The lab's honest boundary: what Monte Carlo simulates — and what it can't.

What goes in

Assumptions, not history — and the lab is faithful to a fault.

Feed the lab your plan's assumptions — branch weights, hit rates, exit structure, risk pool, cost model — and it will simulate thousands of full trading years under exactly those terms. That faithfulness is the point and the trap at once. If you feed it an optimistic edge, it will faithfully simulate an optimistic account. Nothing inside the engine questions whether your assumed hit rate is real; that verification belongs to the live-evidence stack — the journal, the Weekly EV Scorecard, the Advanced EV Analytics Lab. The MC Lab's job begins where theirs ends: given this edge, what does a year of it look like across every ordering fate could deal?

What it actually answers

Sequence risk: the same edge, dealt in every order.

A positive-expectancy system can still destroy an account if the losses arrive bunched at the start. That is sequence risk, and it is invisible in a spreadsheet average — EV per trade says nothing about the order of outcomes. The lab's core service is dealing your exact assumed edge in tens of thousands of different orderings and reporting the distribution of fates: how many paths reach target, how many hit System Lock, how deep the drawdowns run, how long recovery takes. It converts 'my system has positive EV' into 'my system, sized this way, survives X% of its own possible years.' Those are radically different statements, and only the second one deserves capital.

What it cannot do

Three jobs the lab will never perform.

First, it cannot discover an edge. No amount of simulation turns a losing assumption set into a winning one — it can only show you how the losing happens. Second, it cannot predict your next month. The output is a distribution, not a forecast; the median path is the centerline of possibility, not an appointment. Third, it cannot grant permission. A beautiful simulation does not override the gate ladder, expand a tier cap, or excuse an override — capital-state authority always outranks simulated confidence. Operators who forget any of these three turn a survival instrument into a rationalization machine.

• Edge discovery lives in live evidence: journal → scorecard → EV Lab. The MC Lab consumes their verdicts; it never issues them.
• The distribution is honest precisely because it refuses to pick your future for you.
• Simulated confidence is interpretation-layer material. Gate/Brake state remains the highest capital authority.

The fragility service

Its real value: exposing where a 'good' system quietly becomes a lockout machine.

The commercially unglamorous truth is that the lab's highest-value output is bad news delivered early. A thin edge that looks tradeable in a backtest average can carry a lock probability that makes it uninvestable at the intended size — and the lab is the only instrument in the stack that will say so before live capital finds out. Where a fee-heavy cost model, an aggressive risk band, or an over-weighted branch quietly turns positive expectancy into probable ruin, the simulation surfaces it as a number: lock rate, adverse-percentile drawdown, survival probability. Fragility you can see is fragility you can size around.

The key idea

The lab is a ruler for futures, not a window into one.

Everything the simulation layer feeds — the Dynamic 7-Tier Benchmark, the percentile bands, the live-comparison workflow — inherits this boundary. When live results are later read against simulated bands, the comparison is only meaningful because both sides agree on what was assumed. Keep the boundary and the whole simulation stack stays honest: assumptions in, distribution out, verification elsewhere, authority untouched. Blur it, and every percentile chart on the rail becomes decoration.