The stability branch: why the smallest allocation is not the least governed.

What it is

Controlled participation in additional high-quality opportunity.

Overflow began as the +2 strong-trend extension and remains exactly that: a way to participate in extra flow when the broader system and the volatility environment justify it. The MAE/MFE Lab's branch table describes its expected behavior in four words — controlled supplemental opportunity flow — and the branch's design follows from that description rather than from any ambition about returns. It is capped at 10% of the blend. It is primarily a static-exit branch. Its holds are short. Its friction should be negligible. A structurally healthy Overflow week is boring: small count, small positive R, minimal cost drag.

Why 'stability'

The branch contributes shape, not magnitude.

Calling a 10% allocation the stability branch sounds like a promotion it hasn't earned, until you look at what it actually does to the blend. Overflow's outcomes resolve quickly and deterministically, which means its contribution to weekly R arrives with almost no variance attached. In a week where the trend branches are producing lumpy, trail-dependent outcomes and Normal is doing its steady work, Overflow's contribution is the most predictable component in the mix. That is what a stability branch does — it is not there to move the number, it is there to not move it unpredictably while adding genuine opportunity coverage.

The diversification rule

Participation is tracked because absence is a drift, not a preference.

The Scorecard manual states the purpose plainly: track Trend No-Partial and Overflow participation so the system does not quietly drift away from its fat-tail and diversification rules. Overflow sits on the diversification half of that sentence, and the pairing is instructive — the two branches at opposite ends of the variance spectrum are the two the system watches for disappearance. Both are easy to stop selecting for reasons that feel like discipline. An operator who quietly abandons Overflow is not being conservative; they are running a different profile than the one their blended EV assumes, and the Monthly_Quota_Ledger exists to make that visible across a full year rather than a single flattering month.

• Legend_Constraints carries the governance constants: TNP minimums, Overflow minimums, weekly min/max rules, urgency thresholds.
• Monthly_Audit_Card reports actual counts against those constraints, with urgency flags and recovery triggers.
• The ledger runs the full year — quota is a long-horizon instrument, not a weekly scoreboard.

The misreading

Small allocation, maximum scrutiny — and the asymmetry is deliberate.

The intuitive governance model scales oversight with allocation: the 45% branch gets the most attention, the 10% branch gets the least. Overflow inverts it, and the reason is that Overflow's risk is behavioral rather than financial. Ten percent of the blend cannot damage an account through ordinary underperformance. It can do serious damage as a route around the system's other controls, because it is the branch with the loosest natural identity — 'additional high-quality opportunity' is a description an operator under pressure can stretch to cover almost anything. The scrutiny is calibrated to that failure mode, not to the branch's size.

The boundary

Overflow is not a place to express conviction the other branches wouldn't authorize.

Every branch in MARS has an evidence stack that gates entry into it, and Overflow's looser definition makes it the natural destination for trades that failed those stacks elsewhere. A setup that couldn't clear Trend No-Partial's seven lenses does not become acceptable by being relabeled supplemental flow; a trade taken past the throttle's comfort does not become compliant by arriving under a branch with spare quota. Overflow's identity is opportunity that the system and the volatility environment already justify — the justification is a precondition, not something the branch supplies. Anything else is branch-label laundering, and it corrupts the branch's expectancy estimate along with the account.