The risk deployment gradient: capital flows toward proven quality.

The gradient

Seven tiers, one principle: evidence rents authority.

Each tier authorizes a per-trade risk level, and tier assignment is not a mood — it is a function of account state, gate position, and the setup's demonstrated quality. A branch with two hundred trades of positive net expectancy earns access to higher deployment than a variant still in R&D. The gradient converts the boring accounting of evidence into the live sizing decision, which is precisely where discretionary traders leak the most edge. Equal sizing is not neutrality; it is an active decision to fund a weak branch at the expense of a strong one.

Deployment as a skill

You can pick well and still deploy badly.

MARS separates selection quality from deployment quality because they fail independently. The Risk Deployment Quality Score asks: did the biggest risk go to the best-evidenced setups, or did conviction and sizing drift apart? A trader can run a positive-expectancy book and still starve their best branch while overfeeding a mediocre one — profitable, but leaking alpha invisibly. The gradient makes that leak measurable. Two traders with identical selection ability and different deployment discipline will not produce similar curves.

How a tier is earned

Evidence rents authority, and the lease is reviewed.

Tier access is not granted permanently. A branch reaches a higher deployment tier by accumulating trades at positive net expectancy with acceptable stability, and it holds that tier only while both conditions persist. Structural drift, a widening stability band, or a run of conditional weakness in a particular regime can send it back down without any single catastrophic event. This is what makes the gradient a live mechanism rather than a classification exercise: the question is never whether a setup was good, it is whether the evidence currently supports the size being deployed behind it.

• Sample depth and net expectancy set the ceiling.
• Stability, not just average, decides whether the tier holds.
• Demotion is routine maintenance, not a verdict on the operator.

The failure the gradient prevents

Conviction and evidence drift apart quietly, and always in the same direction.

Left ungoverned, deployment migrates toward whatever feels most compelling recently — which is usually the branch that just paid, not the branch with the deepest evidence. The result is a book that looks diversified and is in fact chasing: the best-evidenced setup receives ordinary size while a hot variant receives outsized risk on a handful of trades. Because both are profitable in the short run, nothing announces the problem. The Risk Deployment Quality Score exists to name it in a number, before a mean reversion in the hot variant delivers the message the harder way.

The gradient and the gate are different instruments

One decides how much. The other decides where it goes.

These two mechanisms are frequently confused because they both reduce risk, but they operate on different axes and neither can substitute for the other. The gate ladder sets the total pool available to the account based on drawdown from peak — a scalar. The deployment gradient decides how that pool is distributed across branches based on evidence quality — a shape. A compressed gate with an intact gradient is a smaller book allocated correctly. A full gate with a flattened gradient is a large book allocated badly, and it is the more dangerous of the two states precisely because nothing about it looks defensive. Reducing size is easy to notice. Losing the shape of an allocation is not.

Under compression

The gradient survives the gate ladder — compressed, not abandoned.

When gate states tighten the pool, the gradient does not flatten into equal small bets. Proportionality is preserved: the best-evidenced deployments keep their relative priority inside the smaller pool. Defense reduces total exposure; it never reverses the principle that quality is where remaining risk belongs.