Dynamic 7-Tier Benchmark · The Doctrine
A ruler, and the seven readings on it.
Simulated paths
50,000
Observed hit rate
87.4%
Observed lock rate
12.0%
Median hit week
17
Read this first
The calibration anchor
01Two runs, held side by side, so drift in the ruler is visible.
The workbook preserves both the original benchmark anchor and the current observed run: hit 87.6% against 87.406%, lock 12.4% against 12.018%, hit-week quartiles of 14/17/21 against 15/17/20. Keeping both on the same sheet is deliberate. A benchmark that silently re-anchors itself each time the model changes cannot detect that the model changed, and the operator loses the ability to say whether this month's comparison used the same ruler as last quarter's.
The hit-week trap
02Median hit week is a property of the survivors.
Hit-week percentiles are calculated only among paths that reached the target. Failed paths and locked paths are excluded from that statistic entirely. A fast median hit week therefore says nothing about safety on its own, and it is never a deadline. Read hit speed beside completion probability and lock probability or the number will flatter the system by construction, because it is the average finishing time of the runners who finished.
Anti-misuse
03The rules exist because each one has a failure mode attached.
The standard writes its prohibitions out rather than leaving them to judgement, because every one of them describes a way an operator has talked themselves into risk using a benchmark as the excuse.
- Do not use the benchmark to justify revenge trading or catch-up trading.
- Do not compare a partial week to a full-month band without labelling it provisional.
- Do not judge the system by equity alone: equity, drawdown, gate dwell, tier usage, exposure, and throttle efficiency are read together.
- Do not override the throttle because the live account is below benchmark. Underperformance is a reason to diagnose, not to force risk.
- Do not modify benchmark assumptions without versioning the workbook and updating the standard.
The strong rule
04Below benchmark and worse drawdown is never an argument for more risk.
The standard states this as its one unhedged rule, and it is the position most likely to be argued with in the moment. When live performance sits under the bands and drawdown sits outside them, the two readings together describe a system that is both underperforming and over-risked. The answer is diagnosis, containment, and a model-to-live mismatch review. Adding exposure to a system in that state converts a recoverable quarter into a locked account.
How MARS uses this
A ruler needs marked gradations, and this table is the marking. Each percentile row carries a defined operator read — the doctrine that live performance is compared against bands, not wished against targets, starts by giving every band a name and a permitted meaning.
| Band | 13 wk | 26 wk | 52 wk | Operator read |
|---|---|---|---|---|
| P90 | +64.2R | +131.8R | +268.4R | hot path — do not extrapolate |
| P75 | +47.5R | +96.3R | +197.0R | strong but ordinary |
| P50 | +31.1R | +63.9R | +130.6R | the median story |
| P25 | +16.8R | +35.4R | +74.2R | slow — still inside the model |
| P10 | +4.3R | +11.7R | +27.9R | the path sizing must survive |
Cumulative R by horizon · 50,000 resampled paths
How it benefits you
Percentiles become vocabulary instead of decoration. When you say 'we are tracking P60', everyone - including future you - knows exactly what that means, and sizing debates end by pointing at the P10 row rather than negotiating with optimism.
The ruler, printed: each band with its horizon values and the one interpretation it is allowed to carry.
Reference
Percentile position and the response it authorises
| Live position vs band | Meaning | Operator response |
|---|---|---|
| Above P90 | Exceptional outperformance | Confirm it is not excess risk, rule-breaking, or override behaviour |
| P75 to P90 | Strong outperformance | Continue if drawdown and gate behaviour are normal; document what is working |
| P50 to P75 | Above the centerline | Healthy. No need to force more aggression |
| P25 to P50 | Below median, still normal | Monitor. Check execution, fees, branch mix, and sample size |
| P10 to P25 | Weak relative performance | Review branch quality, EV, drawdown, fee drag, throttle usage, SDE context |
| P5 to P10 | Danger-zone underperformance | Formal diagnostic review; consider reduced aggression until cause is found |
| Below P5 | Outside normal benchmark behaviour | Audit execution, risk usage, gate compliance, and model assumptions |
Connected inside MARS
This module doesn't work alone.
Go deeper
Operator briefs on this territory.
Deep dive — 01
Why the benchmark moves with your tier path.
Static simulations model a trader the system forbids. The dynamic benchmark models the governed one.
Read the full brief →
Deep dive — 02
The same number, read two ways — and only one of them is legal.
Eight prohibitions, all of them describing a moment when a ruler becomes a target.
Read the full brief →
Deep dive — 03
A benchmark measuring a system you no longer run.
Monthly bands, quarterly survival, event-driven rebuilds — and a regeneration log saying which ruler graded which weeks.
Read the full brief →
Every module ships in the complete MARS package.
One price. Eleven workbooks, three TradingView indicators, and the full manual library — $497.