Operator brief · 175

Six ways to read the bands correctly and get the wrong answer.

The key idea

The cap

High percentiles flatten near target, and the flattening is not a ceiling on skill.

Simulated paths stop at completion — a path reaching the target is done, and its equity is recorded there. The consequence for the bands is structural: the upper percentiles compress as the horizon advances, because an increasing share of the paths above them have already finished and stopped growing. An operator reading the late-year bands without knowing this sees the P95 line bending toward the P90 and infers that exceptional outcomes converge, or that the model penalises success. Neither is happening. The distribution is being censored at the top by its own completion rule, and the standard lists understanding this as a precondition for using the table at all.

FigureTarget-capped equity — why the upper bands converge late in the horizon

Schematic geometry, not a workbook reprint: the upper percentiles bend toward the completion threshold as paths finish and stop. The convergence is the cap, not a statement about how good outcomes behave.

The dead paths

Locked paths are terminal residents, not slow performers.

The checklist requires confirming that locked paths are interpreted correctly and not mistaken for paths still trading. The distinction changes the meaning of the low bands entirely. A locked path is an account that breached the terminal drawdown boundary, stopped permanently, and had its equity frozen where it fell — it is not a path having a poor year and it will never recover. Because those frozen outcomes stay resident in every distribution, the adverse percentiles carry the full weight of the futures that died. Read the low bands as the experience of struggling-but-live accounts and they look survivable; read them correctly and a portion of that territory is not a struggle but an ending.

The sign and the label

Two clerical conditions that silently invert a verdict.

Drawdown must be expressed consistently as a negative decimal or percentage, and live period labels must match the benchmark's periods. Both sound like formatting. Both are capable of producing a confident wrong answer with no visible symptom: a sign inconsistency can place an adverse drawdown on the favourable side of a band, and a misaligned period can compare a five-week live month to a calendar-month benchmark and attribute the difference to performance. These are the failures that survive review, because nothing about the output looks unusual. The checklist catches them at the only point they are cheap to catch, which is before the comparison rather than after the conclusion.

The two comparison conditions

Tier usage reads against its cap; open exposure means downside risk.

Two further items constrain what specific figures are compared to. Tier usage must be judged against the gate cap rather than against raw frequency — a run of high-tier deployment is either earned acceleration or aggression drift depending entirely on what the gate authorised at the time, and the frequency alone cannot tell the two apart. And open exposure must be measured as active downside risk, never as floating profit. An account carrying several open positions in profit has capacity consumed by their risk, not released by their unrealised gain, and reading the figure the other way makes a compressed deployment week look like an underused one.

The currency condition

The production record closes before the comparison opens.

The remaining item is the one most likely to be honoured in principle and skipped in practice: live data for the period must be complete before it is compared. The comparison consumes the production stack's outputs as its inputs, so a half-reconciled journal produces a placement built on numbers that will change. This condition is covered in depth by the comparison method itself; the reason it appears in the pre-read checklist too is that it fails differently here. A stale journal does not block the comparison. It produces a percentile, the percentile enters a review, and the correction — arriving days later — has no mechanism to reach the conclusion it invalidated.

The key idea

The instrument has preconditions, and none of them are about performance.

Six checks, and not one asks whether the account did well. They ask whether the bands mean what the reader assumes, whether the live figures are comparable to them, and whether the record is settled enough to be compared at all. That is the shape of every serious measurement discipline — the interesting question is gated behind dull conditions, and skipping the conditions does not prevent an answer, it just detaches the answer from the thing it appears to describe. Two questions govern the benchmark. Six conditions govern whether asking them means anything.

Connected inside MARS

The Two Questions →

Equity Percentiles →

Compliance Panel 3 →

Every brief documents the same shipped system.

The complete MARS package — eleven workbooks, three TradingView indicators, the full manual library — $497.

Get MARS Pro — $497 Explore the Platform