Position Management, Review, and Rule Validation

29.1 Price and Time Expectations After Entry

After entry, the account bears losses from adverse price changes, slippage, and costs. At each management point, ask whether continuing the remaining exposure from now, given current information, is preferable to reducing or exiting. Past fills and incurred costs belong in total performance, but having already paid them does not justify further risk. Asking whether one would establish equivalent exposure at current prices, boundaries, and costs if flat can test the holding rationale, though existing holdings also involve exit costs, taxes, and limits on instantaneous execution. Before holding, specify expected price development, its location, how long to wait, and behavior indicating a weakening view.

Price expectations describe the sequence of structural development. A long on a Spring test may expect selling near the test low to produce less decline, followed by a break of the local high and progress into the range. A breakout-retest long may expect price to hold near former resistance, avoid renewed acceptance inside the old range, and resume net advance after the retest. Not every candle must point the same way, but define permissible pullback areas, required confirmation points, the first obstacle, and targets. The observation scale must match the entry model: minute noise cannot disprove daily structure, and weekly directional judgment cannot replace intraday trade invalidation.

Time expectations specify when these developments should occur, measured in completed bars, sessions, swings, or event windows rather than anxiety. Post-breakout imbalance should usually continue within a reasonable period. Repeated lingering near the boundary and renewed balance weaken the expectation of rapid continuation. A range reversal may require several internal swings and cannot use the same deadline. Base time rules on historical model observations and fix them beforehand, for example: if price has not exceeded the confirmation swing within N completed bars of this scale after triggering, reassess or exit. N is a parameter to validate, not a universal constant.

Assess price and time together. Repeated failures to exceed the planned confirmation point may trigger a predefined cumulative time-exit rule even before the stop. Clear opposing structure may justify discretionary exit early, without waiting for the deadline. Conversely, a position lacking open profit but still within its permitted area, with sellers unable to push farther down, does not justify changing the plan merely because waiting is uncomfortable.

29.2 Meeting Expectations, Underperforming, and Contrary Evidence

Meeting expectations means the predefined key behaviors appear in sequence: support or resistance works, price progresses in the trade direction, pullbacks remain controlled, and confirmation or target is approached within the allotted time. Continue holding, reduce at original targets, or move protection favorably using newly formed structure. Open profit alone does not mean expectations are met; a random jump immediately retraced may establish none of the expected acceptance.

Underperformance means insufficient or deteriorating evidence without decisive contradiction: limited breakout progress, repeated returns to the boundary, increasing effort without matching result, or failure to pass the confirmation swing in time. Responses may include stopping additions, reducing size, tightening to a newly valid boundary, or a time exit. Predefine actions and thresholds so every slow move cannot become a reason to wait longer.

Contrary evidence is an observable fact directly conflicting with the trade rationale. A Spring-test long breaking the shock low and gaining acceptance below, an LPS failure with expanding supply, or an expected valid breakout becoming a failed breakout back inside the range is stronger than an ordinary pullback. A discretionary close may be appropriate; if protection triggers first, actual execution ends exposure. Contrary evidence need not prove final reversal of the larger structure; it only needs to remove the current trade's conditions.

Assess the three cases at the same decision time, with defined actions for each. Reassess after a completed bar, key trade, or event-window end rather than changing plans on every tick. If several interpretations remain, let risk constrain action—no additions, original protection, or reduction—instead of selecting only the explanation favoring continued holding.

29.3 Changes in Holdings, Cost, and Risk After Scaling In or Out

Every fill changes the net position. Let the signed quantity of fill i be Δqᵢ, positive for buys and negative for sells. After fill k, net position Qₖ＝Σ(i＝1 to k)Δqᵢ.

After same-direction additions, calculate average holding cost from the traded amounts and quantities of lots still open, separately specifying whether commissions and other fees are included. Reductions realize some profit/loss but generally should not arbitrarily redefine the remaining economic cost as zero or negative. Keep realized trading-performance profit/loss separate from remaining book cost. If a long is reduced to zero and then reversed short, separate closing the long from opening the short as economic events; do not mix average prices across directions.

FIFO, LIFO, and moving-average costing recognize partial profit/loss as realized or unrealized at different times. With identical fills, fee treatment, and currency conversion, and ignoring tax differences, total net profit/loss after full closure should agree; costing mainly changes the interim split. Records should include fill time, side, quantity, price, fees, net position, remaining average price, realized profit/loss, and unrealized profit/loss marked to executable quotes.

A better new average price cannot replace total risk assessment after an addition. Place every open quantity at its valid protective price or a common stress exit and calculate account loss from realized profit/loss, unrealized profit/loss, expected exit fees, and slippage if that scenario occurred immediately. For same-currency linear positions, calculate each lot as signed quantity times exit-minus-entry price times multiplier; use contractual payoff functions for nonlinear or cross-currency products. Compare stress-scenario net loss measured from trade inception with the fixed trade budget B. Only when calculating additional future drawdown from current net liquidation equity should it be compared with current remaining risk capacity. Do not adjust remaining capacity for realized profit/loss and then include that same realized result again in total trade profit/loss. Both approaches also remain subject to portfolio limits.

Realizing profit through reduction lowers the remaining market exposure but does not retrospectively reduce the initial risk benchmark. Add only when new evidence qualifies, protection is clear, and the maximum planned loss after addition is acceptable. Averaging down an unrealized loss adds same-direction exposure as evidence deteriorates and is not a valid reason to add. When portions use different stops, ensure their total protected quantities match the actual net position and update after partial fills to prevent accidental over-reversal.

Futures accounting cannot treat margin cash flows plus q times notional price as account equity. Futures generate daily or real-time profit/loss under contract specifications while consuming margin and costs. Notional value describes represented exposure; margin secures performance. Neither is the market value of an owned asset. Compute equity under broker or clearing rules from cash balances, applicable unsettled profit/loss, and undeducted costs. Do not add settled profit/loss again if already included in cash.

29.4 Structural Exits, Time Exits, and Moving Stops

A structural exit occurs when price breaks the predefined local trade structure. It may execute automatically via a stop or manually after a completed bar confirms a breach and acceptance. The first reacts quickly but is more vulnerable to wicks and ticks; the second waits for information but may increase loss. Predefine intraday touch, closing confirmation, or sustained presence and the exit order type; do not choose whichever interpretation best suits the result after execution.

A time exit addresses expectations not fulfilled within the valid window: a breakout that should continue but stalls, an expired event-driven window, or a short-term model entering a different-liquidity session. At expiry, fully exit, reduce, or require the next price confirmation according to a fixed rule. It is not an arbitrary sell after several days without gains rule; deadline, scale, and exceptions must be reconstructable in historical samples.

Moving stops adjust protection upward or downward with favorable development. Bases include confirmed swing lows/highs, structural retest boundaries, volatility scales, or fixed trailing distance. Structural movement updates only after a new swing completes, avoiding unconfirmed lows as support. Volatility-based movement requires a fixed indicator, window, and calculation time; platform trailing stops require verification of reference price and update mechanics. Protection normally must not then widen toward greater risk, or the original risk ceiling loses force.

All three exits can coexist if priorities are specified. A rapid protective touch executes the stop first; intact structure at a time deadline follows the time rule; subsequent trend progress uses a moving stop to protect the remainder. Targets, stops, and manual reductions must track actual remaining quantity so a fill on one side does not leave the other at its old size.

Structural exits follow invalidation of the trade rationale; time exits follow missing expected behavior within the permitted period; moving stops change the protective trigger as new structure forms. This course permits protection to move only toward reduced risk, with recalculation for remaining quantity. That is an explicit rule choice, not a claim that all methods must do so. Stops behind a newly formed but unconfirmed tiny turn are easily swept by normal movement. Reaching a profit target need not mean full exit if staged exits are planned, but quantities and remaining protection must be specified. Clear contrary evidence before target also should not be ignored while waiting for theory. All exits depend on actual fills; trigger price is neither execution price nor a guaranteed worst fill.

29.5 Improvised Plan Changes Versus Evidence-Based Rule Revisions

Positions may respond to new information, but changes require predefined grounds. Valid revisions include moving a stop to a nearer valid boundary after new structure forms, reducing remaining quantity after a target fill under original rules, applying contingencies when venue status changes or liquidity materially worsens, and ceasing reliance on a signal after a confirmed source error. Each identifies a new fact, applicable rule, calculation, and effective time.

Improvised changes usually lack new verifiable evidence and move the standard when price becomes adverse: lowering a long stop after a decline, chasing after missing the trigger, canceling a near profit target in pursuit of a farther one, or relabeling a short-term trade as a long-term holding. This course prohibits widening protection on the current position. Widening without reducing quantity increases expected stop loss under normal execution; gaps, suspensions, price limits, or unfilled stops can make actual loss larger still. A later larger structure may inform the next trade or rule version, not loosen the current stop.

Retain the old value, new value, time, triggering evidence, risk change, and responsible operator for every modification. Permission to adjust without defined data, thresholds, and actions leaves excessive room for hindsight. New live-trading ideas may be recorded for research, but should not immediately be tested on current holdings with favorable outcomes credited to the system.

Ambiguous rules discovered while holding can be revised for the next trade or version. A temporary boundary change on the current trade must arise from new facts allowed before entry. The distinction is whether evidence is independently observable then and the change can be applied consistently in favorable and unfavorable cases. Expanding risk near a stop or raising a target as price approaches it usually changes rules under outcome pressure. If sudden changes in market arrangements, contract specifications, or data sources truly make the original order infeasible, minimize risk and record an exception rather than assign the result to the original strategy.

29.6 Execution Deviations and Their Technical Consequences

Chasing occurs when an order is submitted after the planned trigger has passed and actual entry lies outside the permitted zone. Entry-to-protection distance increases, target room shrinks, and original size and reward-to-risk calculations become invalid. Recalculate at current executable prices; only if conditions still qualify create a new plan, otherwise await a retest or new structure instead of retaining old quantity.

Widening a stop moves the boundary adversely as price approaches original protection. This course forbids it for current positions because it increases unit planned risk; unchanged quantity raises whole-trade planned loss and may exceed portfolio stress limits. Gaps, suspensions, price limits, or unfilled stops may still exceed the planned amount. A newly useful larger structure can inform only a future opportunity. Avoiding recognition of a loss is not valid evidence. Other methods permitting predetermined widening with simultaneous reduction must be described as independent rules, not mixed with this course's rules.

Revenge trading rapidly increases frequency, reverses direction, or increases size after a loss without an independent opportunity or in breach of risk intervals. Order records reveal absent triggers, unusually short intervals, excessive quantities, or repeated reversals within the same noise. Rules can require an independent plan ID, a fresh context–location–trigger check after losses, and a halt to new risk when daily loss or order-count limits are reached.

Overtrading means orders and exposure exceed the model's qualifying opportunities, not merely that trade count is high. A high-frequency model may generate many compliant orders, while repeated midrange entries in a low-frequency structural model may be excessive. Useful fields include concurrent positions, accumulated budget by risk source, orders per unit time, proportion of fills without triggers, costs as a share of gross profit, and cancellation/modification counts. Psychological labels do not replace technical diagnosis. Identify the violated order rule, added risk, and needed system constraint.

Execution deviations include late entries, chasing, wrong quantity, missing protection, excessive stop slippage, and unsynchronized partial quantities. Each changes actual risk and requires more than a note saying the plan was not fully followed. Chasing with the original stop increases distance; missing protection removes the planned exit mechanism; early exit may reduce profit and delayed exit may increase loss. Compare possible original-plan profit/loss with actual-fill results to estimate execution drag, while acknowledging that original orders also might not have filled ideally.

29.7 Four-Layer Review: Observation, Interpretation, Decision, and Execution

Observation records only facts available then: price, volume, boundaries, market state, quotes, orders, and fills. Do not backfill later highs/lows or treat an inference such as operator bull trap as fact. Interpretation explains why facts support a Spring, LPSY, valid breakout, or another hypothesis, listing competing explanations and counterevidence. Accurate observation with wrong interpretation differs from an error in raw data.

Decision review asks whether trading, waiting, canceling, or exiting followed the rules given information available then. A later loss does not automatically make the decision wrong; a fully qualified trade with acceptable risk can fail normally. A profitable unauthorized chase remains a decision deviation. Execution review checks submission time, type, size, fill, slippage, partial fills, stop linkage, and manual changes, separating strategy edge from execution drag.

Separate event success, structural judgment, trading profit/loss, and execution deviation. A genuine Spring may still produce a noise-stopped local entry; an incorrect structural view may profit from an unexpected gap; a loss may arise entirely from excess slippage; a rule violation may temporarily profit. Code these outcomes separately to determine whether to change the model, risk parameters, order process, or nothing.

Review using original plan versions, contemporaneous screenshots or data copies, order and fill logs, cost statements, and every change in position and profit/loss while holding. If contemporaneous information is missing, mark the judgment unassessable rather than call a conclusion obvious from a complete historical chart.

29.8 Initial Risk Amount R₀ and After-Cost R

R₀ is the initial planned risk amount fixed when the trade is established, including planned-entry-to-initial-protective-exit price loss, expected round-trip costs, and reasonable slippage, converted to a common account currency. R₀ must be positive. If zero, negative, or missing, do not calculate R; check prices, quantity, multiplier, costs, and conversion. If staging is part of the original plan, define R₀ as its maximum permitted initial-plan loss. If independent risk is added later, retain both original R₀ and incremental risk of the addition rather than revise the benchmark afterward to improve performance.

Profit and Loss (PnL) denotes a trade's gain or loss. After full closure, after-cost R is: R_final＝NetPnL_final÷R₀.

Net PnL includes realized results of every portion, less actual commissions, transaction fees, borrowing or funding costs, and applicable exchange-rate changes. R₀ normalizes trades of different sizes; it does not guarantee actual loss equals 1R and should not shrink after profit-taking or a stop move. Replacing a winner's denominator afterward with remaining risk systematically exaggerates R multiples.

Distinguish planned R, interim R, and final R. Planned reward to risk uses expected entry, stop, target, and costs; interim net PnL divided by fixed R₀ creates a real-time R curve; final R is known only after the whole trade cycle ends. Cross-market comparisons need consistent cost treatment, exchange-rate sources and conversion times, fill timing, and trade-cycle definitions, or identical R labels are not comparable.

Initial Risk Amount and After-Cost R

R0 must be positive and remains unchanged after reducing size or raising protection.

Figure 29-1 Initial Risk Amount and After-Cost R (from Section 29.8 of this course)

29.9 MAE, MFE, and the Entire Trade's Continuous Net PnL Curve

This course calculates Maximum Adverse Excursion (MAE) and Maximum Favorable Excursion (MFE) from the whole trade's time series of net liquidation PnL, beginning at the zero-PnL baseline before entry. Let N(t) be cumulative net PnL through time t if all remaining positions were closed at executable quotes then. MAE＝max[0, −minₜN(t)], and MFE＝max[0, maxₜN(t)]. Platforms may instead define these as the distance from entry price to the worst or best price during holding. Confirm formulas before comparing; do not mix the definitions.

Calculate N(t)＝realized gross PnL＋gross PnL from liquidating remaining positions at executable quotes−paid costs−estimated exit costs＋signed net holding cash flows not already recorded as costs. Do not include financing, borrow, or funding fees again if already deducted. Calculate components gross before deducting costs consistently. If actual fills or Bid/Ask valuation already include spread, do not deduct it twice. Value longs at sellable Bid and shorts at coverable Ask, or other market-defined executable prices. Use actual results for settled and converted PnL; only balances still held in foreign currency use the chosen source's rate at t. Initial opening fees naturally put the net PnL series slightly below zero immediately.

With additions and reductions, reconstruct holdings and net PnL at each time in fill order. The full-period high or low minus a single average entry price cannot represent the whole dynamic trade because quantities differ, some results are realized, and costs occur at different times. Specify data frequency. OHLC does not reveal whether high or low came first within a bar, so do not invent sequence. Use finer data or report a conservative range under both ordering assumptions.

Divide net MAE and MFE by fixed R₀ to obtain MAE_R＝MAE÷R₀ and MFE_R＝MFE÷R₀. If research also reports maximum adverse price points or percentages after entry, specify units and calculation object, keeping them separate from net R that includes costs, staging, and exchange rates. MAE studies stops and normal movement; MFE studies profit giveback and targets. Neither alone proves a new parameter superior.

Figure 29-2 Whole-Trade Net PnL Curve and MAE/MFE (schematic)

29.10 Operational Definitions of Trading Rules

A testable rule includes timing, data, scale, trigger, cancellation, action, and costs. For buy the retest after a breakout, specify market and session, how range boundaries were defined then, what crossing counts as breakout, permissible retest depth, intraday or closing confirmation, cancellation deadline, entry order, and slippage/cost treatment. Missing any key item allows interpretation to change across samples.

High volume, fast, holds, and strong must become calculable, consistently assessable conditions. Volume can reference a fixed past window, the same session, or the prior same-direction swing. Fast can use bar count or duration at the chosen scale. Holding can mean no close back inside, or recovery within a specified number of bars after returning. Parameters need not pretend universal correctness, but must be fixed, saved, and tested.

Subjective structural rules can also be operationalized. Two analysts may independently apply the same annotation manual, recording agreement and reasons for disputes. The manual should define swing confirmation, boundary adjustments, nonclassic structures, and data gaps. Rules must also specify deduplication of repeated signals, risk allocation among simultaneous holdings, simulation of unfilled/partial orders, and whether missing data cause skipping or conservative substitution.

Costs are part of the rule. Use executable-side quotes or a consistent slippage model at entry and exit, add commissions, taxes, funding, and borrow costs, and explain variation with quantity, volatility, and liquidity. Only a complete operational definition lets historical testing, live execution, and review discuss the same object.

29.11 Samples, Future Information, Out-of-Sample Testing, and Costs

Define the market universe, period, liquidity conditions, and inclusion rules before selecting samples. Keeping only currently listed or subsequently successful instruments creates survivorship bias; selecting only visually textbook Springs omits ambiguous candidates inevitably encountered live. Handle delistings, contract rolls, suspensions, corporate actions, and missing data appropriately to the study and retain records.

Look-ahead bias uses information unavailable at the time. Confirming swings with future extremes, filling at the close with an indicator available only afterward, and backfilling revised data overstate performance. If one OHLC bar covers both stop and target, their order is unknown without finer data. Use conservative rules, explicit fill assumptions, or finer data rather than always choose the profitable sequence.

Training data propose and estimate rules, validation data select them, and a final out-of-sample period assesses unseen data. Do not arbitrarily shuffle time series. Walk-forward testing estimates on an earlier window, tests on the immediately following window, and rolls forward using only information then available. Frequent parameter reestimation increases complexity and overfitting risk and must include actual repositioning and execution costs.

Trying many indicators, windows, stops, and targets but reporting only the best combination creates multiple-testing bias. Save every attempt, limit research flexibility, assess stability of simple neighboring parameters, and use data excluded from selection. Overlapping signals, repeated entries in one trend, and correlated instruments make samples nonindependent. Effective information is smaller than trade count, so ordinary independent-sample formulas must not exaggerate certainty. Assess by event, time block, or shared risk source.

Out-of-sample deterioration should not immediately trigger parameters chasing recent markets. First check changes in data, costs, market structure, execution implementation, and exposure composition, then distinguish randomness, fading edge, or rule failure. If deterioration exceeds predefined tolerance, reduce risk or pause new trades and research a new version without contaminating the original out-of-sample record.

29.12 Win Rate, Profit/Loss Distribution, Expectancy, and Drawdown

Win rate is profitable trades divided by all completed trades, with explicit rules for zero results, partly open trades, and multiple fills under one plan. Average gains and losses should be after costs, with medians, quantiles, and tails because a few extreme wins can lift the mean. Normalize each trade by its own fixed initial R₀; sample mean R is an empirical estimate of expectancy per completed trade: E[R]＝p_win×average winning R＋p_loss×average losing R＋p_breakeven×0. All three proportions use all completed trades as denominator, and average loss retains its negative sign. Every completed trade belongs to one category, so weight by actual frequencies in the full sample. Removing zero-PnL trades or merely taking within-category means changes frequencies and no longer gives the sample average per completed trade. If a category is absent, mark its mean not applicable and its contribution directly 0, avoiding multiplication of zero by an undefined mean. Summing all final R values and dividing by trade count gives the same empirical mean with consistent classification. Its ability to represent the future remains constrained by selection, dependence, and market changes in 29.11.

High win rate does not imply positive expectancy: occasional huge losses can overwhelm many wins. Low-win-rate methods may rely on a profitable right tail but suffer long losing streaks. Jointly report trade count, win rate, mean and median profit/loss, quantiles, expectancy, cost share, holding time, MAE/MFE, and results by market state. Staged or overlapping trades also require simultaneous capital usage and portfolio-level PnL.

Drawdown uses the chronological account-equity curve. Let equity at t be Eₜ and prior peak Hₜ＝max(Eₛ, s≤t). Monetary drawdown Dₜ＝Hₜ−Eₜ; percentage drawdown is defined as Dₜ÷Hₜ only when Hₜ＞0. Maximum drawdown is the largest observed in the sample, not a future loss ceiling. Duration, recovery time, and worst rolling windows matter alongside magnitude. Use unitized equity or a predetermined method removing deposits and withdrawals; withdrawals are not trading losses and deposits are not strategy returns. Futures equity must follow applicable cash, unsettled PnL, and undeducted-cost rules. Do not add settled PnL already in cash again, or treat margin and notional contract value as assets.

Finite-sample expectancy and drawdown have substantial uncertainty. Report mean confidence intervals, distributions resampled by time block or event cluster, and results under cost/slippage stress; overlapping correlated samples are not independent trades. Also assess neighboring parameter stability, consistency across years and states, and whether out-of-sample results retain the same direction. Highest return, highest win rate, or lowest drawdown alone cannot select a rule.

29.13 Rule Changes and Version Control

Each rule set needs a unique version, adoption date, applicable markets, data and time zone, parameters, order assumptions, cost model, and risk limits. Retain originals; state each new version's exact changes, supporting sample, expected mechanism, and effective date. Evaluate historical trades under the version used then, not new rules retroactively applied.

Changes may be triggered by persistent out-of-sample deterioration, execution-environment changes, contract or fee changes, clear data errors, or recurring operational ambiguity. Identify whether the problem concerns signal, exit, size, costs, or execution, then make the smallest explainable change. Changing entry, stop, target, and filters simultaneously makes improvement difficult to attribute.

Candidate versions require new training, validation, and forward samples, recording all research attempts. Once old out-of-sample data select new parameters, they become training information and cannot again count as unseen testing. Deploy a new version first at smaller risk or in shadow execution to compare expected and actual fills. Expand only after predefined criteria under the risk framework, not a few profitable trades.

Version archives should retain code or formulas, configuration, data-snapshot identifiers, random seeds, generated results, review comments, and live deviations. If out-of-sample results worsen, compare the current version with predefined tolerance; set pause, rollback, and observation conditions beforehand. Version control does not freeze rules. It ties each change to contemporary evidence so research returns, live returns, and execution deviations can be honestly reconciled.

Four Requirements for Rule Changes and Version Control

Invalid changes include widening stops after losses, removing confirmation after missing a move, and repeatedly adjusting parameters to improve the historical curve.

Figure 29-4 Four Requirements for Rule Changes and Version Control (from Sections 29.5 and 29.13 of this course)