setup name.
2. Overall rule_breach Percentage — the mean of the rule_breach column multiplied by 100%. In spreadsheets, use standard AVERAGE.
3. Median MAE and MFE per Setup — the 50th percentile of mae_r and mfe_r for each setup. In modern spreadsheets, use MEDIAN(FILTER(...)) or array formulas.

Formula syntax and argument separators vary between language localizations and platforms (e.g., Polish Excel: ŚREDNIA and MEDIANA; English: AVERAGE and MEDIAN; commas or semicolons also depend on spreadsheet locale). The descriptions above outline underlying computational logic; translating to your specific spreadsheet requires seconds in documentation or an AI prompt.

This does not constitute an advanced institutional audit — it lacks bootstrap confidence intervals, Monte Carlo drawdown surfaces, and multi-cluster MAE/MFE scatter plots. However, it is fully sufficient to halt emotional tampering after five consecutive losses: three concrete figures extracted from six columns answer the vast majority of panicky doubts. A professional audit expands upon this baseline once analytical tools are integrated — never the reverse.

Three Audit Tiers — Basic / Standard / Advanced

Operational selection rule: your trade volume determines your audit tier, not your ego. With several dozen trades per year, complex Monte Carlo simulations add minimal actionable value — rigorous tracking of six columns in a spreadsheet provides far superior diagnostic clarity. With hundreds of trades, spreadsheets become unwieldy, necessitating programmatic automation. One critical caveat: a confidence interval for expectancy is not an optional luxury of the Advanced tier; it is a prerequisite for resolving Gate 4. On Basic and Standard tiers, calculate it using the closed-form formula from Section 4 — or leave Gate 4 open in observation mode. Tier selection governs analytical depth, never the rules of strategic decision-making.

Phase 1: Data Preparation & Reconciliation (20 Minutes)

1. Export All Closed Trades for the Quarter — from your trade journal and independently from your broker’s official statement. Core fields: trade ID, date, pair, setup ID, tier, playbook_ver, direction, entry, SL, TP, exit, net R, slippage, rule breach flag, MAE, MFE.
2. Calculate Aggregate Performance Metrics — trade count n, win rate %, net expectancy in R with 90% confidence interval, profit factor, drawdown on the equity curve with sampling frequency specified, longest losing streak, median slippage, and overall rule_breach rate.
3. Segment Performance Metrics per Setup — generate the identical metric summary broken down individually for each setup.
4. Benchmark Against Historical Baseline — compare current metrics against the pooled baseline of the prior four quarters (or backtest benchmarks if conducting a first quarterly audit).

Journaling software is not an edge in itself, but the inability to export clean data destroys an audit. A comparative overview of trading journals and analytical tools is provided in Article 15.9. That comparison covers native calculations, custom formulas, and export formats, with dates because prices and features change faster than the audit procedure. For Phase 1, you require three specific deliverables: a complete export of all closed trades, the broker statement for reconciliation, and recorded MAE/MFE columns.

Phase 2: The Four Audit Gates (30 Minutes)

5. Gate 1: Data Integrity & Execution Discipline. Reconcile the trade journal against the broker ledger (trade count, volume, swaps, commissions). Verify playbook version assignments, check completeness, compute rule_breach percentage, the proportion of stops modified, and premature exits before TP1. Log findings in the audit ledger.
6. Gate 2: Friction Costs. Compare median entry slippage, SL slippage, TP slippage, commission rates, and financing swaps against historical baseline. Convert the cost difference into R.
7. Gate 3: Market Regime. Inspect objective regime metrics calibrated per Section 6 (volatility ratios, ADX persistence). Determine whether market conditions fell within setup card specifications.
8. Gate 4: Statistical Edge Evaluation. Calculate net expectancy with 90% confidence intervals, overall and per setup. Populate the mandatory audit matrix: observed metric, root-cause hypothesis, missing proof, immediate operational decision, and re-evaluation trigger.

Phase 3: Setup Tier Calibration (20 Minutes)

9. Evaluate Each Setup Against Tier Criteria — consult the Section 4 criteria table using confidence intervals. Classify each setup into: Maintenance, Observation (Deferral), Demotion, or Deactivation.
10. Enforce the Two-Quarter Rule — verify that any proposed tier demotion, promotion, or deactivation has satisfied statistical conditions across two consecutive audit cycles under clean execution (rule_breach < 8%) and reconciled friction. Otherwise, enforce Observation mode with documented triggers.
11. Align Position Sizing with Setup Tiers — ensure assigned risk percentages match playbook specifications (15.2); execute capital scale adjustments per Article 15.11.

Phase 4: Operational Decision & Parameter Rules (15 Minutes)

12. Enforce the Maximum 1 Parameter Change Rule — if a rule adjustment is indicated, select strictly one parameter per setup supported by the strongest diagnostic evidence.
13. Document the Rationale for Every Modification — what parameter is modified, exact diagnostic justification from journal data, anticipated outcome, and mandatory observation horizon (until the next quarterly audit).
14. Implement Execution & Behavioral Interventions — if the audit diagnosed an execution breakdown, document structured behavioral safeguards (e.g., reciting playbook rules aloud for 10 minutes pre-session, scaling risk down for 4 weeks via 15.11, establishing peer accountability).

Phase 5: Roadmap for the Next Quarter (5 Minutes)

15. Target Trade Volume — forecast expected trade count for the upcoming quarter based on historical setup frequency.
16. Projected Return Range — establish an expected performance envelope based on baseline expectancy and planned trade volume, expressed as a range rather than a profit target that must be achieved. Converting R to a capital percentage requires explicit sizing. This is not the confidence interval for the historical mean: that interval describes uncertainty about the previous mean, while a future quarter adds its own outcome variability, making its result range wider.
17. Designated Monitoring Indicators — identify specific metrics to monitor during weekly execution reviews (e.g., median SL slippage, percentage of manual early exits).
18. Lock Next Audit Date & Playbook Version — record the exact calendar date of the next quarterly audit and increment the playbook version if rules were modified.

Universal Traps During the 90-Minute Audit

Three friction points where a 90-minute audit commonly breaks down:

• Phase 1 Consumes 45 Minutes Instead of 20 — caused by neglecting daily trade logging, forcing the trader to reconstruct records during the audit. Remedy: log trades daily; an audit aggregates existing data, it never collects it.
• Phase 2 is Performed Superficially — skipping uncomfortable execution gates because confronting trading errors is emotionally unpleasant. Remedy: set a 30-minute timer; require every gate to have a written analytical paragraph in the audit log.
• Phase 4 Explodes into 5+ Parameter Changes — inability to prioritize a single modification. Remedy: enforce the iron rule of “one parameter change per setup per quarter”; log all other ideas in a future research queue.

Appendix: Formal Strategy Audit Log Entry Template

Every quarterly audit generates an official entry in your strategy audit log — stored in a dedicated document separate from your daily trade journal. Below is a completed template entry for Q1 2026:

• Audit Date: 2026-04-26
• Audit Period: Q1 2026 (January 1 – March 31, 2026)
• Active Playbook Version: v2026.1, active since 2026-01-02 (prior version v2025.4 retired; zero Q1 trades executed under old rules).
• Gate 1 (Data Integrity & Execution): Journal reconciled against broker statement — 87 trades on both ledgers; commission charges and financing swaps match. Zero data gaps. Overall rule_breach rate: 6.9% (6 of 87 trades; historical baseline 5.0%; operational tolerance threshold < 8.0%). 11% of trades closed manually prior to TP1 (baseline 6.0%) — elevated, requiring monitoring, but within overall compliance limits. Gate 1 passes; edge metrics in Gate 4 are valid to analyze.
• Gate 2 (Friction Costs): Median entry slippage and stop-loss slippage showed no material change from the prior four-quarter baseline; commission fees and swaps unchanged. Net R calculated using the identical accounting convention. Gate 2 passes.
• Gate 3 (Market Regime): One indicator was slightly outside its band; remaining volatility and directional persistence metrics remained within setup card boundaries. Zero evidence of a structural regime breakdown. Gate 3 passes.
• Gate 4 (Statistical Edge Evaluation) [SYNTHETIC FIGURES FOR ILLUSTRATION]: Total sample n = 87, Win Rate 52%, net expectancy +0.32R, 90% confidence interval (+0.09R; +0.55R), profit factor 1.66. Setup breakdown: LON Breakout: n = 34, E = +0.56R, 90% CI (+0.17R; +0.96R) NY Reversal: n = 22, E = +0.18R, 90% CI (−0.27R; +0.64R) Asia Range: n = 18, E = +0.28R, 90% CI (−0.11R; +0.68R) News Fade: n = 13, E = −0.04R, 90% CI (−0.59R; +0.52R); pooled with prior quarter: n = 36, E = −0.03R, 90% CI (−0.36R; +0.31R).
• Operational Decision 1: LON Breakout, NY Reversal, and Asia Range: observation, without class changes in either direction. Each setup's interval includes its class decision threshold, so Observation overrides formal maintenance, demotion, and promotion. Reassess at the next audit. NY Reversal can be demoted to B only if the table condition (E < +0.15R or the upper interval bound below +0.30R) is met in two consecutive audits, with rule_breach < 8% and reconciled costs in both, and with the Observation rule respected. The current audit does not qualify: E is +0.18R and the upper bound +0.64R, so the counter starts at zero. The first qualifying audit can trigger a scale reduction, not a class demotion.
• Operational Decision 2: News Fade — conditional deferral with capital scaling reduction per Article 15.11. Re-evaluation condition: deactivation if the upper bound of the 90% CI falls below +0.15R in the next audit. (The condition of negative expectancy across two consecutive quarters is already satisfied; the interval remains unresolved; however, the current 15% rule_breach also fails the clean-execution condition, so shutdown cannot be automatic solely on the next interval. The Section 4 prerequisites must all be satisfied). The 15% rule_breach rate on this specific setup is assigned to Decision 3 as an execution defect.
• Operational Decision 3 (Execution Discipline): Focus on strict adherence to codified exit rules established in setup card v2026.1 — no new rules added; playbook text and version number remain unchanged. Action: daily execution review for four weeks. Q2 objective: reduce exits before TP1 below the reference rate. If the wording of the exit rule itself were changed, that would be a playbook amendment requiring a new version, metric, threshold, and verification date like any other change.
• Rule Modifications: Zero. Playbook version remains locked at v2026.1.
• Q2 2026 Operational Roadmap: 70–90 planned trades. Primary diagnostic criteria for next audit: confidence interval widths across all four setups, median slippage stability, and rule_breach rate. Off-calendar audit trigger: equity drawdown threshold defined in trading plan.

FAQ — Frequently Asked Questions