Fifty-eight structures. Thirty-seven clear the defined-risk gate.

Defined risk has a narrow, mechanical meaning here: the maximum loss of a position is computable before the order exists, from the structure alone. A call debit spread can lose the debit paid and nothing more. A credit spread can lose the strike width minus the credit received. A long option can lose its premium. None of these numbers depend on a forecast, a volatility model, or an exit executed in time. They are properties of the legs.

That property is the precondition for Gate 1, Radon’s convexity gate. The gate requires gain of at least 2 times loss, on defined-risk structures only, and a ratio needs a denominator. If the maximum loss is unlimited, or contingent on where the underlying opens after a gap, gain over loss is not a number that can be checked at submit. The gate would be reduced to an opinion.

The same denominator drives sizing. Radon sizes positions with fractional Kelly under a hard cap of 2.5 percent of bankroll per position, and Kelly arithmetic starts from the amount at risk. A structure without a computable max loss cannot be sized, only guessed at. Why defined loss makes sizing computable is the subject of its own dossier.

The catalog’s edge case makes the definition precise. The Short Put (Cash-Secured) entry is marked defined with a max loss of strike × 100, and its notes record the collateral requirement: strike × 100 per contract, cash that covers assignment. The worst case, assignment with the stock at zero, is a large number. It is also a known number. Defined risk means known loss, not small loss.

Two disclosures close the ledger. First, an ALLOW from the guard is not a pass through the gates: Gate 1 accepts defined-risk structures only, so an undefined risk profile stops a trade at convexity regardless of what the leg-matching ladder concluded. Second, Gate 4, the no-naked-shorts gate that consumes these guard decisions, is currently disabled by operator policy, with the blocking logic preserved for re-enable. The catalog keeps stating ALLOW and BLOCK either way: the taxonomy records what the logic decides, not whether the operator currently lets it act.

Structure selection is one step in a longer chain: flow signal, regime corroboration, level mapping, then structure, size, verification, and routing. Where structure selection sits in the method is covered in the flow dossier, from dark-pool print to routed order.